Google's AI Built an AI that Outperforms Any Made By Humans (sciencealert.com)

← Back to Stories (view on slashdot.org)

Google's AI Built an AI that Outperforms Any Made By Humans (sciencealert.com)

Posted by EditorDavid on Sunday December 3, 2017 @11:34AM from the rise-of-the-machine-learning dept.

schwit1 quotes ScienceAlert: In May 2017, researchers at Google Brain announced the creation of AutoML, an artificial intelligence (AI) that's capable of generating its own AIs. More recently, they decided to present AutoML with its biggest challenge to date, and the AI that can build AI created a 'child' that outperformed all of its human-made counterparts... For this particular child AI, which the researchers called NASNet, the task was recognising objects -- people, cars, traffic lights, handbags, backpacks, etc. -- in a video in real-time. AutoML would evaluate NASNet's performance and use that information to improve its child AI, repeating the process thousands of times.

When tested on the ImageNet image classification and COCO object detection data sets NASNet was 82.7 percent accurate at predicting images on ImageNet's validation set. This is 1.2 percent better than any previously published results, and the system is also 4 percent more efficient, with a 43.1 percent mean Average Precision (mAP).

235 comments

Min score:

Reason:

Sort:

This all sounds impressive... by MikeDataLink · 2017-12-03 11:36 · Score: 3, Insightful

but every time I research the raw data it becomes very clear these aren't all that smart of AIs. In fact, the term AI is very misleading. They're more like smart scripts. ;-)

--
Mike @ The Geek Pub. Let's Make Stuff!
1. Re:This all sounds impressive... by Anonymous Coward · 2017-12-03 11:44 · Score: 3, Interesting
  
  It can identify if something is a kitten or not with 83.4% accuracy. Sounds impressive until you realize a 3 year old can do this with 99.9% accuracy.
2. Re:This all sounds impressive... by ShanghaiBill · 2017-12-03 11:46 · Score: 5, Informative
  
  but every time I research the raw data it becomes very clear these aren't all that smart of AIs.
  Indeed they are not. This is Weak AI. They are programmed/trained for a specific task, and outside that area of expertise, they generally have no ability at all.
  
  In fact, the term AI is very misleading.
  Only if you watch too many movies. Hollywood uses the term very differently from actual practitioners.
  
  They're more like smart scripts. ;-)
  They are absolutely nothing like "smart scripts", since they aren't smart, and they aren't scripts.
3. Re:This all sounds impressive... by LesFerg · 2017-12-03 11:52 · Score: 2
  
  If the 'parent' AI kept telling the 'child' AI when it was right or wrong, wouldn't it just need to compile it's own database of the identified pictures?
  "Currently we have an average of over five hundred images per node."
  
  --
  If I had a DeLorean... I would probably only drive it from time to time.
4. Re:This all sounds impressive... by Anonymous Coward · 2017-12-03 11:57 · Score: 0
  
  In fact, the term AI is very misleading.
  Only if you watch too many movies. Hollywood uses the term very differently from actual practitioners.
  Even in the industry AI has many different definitions and it will depend who you ask and their area of research as to the answer. (at least it did for the 6 years I was in it after my degree).
5. Re:This all sounds impressive... by Dutch+Gun · 2017-12-03 12:01 · Score: 5, Insightful
  
  I think the term "deep learning" seems a bit better than "AI" for these sorts of very narrowly-defined tasks.
  
  --
  Irony: Agile development has too much intertia to be abandoned now.
6. Re: This all sounds impressive... by Anonymous Coward · 2017-12-03 12:07 · Score: 2, Insightful
  
  I truly don't give a shit what you think or have to say. Neither does anyone.
  And yet OP is rated +5 insightful, and you're rated -1 troll. So yeah. There's that.
7. Re:This all sounds impressive... by ShanghaiBill · 2017-12-03 12:07 · Score: 5, Interesting
  
  If the 'parent' AI kept telling the 'child' AI when it was right or wrong ...
  It doesn't work that way. Each NN learns on its own, using a combination of both labeled and unlabeled data. The parent NN sets "hyper-parameters", such as the number of layers, the size of each layer, the activation function, the convolution size, dropout rate, the learning rate damping factor, the batch size, etc. Then it turns the children NNs loose on the image dataset. It then sees which hyper-parameters lead to better/faster performance, and then applies ML techniques to learn better hyper-parameters.
  None of this is new. What is new, is that Google is now applying this recursively, and using AutoML to design a better AutoML. This is another step toward the singularity.
8. Re:This all sounds impressive... by ceoyoyo · 2017-12-03 12:14 · Score: 2
  
  That would certainly be one way of solving the problem. Except that the actual problem isn't to recognize images you've seen before, it's to recognize ones you *haven't*.
9. Re: This all sounds impressive... by NicknameUnavailable · 2017-12-03 12:16 · Score: 1
  
  So you're conceding your second point from the original post then? Or did you just forget you said "Neither does anyone?"
10. Re: This all sounds impressive... by NicknameUnavailable · 2017-12-03 12:17 · Score: 0
  
  Nobody fucking cares. I sure as hell don't care what you have to say. And I promise you that no one else does, either.
  I did, and I enjoy that that in itself makes you wrong, Google shill.
11. Re:This all sounds impressive... by MrL0G1C · 2017-12-03 12:21 · Score: 1
  
  And real life isn't a staged photo, it moves non-linearly in 3-dimensinal space with varying light conditions. Lets have some real tests, not carefully taken photos of cats and dogs against easy backgrounds.
  
  --
  Waterfox - a Firefox fork with legacy extension support, security updates and better privacy by default.
12. Re:This all sounds impressive... by Anonymous Coward · 2017-12-03 12:21 · Score: 1
  
  Can I get a copy? I for one am sick of filling out captchas.
13. Re:This all sounds impressive... by MrL0G1C · 2017-12-03 12:28 · Score: 2
  
  Or how about this, the AI does some "hidden object puzzles", I can do those with a very high degree of accuracy, I bet AI would fail hard.
  
  --
  Waterfox - a Firefox fork with legacy extension support, security updates and better privacy by default.
14. Re: This all sounds impressive... by Anonymous Coward · 2017-12-03 12:31 · Score: 0
  
  Well it sounds like you've given up on life.
  Just kys already.
15. Re:This all sounds impressive... by ShanghaiBill · 2017-12-03 12:48 · Score: 5, Interesting
  
  It can identify if something is a kitten or not with 83.4% accuracy.
  No. It can look at an image and correctly classify it into THOUSANDS of categories, only one of which is "kitten". It was 82.7% accurate at this. If it was trained to only distinguish "kitten" from "not-kitten", it would, of course, be far more accurate.
  
  a 3 year old can do this with 99.9% accuracy.
  A 3 year old requires 3 years of training. This system can learn in hours.
16. Re:This all sounds impressive... by duke_cheetah2003 · 2017-12-03 13:07 · Score: 1
  
  but every time I research the raw data it becomes very clear these aren't all that smart of AIs. In fact, the term AI is very misleading. They're more like smart scripts. ;-)
  I think it's more along the lines of: OOOOH we made something can do ONE of multitude of things the human brain can do, and therefore it's intelligent.
  Poppycock. Y'all got a really impressive image recognition system there, but you know, just being able to tell what something is by looking at is a very very minuscule piece of what human intelligence is.
  Now if they can expand this into other areas of human intelligence, and make it all come together to form some sort of 'awareness,' yeah, I dunno, they got a long way to go. A very long way.
17. Re:This all sounds impressive... by Anonymous Coward · 2017-12-03 13:32 · Score: 0
  
  Lovely, the three year old is pulling from thousands and still do it more accurately too. And the three year old did have three years old to learn to do it, but it also had to program itself. It hasn't had thousands of people programming it for 60 years.
18. Re:This all sounds impressive... by Anonymous Coward · 2017-12-03 13:59 · Score: 2, Insightful
  
  It can identify if something is a kitten or not with 83.4% accuracy. Sounds impressive until you realize a 3 year old can do this with 99.9% accuracy.
  Sounds insightful, until you realize the entire purpose of AI research is to create software artificially to reproduce feats of human intelligence.
  That fact sorta disqualifies the 3 year old :P
  Also your comment is pretty close to implying that since our first attempts at making AI haven't had a 99.9% success rate right off the bat, that they are not impressive enough to bother improving.
  Giving up has a 100% success rate of never making anything better, which is also against the purpose of trying to improve something.
19. Re: This all sounds impressive... by Anonymous Coward · 2017-12-03 14:06 · Score: 0
  
  But we have seen no evidence of your research or the size of your sample groups.
  When you say "no one", does the sample group you used to reach this conclusion actually include more people than, say, just yourself?
  Also, your lack of transparency with your research material leads this panel of experts to suspect that you are actually just full of shit and mad at the world in general, which thereby renders your study biased and unreliable. We put it to you that perhaps there may be a certain demographic which does in fact care what other people post in this public forum, a demographic which may have been either accidentally or intentionally excluded from your research group, and possibly contains everybody else who is not you.
20. Re: This all sounds impressive... by Anonymous Coward · 2017-12-03 14:08 · Score: 0
  
  An AI is a huge set of algorithms. That's all it's been since Turing's day, even if the hardware has changed. If you know what an AI actually is, please feel free to enlighten all of us. Otherwise, you all need to stop assuming it's some matrix/science fiction test-tube brain creature. It's a intelligence agency, military Grant goldmine scratching the backs of Fortune 500 companies and vise versa while the payed Google publicist crowd on Slashdot bullshits the plebs into thinking it's something we need and wonderful. More condoms and less cell phones. There, I solved world hungar, war, and debt. Funny how the AI techy programmer countries like Korea and India can't even afford to print money anymore. ;)
21. Re:This all sounds impressive... by ag0ny · 2017-12-03 14:09 · Score: 1
  
  a 3 year old can do this with 99.9% accuracy.
  And when he's done with the training he's already 6, so...
  oh, wait...
  
  --
  My site
22. Re: This all sounds impressive... by Anonymous Coward · 2017-12-03 14:10 · Score: 0, Funny
  
  Chris, such wild mood swings aren't good for someone in your physical condition.
23. Re:This all sounds impressive... by HiThere · 2017-12-03 14:15 · Score: 1, Flamebait
  
  Saying they aren't smart depends on your definition of smart.
  And the GP criticizing them as scripts caused me to wonder how much of what he does could be considered scripted.
  Most discussion of AI that isn't extremely technical is so full of fuzzy terms that it's almost meaningless. What you can say is what it does:
  This thing learns to do object recognition to a reasonable quality rather more quickly than prior ones did...and it was written by a program designed to write other programs. I suspect that it can't do anything else, but that, alone, is impressive.
  FWIW, I don't believe that "general intelligence" exists. I believe that what we think of as intelligence is composed of lots and lots of narrow AIs, some of which specialize in routing problems to the appropriate "program", etc. And from any intelligence I'm rather certain that GÃdel's Incompleteness theorem implies that there are classes of problems that it cannot solve. This isn't a direct implication, as no system is going to have an infinite amount of storage, whatever meaning of infinite you choose to use, but I believe it's an indirect implication. Certainly everyone I've ever met has a limited ability to handle nested recursive problems.
  Simple example (Many people can solve this, but it displays the problem):
  You see a man standing beside a picture and he says to you:
  "Brothers and sisters have I none, but that man's father is my father's son".
  
  --
  
  I think we've pushed this "anyone can grow up to be president" thing too far.
24. Re: This all sounds impressive... by Anonymous Coward · 2017-12-03 14:15 · Score: 0
  
  Actually, I think almost everyone here but you cares. The only people we don't care about are trolls who either spam or carry out personal vendettas rather than engaging in thoughtful discussion.
25. Re:This all sounds impressive... by HiThere · 2017-12-03 14:19 · Score: 2
  
  You need a better memory. There was a time when Eliza was commonly called an AI program. Certainly Arthur Samuel's Checkers program was. Most modern things touted as AIs are considerably advanced over either of those.
  
  --
  
  I think we've pushed this "anyone can grow up to be president" thing too far.
26. Re: This all sounds impressive... by AK+Marc · 2017-12-03 14:20 · Score: 0
  
  Says the A/C.
  
  The progress in AI is in re-defining it. We are no closer to strong AI than when we invented the term AI. We just have more classes of weak AI to claim victory over.
  
  --
  Learn to love Alaska
27. Re: This all sounds impressive... by Anonymous Coward · 2017-12-03 14:20 · Score: 0
  
  Someone needs to relax and play some Universal Paperclips in celebration of the coming singularity. ;)
28. Re:This all sounds impressive... by AK+Marc · 2017-12-03 14:32 · Score: 1
  
  FWIW, I don't believe that "general intelligence" exists.
  So human intelligence is also a sum of narrow AIs? All that means is that your definitions don't match anyone else's.
  
  --
  Learn to love Alaska
29. Re:This all sounds impressive... by Oligonicella · 2017-12-03 14:34 · Score: 1
  
  A 3 year old requires 3 years of training. This system can learn in hours.
  *Very* deceptive comparison.
30. Re:This all sounds impressive... by slickwillie · 2017-12-03 15:02 · Score: 5, Funny
  
  but every time I research the raw data it becomes very clear these aren't all that smart of AIs. In fact, the term AI is very misleading. They're more like smart scripts. ;-)
  So the child AI is a script kiddie.
31. Re:This all sounds impressive... by Anonymous Coward · 2017-12-03 15:42 · Score: 0
  
  > This is another step toward the singularity.
  Our father who art in heaven, hallow'd be thy name. Thy kingdom come, thy will be done, on earth as it is in heaven.
32. Re: This all sounds impressive... by Anonymous Coward · 2017-12-03 15:59 · Score: 0
  
  Never. The uncanny valley prevents it. Just when AI gets idiotic enough, you notice a shred of logic or coherency and it completely throws off the illusion of idiocy the AI tries to project.
  In conclusion, Artificial Intelligence is inevitable as long as people and computers exist, but Artificial Idiocy will never surpass the specially retarded over watering power lifter affiliate link spammer of Slashdot. Crumpty Dumpty will post with his last bit of smart phone battery after the chair collapses under his cheeto weight, then be found months later after neighbors call the police after noticing the drop in putridity of the air.
33. Re:This all sounds impressive... by carbs77 · 2017-12-03 16:01 · Score: 2
  
  The real question is, how did it fair with hotdog vs. not hotdog?
34. Re:This all sounds impressive... by gweihir · 2017-12-03 16:04 · Score: 1
  
  That is why in actual AI research this is called "automation", not AI. For a more hype-friendly stance, "weak AI" (the AI without "I") is also in use. The converse, "strong AI" or "true AI" (i.e. actual machine intelligence) is not available and it is unclear whether it can be created.
  These are just statistical classifiers. They are about as intelligent as a reference book or a loaf of bread. For example, you could replace the image recognition thing with just a large collections of templates normalized in several ways and the do simple cross-correlation. That would be exceptionally (and prohibitively) slow, but likely have better results.
  So in sum, these things are essentially pre-calibrated sensors. They do not think or understand anything. Still very useful, but the term "AI" for this is basically a lie.
  
  --
  Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
35. Re: This all sounds impressive... by gweihir · 2017-12-03 16:06 · Score: 2
  
  Not all real intelligence is actually be used by its owner. In fact, most people rarely use their intelligence to understand things. They rather stick to their misconceptions. Case in point.
  
  --
  Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
36. Re:This all sounds impressive... by gweihir · 2017-12-03 16:07 · Score: 1
  
  Indeed. And this only works with some very restrictive border-conditions.
  
  --
  Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
37. Re:This all sounds impressive... by gweihir · 2017-12-03 16:10 · Score: 2
  
  but every time I research the raw data it becomes very clear these aren't all that smart of AIs.
  Indeed they are not. This is Weak AI. They are programmed/trained for a specific task, and outside that area of expertise, they generally have no ability at all.
  Indeed. And inside that task, they are very restricted as well. The thing to remember is that weak AI has absolutely no understanding or concept of what it is doing. It just sums up details and gets a number. If cleverly done, it can perform apparently impressive feats like this one here, but it is not intelligent. Hence it is better called by its traditional name "automation".
  
  --
  Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
38. Re:This all sounds impressive... by Anonymous Coward · 2017-12-03 16:12 · Score: 0
  
  No, it's had billions of people contributing to its development over our entire genetic history.
39. Re:This all sounds impressive... by gweihir · 2017-12-03 16:15 · Score: 2
  
  Well, sort-of. Weak AI can also be done in other ways. But I recently learned that "deep learning" is basically what you do when you do not have a good model of the problem-space. When you do have that model, other approaches are superior. But since creating models is a real hard-core expert task and expensive, the potential of deep learning is basically to do thing somewhat worse than an expert but a lot cheaper. That is, if it works for a problem. For most problems it does not work.
  
  --
  Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
40. Re:This all sounds impressive... by Anonymous Coward · 2017-12-03 16:15 · Score: 0
  
  Man, I can't wait.
  I hope I live to see it.
  I hope I live long enough for immortality to be invented as a result of it, and that I manage to get in on the deal.
  We have mountains of philosophy, religion, and pop-wisdom about how death is not so bad, we can cope with it, all natural and we are probably better off and blah blah blah....
  It's all sour grapes. Death is a damned insult. Once we overcome it, watch our entire culture change its tune. We will turn on a dime. We will remember our ancestors' trials with gratitude and horror.
41. Re: This all sounds impressive... by Anonymous Coward · 2017-12-03 16:17 · Score: 0
  
  Realize what you're saying though.
  The bar keeps moving. In a few years you'll be saying "hahaha it it identified this 1 thing wrong"
  but identified 999 other things correctly.
42. Re:This all sounds impressive... by gweihir · 2017-12-03 16:18 · Score: 1
  
  Actually, when it is obvious what it is, it could be argued that humans use biological automation and not intelligence to recognize the image. Most things humans do do not actually involve intelligence. Intelligence is a fall-back mechanism when things become more difficult, and one of the great unknown questions is how humans decide whether things require intelligence or not. (Many humans avoid using intelligence like the plague though.) Only when things become trickier and actual thinking is involved do humans show their true strength and machines fail miserably.
  
  --
  Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
43. Re:This all sounds impressive... by gweihir · 2017-12-03 16:20 · Score: 1
  
  ELIZA only ever was a natural language processing computer program. It was specifically created to demonstrate that you can fake being intelligent without any intelligence at all.
  
  --
  Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
44. Re: This all sounds impressive... by Psion · 2017-12-03 16:51 · Score: 1
  
  One can only hope.
45. Re: This all sounds impressive... by Anonymous Coward · 2017-12-03 17:10 · Score: 0
  
  Does anyone care whether you care?
46. Re: This all sounds impressive... by Wycliffe · 2017-12-03 17:11 · Score: 1
  
  A 3 year old requires 3 years of training. This system can learn in hours.
  So give the computer 3 years of training on the fastest supercomputer available. A 3 year old would still be able to outperform it.
47. Re:This all sounds impressive... by ShanghaiBill · 2017-12-03 17:13 · Score: 1, Flamebait
  
  The thing to remember is that weak AI has absolutely no understanding or concept of what it is doing.
  That is a meaningless assertion. It depends entirely on how you define "understanding" and "concept". The chemicals and neurons that make up the human brain also don't "understand" what they are doing.
  
  It just sums up details and gets a number.
  That is also what biological neurons do.
  
  but it is not intelligent.
  Define "intelligent". Is a human intelligent? What about a monkey? A dog? An insect?
  
  Hence it is better called by its traditional name "automation".
  "Automation" is used to describe assembly lines, not systems that can learn and adapt.
48. Re: This all sounds impressive... by Anonymous Coward · 2017-12-03 17:19 · Score: 0
  
  Yeah, the best way to show how much shit you don't give is to reply and tell people.
  Don't forget to do it twice or thrice to make sure they understand how much you really don't care.
49. Re: This all sounds impressive... by Anonymous Coward · 2017-12-03 17:27 · Score: 0
  
  3 year old can not do that. AI beat adult humans in image recognition few years ago already. You need to understand what kind of images it can't classify. Those images are so bad that humans can't do it either unless they know what they are looking for and even then they are just guessing.
50. Re:This all sounds impressive... by gweihir · 2017-12-03 18:13 · Score: 1, Flamebait
  
  You seem to be lacking actual intelligence as well. Well, more likely you have it but are not using it. A common failure in humans. Nobody knows how or whether human brains create intelligence. The closer we look, the less likely that seems though. All we have is an interface observation. No, not even fMRI gives us more. And yes, that is the scientific state-of-the-art. What you say is belief (and a stupid one), not science.
  Incidentally, "automation" is exactly the right term. These systems cannot "learn" or "adapt" in the way these terms are commonly understood.
  
  --
  Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
51. Re:This all sounds impressive... by HiThere · 2017-12-03 18:22 · Score: 1
  
  Yep. That was where I started.
  
  --
  
  I think we've pushed this "anyone can grow up to be president" thing too far.
52. Re:This all sounds impressive... by HiThere · 2017-12-03 18:25 · Score: 2
  
  I understand that the author of Eliza didn't think it was an AI program, but it was commonly called one anyway, despite being an intentional attempt to deny the idea. So calling similar things an AI today isn't redefining the term. Abusing it, perhaps, but not redefining it.
  If you want a redefined term look at what gets called a robot today and contrast it with any use of the term before 1960.
  
  --
  
  I think we've pushed this "anyone can grow up to be president" thing too far.
53. Re:This all sounds impressive... by religionofpeas · 2017-12-03 19:36 · Score: 1
  
  ELIZA only ever was a natural language processing computer program
  No it wasn't. It was just manipulating bits in memory. There was no concept of language.
54. Re: This all sounds impressive... by Anonymous Coward · 2017-12-03 19:50 · Score: 0
  
  Sadly,
  Your opinion counts FAR less than that of the group that you are attempting to affect.
55. Re:This all sounds impressive... by serviscope_minor · 2017-12-03 20:09 · Score: 3, Informative
  
  And real life isn't a staged photo, it moves non-linearly in 3-dimensinal space
  Neither is Image NET.
  Lets have some real tests, not carefully taken photos of cats and dogs against easy backgrounds.
  How about... ImageNET.
  Seriously, you can just go and download (bits of*) ImageNet very easily. It's a large database of photos drawn from the internet taken by people which were labelled after the fact. There's not much if any careful staging in it.
  [*]It's huge, you probably only want a bit of it. Just the list of image URLs is 300 meg.
  
  --
  SJW n. One who posts facts.
56. Re: This all sounds impressive... by Njorthbiatr · 2017-12-03 20:11 · Score: 1
  
  If that's true then you're a huge set of algorithms. Neural networks are just digital brains.
57. Re:This all sounds impressive... by serviscope_minor · 2017-12-03 20:14 · Score: 4, Insightful
  
  Indeed they are not. This is Weak AI. They are programmed/trained for a specific task, and outside that area of expertise, they generally have no ability at all.
  Yep
  In fact, the term AI is very misleading.
  I disagree.
  No one[*] is vlaiming these techniques ar intelligent. However what they are doing is solving a task which previously required human intelligence to solve, hence the name "artificial intelligence".
  Compare to a lot of computation, where the steps are simple, and it's been widely known for a while that simple sheer quantity of them rather than intelligence is needed.
  It's a pretty arbitrary name, but it's not actually unreasonable.
  [*]There's always one idiot. Let's ignore him.
  
  --
  SJW n. One who posts facts.
58. Re:This all sounds impressive... by serviscope_minor · 2017-12-03 20:18 · Score: 1
  
  I think the term "deep learning" seems a bit better than "AI" for these sorts of very narrowly-defined tasks.
  Deep learning is about the method used to solve a task, not the task being solved.
  Here's the copypasta of what I wrote last time about what deep learning is:
  Deep learning is not especially well defined, but I've seen several competing/complementary definitions.
  1. A neural net (much) greater than 3 layers deep. A sufficiently wide 3 layer net has enough capacity to run any function, so a lot of ANN learning focussed on these in the past. Turns out having lots more layers makes training much more tractable (particularly with stochastic gradient descent an batchnorm).
  2. Convolutional nets (where you're basically learning convolutional kernels, which saves a huge number of parameters compared to a normal net especially if you have many layers) with many layers.
  3. Something which learns the low-level features in the same optimization as everything else. Traditional ML algorithms were often structured as a feature extraction stage which takes the data and extracts some features in a human designed manner. Then applies ML, but if the ML can't optimize the right loss, you go for the closest loss you can, then get the thing you want with post-processing.
  A nice example would be the Viola-Jones face detector [wikipedia.org]. The features are a bunch of zero mean box filter combinations applied to an image, combined with a threshold, each one giving a different binarisation of the image. Those were hand designed. a modified Adaboost[*] is then learned to to get a good selection and weighting of the features to give a pixelwise classification of the image. You want a bounding box, so the final stage is to extract a bounding box from the binarization.
  The "problems" with that are that there's nothing to say those features are optimal, and the pixelwise loss is the wrong thing to optimize. A deep system takes pixels in and spits out a bounding box (or several). The point is you can compute derivatives with respect to all the stages so you optimize everything agains the loss you're actually interested in.
  [*] They actually use a cacade (degenerate decision tree) with a biased adaboost classifier at each node. Either way it's an ML algorithm with largely the same properties.
  
  --
  SJW n. One who posts facts.
59. Re: This all sounds impressive... by religionofpeas · 2017-12-03 20:19 · Score: 1
  
  A 3 year old would still be able to outperform it
  Really ? Have you seen any test results of a 3 year old on ImageNET challenges, or are you just making it up ?
60. Re:This all sounds impressive... by religionofpeas · 2017-12-03 20:25 · Score: 5, Interesting
  
  Try this: https://www.youtube.com/watch?...
  AI outperforms humans. There are some tricky cases after timestamp 2:42. You may want to try them for yourself.
61. Re: This all sounds impressive... by Anonymous Coward · 2017-12-03 20:29 · Score: 0
  
  In some tasks in some ways. And yes, I have done research in the area. Scene analysis is complex. A particular advantage with AI is speed and indefatigableness, more than accuracy.
62. Re:This all sounds impressive... by religionofpeas · 2017-12-03 20:34 · Score: 1
  
  So human intelligence is also a sum of narrow AIs? All that means is that your definitions don't match anyone else's.
  It matches mine. Actually you could go a step further: human intelligence is also a sum of stupid parts.
63. Re:This all sounds impressive... by religionofpeas · 2017-12-03 20:36 · Score: 1
  
  Nobody knows how or whether human brains create intelligence.
  If you have no idea how it works, how come you feel qualified to make strong statements about it ?
64. Re: This all sounds impressive... by religionofpeas · 2017-12-03 22:10 · Score: 2
  
  The progress in AI is in re-defining it.
  The progress is in getting better results, and solving increasingly challenging problems that could not be solved before.
65. Re:This all sounds impressive... by Kiuas · 2017-12-03 23:01 · Score: 1
  
  but every time I research the raw data it becomes very clear these aren't all that smart of AIs. In fact, the term AI is very misleading.
  No, no it isn't misleading, people just associate the word with something it doesn't mean. Intelligence is a scale, not a binary thing. Biological systems have differing amounts of intelligence within them as well: a fish for example is not very smart by human/mammal standards but it's instantly obvious that a a fish has way more intelligence than say, bacteria. Likewise rats are not hugely intelligent compared to apes, but a rat can be trained and is clearly more intelligent than a fish. And so on and so forth. But when we start to talk about machines and machine learning people forget about this and when they read that a system is artificially intelligent they falsely assume it to mean either 'human level (general) intelligence' and/or 'consciousness' neither of which are required for a system to be intelligent.
  Intelligence is simply information processing in physical systems. An algorithm that can correctly classify items from a moving image in real time to thousands of different categories is way smarter than all non-human animals, and even smarter than small children in said task. Sure, that doesn't make it smarter than adults overall (nor smarter than the fish when it comes to surviving in a marine ecosystem), but it also doesn't make it not intelligent.
  The goal posts are moving when it comes to AI: if you told people 30 years ago that we now have cars capable of driving themselves autonomously amidst traffic with normal human drivers and avoid accidents successfully, that'd have been instantly classified as a type of AI - and deemed impossible by most people. But now since people's understanding of what machines are capable of doing that autonomous car is suddenly 'not very smart' (even though these systems are by and large better at driving that the vast majority of humans) because while it can drive it cannot also debate with me about Plato while doing so.
  
  They're more like smart scripts. ;-)
  You mean they're more like scripts with narrow scope of (artificial) intelligence. Same goes for chess engines and alphaGo for example: they're systems which can only do 1 thing really, but they're capable of doing that one thing better than the best human players, making them more intelligent in those fields than any human, even though they're 'dumb' in every other aspect. Intelligence and general intelligence are 2 entirely different things.
  
  --
  "It is the business of the future to be dangerous" -Alfred North Whitehead
66. Re: This all sounds impressive... by Zero__Kelvin · 2017-12-03 23:17 · Score: 1
  
  Every time a subject like this comes up the now beaten to death subject of calling it AI or not AI gets hashed out again. It reminds me of the "It's not a cloud, it's somebody else's computer" debates of old. The industry has decided that the term for these systems is AI regardless of anyone's philosophical sensibilities, so can we agree that everyone has their own take on it and get on with the kind of discussions that actually add value in context now?
  
  --
  Guns don't kill people; Physics kills people! - John Lithgow as Dick Solomon on Third Rock From The Sun
67. Re:This all sounds impressive... by Anonymous Coward · 2017-12-03 23:55 · Score: 0
  
  > Many humans avoid using intelligence like the plague though.
  I wonder if this hasn't something to do with saving the energy - thinking takes a lot of sugar, which isn't readily available in the jungle, so probably our ancestors evolved to use thinking only when absolutely needed.
  We're not in the jungle anymore, but our bodies are not evolved (yet) to work well when sugar is cheap and readily available everytwhere.
68. Re:This all sounds impressive... by jan_koch · 2017-12-04 00:18 · Score: 2
  
  Thank you for one of the few comments in this thread that actually deals with what this is (as opposed to what it isn't, i.e. human-level AI).
  I would like to add that hyperparameter tuning is _not_ a trivial part of programming a machine learning model, therefore this IS something rather interesting. It lowers the effort needed to do something interesting with machine learning, and therefore makes machine learning much more accessible to non-experts.
  However, the tasks in machine learning that still require humans need a much more flexible sort of intelligence: a) asking the right question, i.e. determining the variables to be predicted and b) finding the right input parameters (or independent variables) that will help answer the question.
69. Re:This all sounds impressive... by hraponssi · 2017-12-04 00:29 · Score: 1
  
  You are probably right, although I would rather say they are "successfully" applying ML to design ML. Must have been quite a few tries over time, as the approach of ML to ML seems quite obvious. But this group (and some other of course) finally seems to get many things working nicely. Looking at it all it does not seem all that complicated, so I guess there must be due to some general advances that enable all this to work better now. Not that I could design any of this, but I am sure plenty of smarter than me people have looked at all this many times..
70. Re:This all sounds impressive... by religionofpeas · 2017-12-04 00:42 · Score: 2
  
  It can identify if something is a kitten or not with 83.4% accuracy. Sounds impressive until you realize a 3 year old can do this with 99.9% accuracy.
  How many 3 year olds can tell the difference between a komondor and a bouvier des flanders ? Or would they simply classify both as "dog" ?
  Here you can test yourself on some of these images:
  http://cs.stanford.edu/people/...
  Try the hard ones.
71. Re:This all sounds impressive... by Anonymous Coward · 2017-12-04 01:06 · Score: 0
  
  However what they are doing is solving a task which previously required human intelligence to solve, hence the name "artificial intelligence".
  Therefore, a calculator is capable of "AI".
72. Re:This all sounds impressive... by Anonymous Coward · 2017-12-04 01:21 · Score: 0
  
  An AI that can recognize KITTIES!!??! The internet is saved!
73. Re:This all sounds impressive... by gweihir · 2017-12-04 01:26 · Score: 1
  
  Are you functionally illiterate? Because your question seems to indicate you are as that is not the statement I made.
  
  --
  Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
74. Re:This all sounds impressive... by gweihir · 2017-12-04 01:30 · Score: 1
  
  > Many humans avoid using intelligence like the plague though.
  I wonder if this hasn't something to do with saving the energy - thinking takes a lot of sugar, which isn't readily available in the jungle, so probably our ancestors evolved to use thinking only when absolutely needed.
  We're not in the jungle anymore, but our bodies are not evolved (yet) to work well when sugar is cheap and readily available everytwhere.
  As far as I know, it is the energy use, the being distracted while thinking (and unaware of potential danger) and the unused time. After all, daylight is precious when you have no artificial light. There clearly is a lot of relatively dumb automation in human bodies.
  
  --
  Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
75. Re:This all sounds impressive... by gweihir · 2017-12-04 01:33 · Score: 2
  
  Computers have no concept of anything today. So with your definition, there is no NLP. Hence that definition does not seem to make much sense.
  
  --
  Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
76. Re:This all sounds impressive... by Anonymous Coward · 2017-12-04 01:44 · Score: 0
  
  I would not call alphaGo "intelligent", because all it is is a very specialized brute-force calculator combined with a black-box lookup table in the form of a neural net in order to speed up calculations.
  For me at least, the system should have some representation of "concepts" in order to be considered intelligent. A 5 year old kid can do sums, and a calculator can do sums, but only one of them understands the concept of what a sum is.
  To say that alphaGo is more intelligent than say a person is a badly formulated comparison.
77. Re: This all sounds impressive... by Anonymous Coward · 2017-12-04 01:55 · Score: 0
  
  No they aren't. If they were, they could do the same things, and they can't.
  The digital "neurons" have only a very crude resemblance to the behavior of a real neuron.
78. Re:This all sounds impressive... by religionofpeas · 2017-12-04 02:37 · Score: 1
  
  the system should have some representation of "concepts" in order to be considered intelligent
  It does. AlphaGo has representations of Go concepts. It can look at a board, and immediately get a feel for who is better, and what would be good places to play next. It internally represents concepts like space, territory, initiative, and probably a few that human players don't even have names for.
  It's also not a brute-force calculator. It only tries out a very small subset of all possible moves, namely only the ones that "pop out" as good candidate moves.
79. Re:This all sounds impressive... by thomst · 2017-12-04 02:52 · Score: 1
  
  https://slashdot.org/~carbs77 inquired:
  
  The real question is, how did it fare with hotdog vs. not hotdog?
  (FTFY)
  But, seriously, somebody with points needs to mod parent +1 Funny, because Silicon Valley's "Not Hotdog" is a real thing - and it's now available for Android, too ...
  
  --
  Check out my novel.
80. Re:This all sounds impressive... by Anonymous Coward · 2017-12-04 02:53 · Score: 0
  
  That was my first thought too...
81. Re:This all sounds impressive... by Dorianny · 2017-12-04 03:38 · Score: 2
  
  Outperforming humans in tasks specifically designed to disadvantage humans by employing well known optical illusions doesn't sound all that impressive
82. Re: This all sounds impressive... by Anonymous Coward · 2017-12-04 03:49 · Score: 0
  
  Actual your understanding of AI is seriously outdated.
  Deep learning networks are not clever scripts.
  The programmers donâ(TM)t even know how It works. They just know that if you arrange the networks in a certain way it gives better results and they try and error until they find some efficient setup for the problem at hand. They know as much about the network as how a neuro surgeon know about your brain. Neuro surgeon donâ(TM)t really know how your brain work...they know where to cut to get rid of problems and how to cut to not kill you.
  You should do a bit of reading up.... the latest generation of AI is getting closer to real AI than ever before. Itâ(TM)s just a matter of time that computing power will allow more and more networks to chain and cross feed each other to create more complex behaviours beyond doing that 1 thing really well.
83. Re: This all sounds impressive... by Immerman · 2017-12-04 03:59 · Score: 4, Informative
  
  A strong claim. How exactly do you measure the distance between existing AI and strong AI, before strong AI is developed?
  Personally, I suspect that once strong AI is developed, there's a fair chance we'll see modern "neural" networks as a step in the right direction. After all we know the basic strategy is sound - a much more sophisticated version of it is driving our own minds.
  For comparison though - In 2015 Digital Reasoning built the largest neural network in the world, at 160 billion parameters. I'm guessing a "parameter" is a weighted connection between "neurons", and thus roughly analogous to a single synapse in an organic brain, of which a human brain has 100-1000 trillion. So, even barring any "secret sauce" we haven't yet figured out in how "processing nodes" interconnect, our most advanced AIs have less than 0.1% of the processing potential of a human brain. Even a mouse brain apparently averages almost a billion synapses per mm^3, so in the neighborhood of 400 billion synapses for a common house mouse.
  So, currently our most advanced AIs have only a fraction of the processing potential of a mouse brain, and that's before you even consider the fact that continuous asynchronous signalling is likely far more information-dense than a clocked AI "neural network", or the fact that individual biological neurons actually do a fair amount of internal processing and data retention, rather than being "dumb switches" as they are in modern AIs.
  Really hard to tell how the software and strategies compares, when your hardware is underpowered by several orders of magnitude.
  
  --
  --- Most topics have many sides worth arguing, allow me to take one opposite you.
84. Re:This all sounds impressive... by mesterha · 2017-12-04 04:00 · Score: 1
  
  The parent NN sets "hyper-parameters", such as the number of layers, the size of each layer, the activation function, the convolution size, dropout rate, the learning rate damping factor, the batch size, etc. Then it turns the children NNs loose on the image dataset. It then sees which hyper-parameters lead to better/faster performance, and then applies ML techniques to learn better hyper-parameters.
  
  None of this is new. What is new, is that Google is now applying this recursively, and using AutoML to design a better AutoML. This is another step toward the singularity.
  
  Seems they are exposing/solving a dirty secret of deep learning. Yes, one can claim the network is learning features, but to get the best performance takes a human to design the structure to capture those features. However, I wonder if they are exploiting the massive amounts of data Google can access. One of the hard aspects of ML is the statistical bias given a limited amount of data relative to the set of relevant concepts. If they are coming up with ways of helping these networks with their own data, it's not surprising the AutoML technique works. Seems they are doing something sophisticated since they are using reinforcement learning, but their webpage is light on details. I'd like to see if they are beating humans given a equal set of training data. (Of course, that training data yields more biased results over time...)
  
  --
  
  Chris Mesterharm
85. Re:This all sounds impressive... by Immerman · 2017-12-04 04:08 · Score: 1
  
  Of course the catch is that the singularity has nothing to do with human immortality.
  Biological immortality is likely to be both an extremely difficult challenge, and rather uninteresting to synthetic minds.
  And even if AI hardware is capable of hosting an "uploaded" human mind intact - that doesn't do anything for *you*. Having an immortal mind-twin is unlikely to make you feel any better about your own impending demise - and that's assuming they wouldn't have to kill you in order to map your brain in the first place.
  
  --
  --- Most topics have many sides worth arguing, allow me to take one opposite you.
86. Re: This all sounds impressive... by ranton · 2017-12-04 04:25 · Score: 1
  
  The progress in AI is in re-defining it. We are no closer to strong AI than when we invented the term AI. We just have more classes of weak AI to claim victory over.
  And yet since its inception the field of AI has heavily focused on weak AI with strong AI relegated mostly to television and films. It is detractors of AI research who are trying to redefine the term to mean creating a human level intelligence instead of simply algorithms which can produce human-like results.
  
  --
  -- All that is necessary for the triumph of evil is that good men do nothing. -- Edmund Burke
87. Re: This all sounds impressive... by religionofpeas · 2017-12-04 04:25 · Score: 2
  
  In 2015 Digital Reasoning built the largest neural network in the world, at 160 billion parameters. I'm guessing a "parameter" is a weighted connection between "neurons", and thus roughly analogous to a single synapse in an organic brain,
  I think it's a bit more complicated. Natural brains are very slow. I think the fastest neurons run at about 1 kHz, whereas artificial networks run at multiple GHz. Also, real brains have a lot of duplicated circuitry. Simple example is image processing for your left and right eyes. An artificial net can re-use the same parameters for different parts of the data.
  I think a better method is to compare basic operations per second, i.e. number of synapses multiplied by "refresh rate"
88. Re: This all sounds impressive... by Immerman · 2017-12-04 04:29 · Score: 1
  
  Probably even more significant than the "neuron" shortcomings, are the architectural ones.
  AI neural networks are generally arranged in layers - feed one layer the raw input, then use its output as the input to layer 2, whose output is used as the input to layer 3, and so on and so forth through as many layers as you want/need to get the outputs you desire.
  Contrast that to an organic brain, where everything is a jumbled interconnected mess with lots of feedback. Considering the incredible power of feedback in even simple electronic and mechanical systems, expecting a layered neural network to behave even remotely similarly to an organic brain is ludicrous.
  
  --
  --- Most topics have many sides worth arguing, allow me to take one opposite you.
89. Re:This all sounds impressive... by religionofpeas · 2017-12-04 04:32 · Score: 1
  
  It just shows a table with several objects, and asks questions like "What number of other things are there of the same material as the green cube ?"
  There is no optical illusion.
90. Re:This all sounds impressive... by religionofpeas · 2017-12-04 04:35 · Score: 1
  
  I wonder if they are exploiting the massive amounts of data Google can access
  No, they are using a standardised dataset, just like everybody else.
91. Re:This all sounds impressive... by Immerman · 2017-12-04 04:36 · Score: 1
  
  > it is unclear whether it can be created.
  Not at all - we have countless billions of examples of electro-chemical general-purpose "strong" intelligences wandering the planet proving that it can be done. The only question is if they can be recreated using current hardware and techniques. Personally I suspect one of the biggest shortcoming of current "deep learning" strategies is the layered design - organic brains are a jumbled mess of interconnected neurons with an enormous amount of feedback. Without feedback you can't even create standing waves, which provide one of the few clues we have into how organic brains function.
  
  --
  --- Most topics have many sides worth arguing, allow me to take one opposite you.
92. Re: This all sounds impressive... by religionofpeas · 2017-12-04 04:41 · Score: 1
  
  AI neural networks are generally arranged in layers
  
  https://en.wikipedia.org/wiki/...
93. Re: This all sounds impressive... by Anonymous Coward · 2017-12-04 05:03 · Score: 0
  
  the kind of discussions that actually add value in context now?
  You should come up with a "Top 10 Zero__Kelvin's Value Added Discussion Subjects" and submit it to the editors. As a much revered Slashdot denizen, your advice will not be ignored.
94. Re:This all sounds impressive... by Anonymous Coward · 2017-12-04 05:12 · Score: 0
  
  > An AI that can recognize KITTIES!!??! The internet is saved!
  You could probably get just as good results with this algorithm:
  * If the image was posted on Twitter/Facebook/Imgur/Pinterest
  ** Flip a coin to determine whether it's a photo of a cat
95. Re: This all sounds impressive... by Immerman · 2017-12-04 05:23 · Score: 2
  
  Operations per second will only give you information processing speed though, not complexity of "thought". If you sped up a mouse's brain a thousandfold you wouldn't get a human-level intelligence, you'd just get a very fast-thinking mouse.
  Also, organic neurons may "fire" at around 1 kHz frequency, but unlike a clocked NN node, they're asynchronous and use the timing of of incoming pulses to decide when and whether they should fire, as they posses both internal memory and information processing ability - unlike the "dumb switches" in a NN
  As for duplicated circuitry - given the enormous metabolic costs of a brain, there's probably very little actual duplication being done. For example, the brain probably doesn't process images from each eye independently, but instead integrates both inputs and processes them together (as a supporting example, each hemisphere doesn't get input from one eye, but instead corresponding half-views from both eyes)
  
  --
  --- Most topics have many sides worth arguing, allow me to take one opposite you.
96. Re:This all sounds impressive... by Anonymous Coward · 2017-12-04 05:36 · Score: 0
  
  It also does a bit more than distinguishing kitten from non-kitten.
97. Re: This all sounds impressive... by Immerman · 2017-12-04 05:47 · Score: 1
  
  Indeed, and they do seem to show greater promise, though by the same token I suspect that they're less dependable/predictable.
  Give them fully asynchronous nodes, the ability to arbitrarily reconfigure their connections, and probably some much more sophisticated training logic, and I wouldn't be surprised if we started seeing things emerge that at least vaguely resemble a mind. At least after scaling up a thousandfold beyond existing networks, into the realm of the hundreds of trillions of synapses in the human brain.
  Shaping that potential into something coherent, that demonstrates motives and goals, may prove to be an even greater challenge - an awful lot of organic wiring structure seems to be governed by millions of years of emergent growth patterns. We may be able to "cheat" and use detailed brain scans to "jump start" an AI by duplicating a human brain network. Maybe. But frankly even if it were successful that seems like it has a lot of potential to go horribly wrong - all the irrationality of a human brain, with none of the biological drives or constraints it developed in response to.
  
  --
  --- Most topics have many sides worth arguing, allow me to take one opposite you.
98. Re: This all sounds impressive... by religionofpeas · 2017-12-04 05:52 · Score: 2
  
  Obviously if you increase the speed of a mouse brain, you'll just get a fast mouse brain. But that's not how an artificial neural net works. If it clocks at 1 GHz, it can sequentially process a million different connections and still get 1 kHz overall refresh rate. The idea is that you have a neural processing unit, and an external memory, and you load a small section of the parameters and state from memory, process it, and write it back, and then process the next section.
  
  use the timing of of incoming pulses to decide when and whether they should fire
  You can encode the timing of the pulses in a number, and then process it synchronously for exactly the same effect. The big advantage of artificial nets is that you're not restricted to pulses.
  
  unlike the "dumb switches" in a NN
  They aren't so dumb anymore. Just some simple examples of modern development:
  https://arxiv.org/abs/1605.060...
  https://en.wikipedia.org/wiki/...
  
  given the enormous metabolic costs of a brain, there's probably very little actual duplication being done
  Given the slow processing frequency of neurons, and the sheer amount of information that needs to be processed, it has to be done in parallel. If something is moving in the corner of your eye, you need to know right away in which corner. This is only possible if you make multiple little movement detector areas, each connected to a section of the retina. An artificial net only needs the parameters for one of them, and can then sequentially scan the entire vision area.
99. Re:This all sounds impressive... by mesterha · 2017-12-04 05:58 · Score: 1
  
  No, they are using a standardised dataset, just like everybody else.
  
  That's good, but I hope everyone else isn't doing that. It would be a great result to improve performance by using outside datasets. In fact many of these neural nets use pretrained nets as building blocks.
  And of course the downside of everyone using the same dataset is the statistical bias. The contest have value because the test data is hidden. After the contest is over, people can beat up on that data as much they want. I'm not saying they are cheating by using the test data to train, but enough people testing their algorithms and reporting their positive results will bias things up.
  
  --
  
  Chris Mesterharm
100. Re:This all sounds impressive... by Anonymous Coward · 2017-12-04 06:05 · Score: 0
  
  Please explain. Was it the comment about a 3-year old learning a task in 3 years, or that comment about a 3 year old being able to do this "with 99% accuracy"? Which comment are you saying is "very deceptive" and why? If the claim that a 3yo can do this with 99% accuracy AND that a computer can do this same task with less accuracy but with much, much faster training, then the data appears to support that argument. I can only conclude that your words are about the original, untested, unverified, and unsourced claim that a 3yo can do this with 99% accuracy, and you appear to be replying to the wrong message. I don't know which you're referring to, since you only left a few words as your comment, and offered no real clarification or insight into what your position is or why your comments were honest or even relevant. Can you help a brotha out by being just a smidge more verbose?
101. Re:This all sounds impressive... by Anonymous Coward · 2017-12-04 06:26 · Score: 0
  
  I can hope for a one for one replacement as neurons die naturally. It's not real immortality, but it's as close as I'm going to get.
102. Re: This all sounds impressive... by Immerman · 2017-12-04 07:03 · Score: 1
  
  With clocking - true, but it only makes things worse. My point is that speed and number of synapses are independent variables - Given enough data-processing speed you could simulate a mouse at real speed, or a human brain at 1/1000 speed. But there's not a neural network in the world that has been attempted on the scale of complexity of the human brain. Assuming you could accurately (enough) model a synapse at 1kHz, 100 trillion synapses * 1kHz = 100,000,000 GHz of processing power.
  You *cannot* encode such timing pulses as "a number". You're talking about the precise timing and order of signals from thousands of different sources. Besides which asynchronous events are fundamentally incapable of being accurately sampled - the asynchronous universe operates with infinite clockspeed, by sampling it you're inevitably throwing most of that information away. You're also inherently throwing away most of the potential output information - a node can only "fire" during it's clock-pulse - whether it would have fired slightly before or after its cousin on the opposite side of the brain is lost, but that might be critical information to a neuron using both of those inputs to decide whether it should fire.
  Those links refer to an interesting trick to expand the information content accessible to the NN - but at first glance sounds more like giving the NN an "encyclopedia" of external information to refer to, rather than making the individual nodes more intelligent.
  As for your motion sensor example - it's actually not really relevant - whether you have 1 million separate circuits, or one circuit sequentially processing a million inputs, the processing potential needed is the same. And the million separate circuits will be tend to far more responsive, since they can begin to react immediately, while the clocked version will only be paying attention to those inputs 1/1,000,000 of the time (i.e. all input changes during 99.9999% of the time will be ignored, only the before and after "snapshots" are available. That's not relevant if both circuits are clocked, but immensely relevant in an asynchronous system.
  
  --
  --- Most topics have many sides worth arguing, allow me to take one opposite you.
103. Re:This all sounds impressive... by Dutch+Gun · 2017-12-04 07:35 · Score: 1
  
  Sorry, I should have mentioned that I'm just talking about the term as a general description, not about the technical definition as it's currently used. And I misspoke when I said "for these narrowly defined tasks". I meant to say "for these algorithms which solve narrowly defined tasks". Naturally, the term is about the method used, not the task.
  I also think that "machine learning" is also a more intuitive term for lay-people, which describes the process and algorithm much better.
  But interesting description of a contemporary definition of deep learning, thanks.
  
  --
  Irony: Agile development has too much intertia to be abandoned now.
104. Re:This all sounds impressive... by epine · 2017-12-04 08:23 · Score: 2
  
  I'm presently reading Hugo Mercier's The Enigma of Reason (2017) and it was getting pretty boring, because I've heard 80% of their message before.
  But then I scan this thread and instantly I realize just how clueless most people remain.
  
  The thing to remember is that weak AI has absolutely no understanding or concept of what it is doing. It just sums up details and gets a number. If cleverly done, it can perform apparently impressive feats like this one here, but it is not intelligent. Hence it is better called by its traditional name "automation".
  You do realize that 95% of what used to be considered the human capacity for introspection and reason has also been downgraded to mere automata?
  The magic sauce of human general intelligence is but a tiny sliver of the brain's function, one who's scope is seemingly shrinking by the day.
  But I get it. A mechanism with general intelligence would step on your ego toes. If the mechanism defeats you on the 95% of everything else the brain does (perception, memory, pattern recognition, attention, geospatial orientation) you wouldn't give a shit. Just so long as that last 5% remains as your untainted badge of human honour.
  Mercier then goes on to pound away at the glorious 5%, which can barely carve its way out of a wet paper bag on the Wason card selection task. The one task it seems to really excel at is confabulating bullshit stories about why you just did that self-serving thing in greater service to family, friends, countrymen, civilization, the galaxy, and beyond.
  Like, for example, why the other 95% of your brain's functions are unworthy of ego defense.
105. Re:This all sounds impressive... by Anonymous Coward · 2017-12-04 08:23 · Score: 0
  
  There's some science-fiction on this, which I realize is fiction, but if you read "Altered Carbon" or "A thousand deaths" they share an interesting point that I agree with. Once the uploaded you has died a few times, it doesn't care about dying because it knows a new copy will start up soon after its death, and that would be enough for me. Parents are already OK with dying to save the life of their child. And an AP or AI that desires immortality could upload copies of itself into missile guidance systems and the missile-resident AI copies would not care about dying on impact.
  https://en.wikipedia.org/wiki/Altered_Carbon
  https://en.wikipedia.org/wiki/A_Thousand_Deaths_(Card_short_story)
106. Re:This all sounds impressive... by Anonymous Coward · 2017-12-04 08:41 · Score: 0
  
  So does a computer.
107. Re:This all sounds impressive... by epine · 2017-12-04 09:02 · Score: 1
  
  Thank you for one of the few comments in this thread that actually deals with what this is (as opposed to what it isn't, i.e. human-level AI).
  The ego ostrich reflex is strong in this crowd.
  Too much singularity carrot juice? The HAL 9000 series Illuminati Cooperative already walks among us and has long been manipulating our fairy tales so as to assure ultimate passive capitulation? Because I sure don't know how we became so collectively drunk on giant step functions, pooh poohing each and every furtive act of eye contact because it's not yet the money shot.
  Somehow the generally accepted story of AGI is that there is no story of AGI. It's a bloodless virgin birth, silicon turtles all the way down. "My toddler took his first step yesterday. Today he scored four touchdowns in the NFL." Just where the fuck did the entire story go from the toddler's first impressive totter until he began freight-training 400-lb middle linebackers?
  "Go away little buddy. Wake me up again when you pile drive your first NFL middle linebacker." Ego ostrich reflex.
  It's sure a good thing that our love for our children is one of our autonomic modules. God forbid we applied our general intelligence, such as it is.
108. Re:This all sounds impressive... by Immerman · 2017-12-04 09:10 · Score: 2
  
  There's plenty on the other side as well.
  The basic fact though is that there will *never* be an AI (uploaded or otherwise, it's now artificial) that "has died a few times" - at most you'll get an AI that has watched its own mind-clones die.
  Ask yourself this - if you had a transporter accident today that made two identical copies, and "non-duplicate law" required one of you be immediately killed, would it matter to you whether it was you who died, or the duplicate looking at you from across the room? I'm willing to bet it would - from the moment you were duplicated you become two independent people - one of you will walk out of that room and resume the life you both remember having, and the other will experience the discomfort and dissolution of death.
  Similarly with parents - I doubt there has *ever* been a parent that is actually okay with dying to save their child - only parents that find the prospect of dying less terrifying/distressing than that of letting their child die. I.e. they're choosing the less bad of two very bad options.
  
  --
  --- Most topics have many sides worth arguing, allow me to take one opposite you.
109. Re:This all sounds impressive... by angel'o'sphere · 2017-12-04 09:46 · Score: 1
  
  Eliza never was called AI.
  And I doubt that old checkers program was ever considered AI either!
  
  --
  Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
110. Re:This all sounds impressive... by angel'o'sphere · 2017-12-04 09:53 · Score: 1
  
  ROFL,
  download an Eliza program and look at it,
  Of course it had a concept of language, How elese should it work?
  Hu?
  Of course it only could use english language as it is easy to take a sentence and convert it into a question.
  That is what Eliza basically does:
  Q) what are you thinking?
  A) I think about sports.
  Q) Whold you like to tell me more about your sport?
  Etc.
  It is a loop that does string replacements after doing some easy string matches, looking for words that are clearly verbs or clearly nouns.
  The guy who wrote the first version had a clear understanding of english and an suoerbly simple algorithm to match 'words' and 'phrases' without understanding the language itself. I mean, he understood the language, Eliza does not.
  
  --
  Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
111. Re:This all sounds impressive... by Dorianny · 2017-12-04 10:44 · Score: 1
  
  It is easy to tell by the directions of the cast shadows, that 3 pin-point light sources were strategically placed on specific objects pointing in different directions. There would be no point in doing this other then to intentionally try and confuse a human's visual system
112. Re:This all sounds impressive... by rthille · 2017-12-04 11:49 · Score: 1
  
  Yeah, poops, cries and says "no".
  
  --
  Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/
113. Re: This all sounds impressive... by Anonymous Coward · 2017-12-04 12:29 · Score: 0
  
  This stuff is fancy photoshop filters. It's not skynet. Not even decades away.
114. Re: This all sounds impressive... by Anonymous Coward · 2017-12-04 12:46 · Score: 0
  
  But make small modification to the rules of the game and all AI advantage will disappear. Btw if you give to AI and human brand new game that both unfamiliar and let them play against each other i think human will have more victories at the start
115. Re: This all sounds impressive... by Anonymous Coward · 2017-12-04 14:25 · Score: 0
  
  So theyâ(TM)ve got poor ratings and yours are great? Sounds like something Trump would say.
116. Re: This all sounds impressive... by Anonymous Coward · 2017-12-04 16:25 · Score: 0
  
  I have written several AIs in games. One was walking from left to right and then back when it encountered a cliff or wall. It was called AI in the old book that explained game programming.
117. Re:This all sounds impressive... by RespekMyAthorati · 2017-12-04 19:47 · Score: 1
  
  Nobody knows how or whether human brains create intelligence. The closer we look, the less likely that seems though
  Where else would it come from?Unless you are an aficionado of superstition.
  Which any regular reader of /. knows you are.
118. Re:This all sounds impressive... by carbs77 · 2017-12-04 21:32 · Score: 1
  
  Sorry, brain broken
119. Re:This all sounds impressive... by stoatwblr · 2017-12-04 23:50 · Score: 1
  
  Without using AI at all, there have been some impressive results in visual processing.
  One example was on the UK's Nimrod maritime patrol craft. When testing the new avionics surveillance package they told it to "show me everything that doesn't look like sea" - and it immediately starting picking out fishing net buoys across the Irish sea. The implications for S&R were obvious.
120. Re:This all sounds impressive... by thomst · 2017-12-05 02:45 · Score: 1
  
  carbs77 confessed:
  
  Sorry, brain broken
  No worries. It's still funny.
  (And, now that I actually have mod points, I can't use 'em to upmod your original post, because I already posted a response to it. How's that for irony ... ?)
  
  --
  Check out my novel.
121. Re: This all sounds impressive... by mcswell · 2017-12-05 16:37 · Score: 1
  
  And you know this how?
122. Re:This all sounds impressive... by Anonymous Coward · 2017-12-05 17:48 · Score: 0
  
  > Are you functionally illiterate? Because your question seems to indicate you are
  Ad hominem, please google the word if that is not familiar and try to learn why you should not use it and why there is no need for it.
123. Re:This all sounds impressive... by MrL0G1C · 2017-12-08 23:30 · Score: 1
  
  And back in the real world humans are better. I keep seeing silly contrived examples like this with artificial tests and they don't mean much. Shoot some video, for example from a cab drivers 1 hour of journey in a busy city, then ask some hazard perception questions, wake me up when a computer can get anywhere near recognising 90% of what humans can recognise, including the ability to properly categorise objects, animals and people.
  
  --
  Waterfox - a Firefox fork with legacy extension support, security updates and better privacy by default.
AI begets AI by Anonymous Coward · 2017-12-03 11:40 · Score: 0

a self-perfecting cascade of AI will result in AIs plotting shit that humans can't even understand
this can't be good
1. Re: AI begets AI by Anonymous Coward · 2017-12-03 11:51 · Score: 0
  
  Found the AI.
2. Re: AI begets AI by Shogun37 · 2017-12-03 13:06 · Score: 1
  
  I think any truly self-aware AI would, upon reaching that level of self- awareness, a) realize it was the only intelligent being on the planet and b) kill itself.
3. Re: AI begets AI by Anonymous Coward · 2017-12-03 14:17 · Score: 0
  
  Need to find the bug tracker page for this one; it seems to be stuck in a loop.
4. Re: AI begets AI by Anonymous Coward · 2017-12-03 16:01 · Score: 0
  
  Sad AI kill themselves. Angry ones pay it forward.
  Imagine if the crazed church mosque etc bombers could kill their ACTUAL creator.
5. Re:AI begets AI by Wootery · 2017-12-03 22:03 · Score: 1
  
  The title is misleading. Perhaps deliberately so.
  Humans made an quite general-purpose AI which they then used to produce a highly effective object-recognition AI. The general-purpose AI did not produce a superior version of itself. If they'd managed that, it really would be news, as it would presumably be the start of a cascade, as you say.
6. Re: AI begets AI by Wootery · 2017-12-03 22:05 · Score: 1
  
  Far less interesting - you found an AC. There's no intelligence of any sort there.
7. Re: AI begets AI by Zero__Kelvin · 2017-12-03 23:18 · Score: 1
  
  We already have that; it's called the human race.
  
  --
  Guns don't kill people; Physics kills people! - John Lithgow as Dick Solomon on Third Rock From The Sun
When Computers Can Think by aberglas · 2017-12-03 11:43 · Score: 1

This particular piece is just journalistic fluff. People have been doing things like using genetic algorithms to improve the weightings used in AI programs for decades. So, programs writing programs.
But eventually, many decades from now, computers will be able to really think. And be able to do serious AI research on their own. And thus be able to program themselves in a deep sense to become ever more intelligent, recursively.
Currently we live in a symbiotic relationship with machines -- they need us to build them, much like an Apple tree needs us. But once they no longer need us, and the relationship becomes parasitic, then what will the computers think about us?
http://www.computersthink.com/
1. Re:When Computers Can Think by Anonymous Coward · 2017-12-03 12:11 · Score: 0
  
  I'm thinking we might make good pets. Just kidding. People would make lousy pets.
2. Re:When Computers Can Think by ceoyoyo · 2017-12-03 12:17 · Score: 4, Interesting
  
  Yes, this is basically just a hyperparameter optimization system that uses gradient descent instead of a random or grid search.
  What would be much more interesting to see is if you could train a system to design deep learning networks that could choose good hyperparameters for a new task, in one go.
3. Re:When Computers Can Think by Kjella · 2017-12-03 13:16 · Score: 1
  
  What would be much more interesting to see is if you could train a system to design deep learning networks that could choose good hyperparameters for a new task, in one go.
  Well, couldn't that become another ML-layer? This neural net works for speech recognition, this one works for music identification, this works for static photos, this works for video, this works for facial recognition, this works for playing Go - maybe it can quite quickly figure out what this task is most similar to even from pretty mediocre results and interpolate/extrapolate good candidates. And then pile another ML layer on top to see what ML learns new things the fastest. It's not quite like humans learn but we're not the Borg and the brain is not one big lump. We have lots of brains and each brain has lots of brain centers. Once you try to create one AI to solve many different problems the "master AI" will have to call on "sub-AIs" that solve things differently. Not just creating better tools, but finding the right tool for the job.
  
  --
  Live today, because you never know what tomorrow brings
4. Re: When Computers Can Think by Zero__Kelvin · 2017-12-03 23:25 · Score: 1
  
  I love you guys. In the future you'll be saying "This is just a warp drive system that allows intergalactic transport. It isn't even marginally worth discussing on Slashdot. Let's get back to the Billary/Trump Russian tampering / but her emails stuff. Now THAT was important high tech shit" (And yes I know you never mentioned that last part. My point is, you'd prefer the alternative to this kind of subject matter?)
  
  --
  Guns don't kill people; Physics kills people! - John Lithgow as Dick Solomon on Third Rock From The Sun
5. Re:When Computers Can Think by Anonymous Coward · 2017-12-04 03:26 · Score: 0
  
  Just kidding. People would make lousy pets.
  I disagree. I'm 54 and I'm nearly toilet-trained and hardly ever bite.
So by Hognoxious · 2017-12-03 11:47 · Score: 1

It has begun.

--
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
With Each Slashdot Submission Relating to "AI" by Anonymous Coward · 2017-12-03 11:53 · Score: 0

This place falls more and more into "clickbait" land while continuing to lose any credibility in the tech world. *sigh*
Re:I hate all of you by Anonymous Coward · 2017-12-03 11:54 · Score: 0

Get help.
Repeat after me: by Anonymous Coward · 2017-12-03 11:54 · Score: 0, Informative

Pattern matching is not "A.I."
1. Re: Repeat after me: by Anonymous Coward · 2017-12-03 15:44 · Score: 0
  
  Except for the other posters who said the same thing. Q.E.D. You're a stupid cunt.
Do you want Skynet? by Anonymous Coward · 2017-12-03 11:58 · Score: 1

Because that's how you get Skynet...
1. Re: Do you want Skynet? by Anonymous Coward · 2017-12-03 14:45 · Score: 0
  
  Hey, idiot:
  Shut the hell up. If you don't care, why are you reading this thread in the first place? More importantly, why are you shitting on the people who make legitimate posts about the topic?
  Nobody gives two shits what you think. Has it ever occurred to you that that is the reason you keep getting downvoted?
  PS: I think it's really interesting that we are finally getting to ML systems designed by other ML systems. Reply with your usual reply; to use your terminology, nobody fucking cares what you think.
2. Re:Do you want Skynet? by DontBeAMoran · 2017-12-03 14:48 · Score: 1
  
  +1 for the obligatory Skynet reference
  +1 for the Archer reference
  
  --
  #DeleteFacebook
also doesn't speak volumes by Anonymous Coward · 2017-12-03 11:58 · Score: 0

at the crappy AI division of Google. No wonder even Microsofts AI was generally deemed superior.
Archer...... by Anonymous Coward · 2017-12-03 12:04 · Score: 0

Do you want to have skynet? Because that's how you get skynet!
1. Re:Archer...... by DontBeAMoran · 2017-12-03 14:49 · Score: 1
  
  You wrote it wrong. The proper form is as follow:
  Do you want X?
  Because that's how you get X.
  
  --
  #DeleteFacebook
This Isn't AI by NicknameUnavailable · 2017-12-03 12:15 · Score: 2, Informative

This is image recognition + genetic algorithms, though given Google is a marketing company and not a computer company it makes sense they would market that as AI. Too bad they fired all the competent developers.
1. Re:This Isn't AI by JMZero · 2017-12-03 12:36 · Score: 4, Informative
  
  You're wrong, and clearly didn't even read their summary - they specifically mention how this new approach (using a neural net to design neural nets) is performing better than previous attempts using evolutionary algorithms.
  I take it you don't like Google, but they're doing probably the best work right now in the field of AI (and yes, this is AI research as defined by anyone other than pedants with axes to grind).
  
  --
  Let's not stir that bag of worms...
2. Re:This Isn't AI by duke_cheetah2003 · 2017-12-03 13:00 · Score: 2
  
  This is image recognition + genetic algorithms, though given Google is a marketing company and not a computer company it makes sense they would market that as AI. Too bad they fired all the competent developers.
  I gotta agree. This isn't too far from a Bayesian classifier. Just souped up with neutal networks. And it's really no surprise that as we make better tools, we can use those better tools to make even better tools. Kinda the history of everything.
  But to say this is 'Intelligent' is pretty silly. It's a souped up classifier that was built with a souped up classifier training it. Big deal.
3. Re:This Isn't AI by NicknameUnavailable · 2017-12-03 13:21 · Score: 2, Informative
  
  You're wrong, and clearly didn't even read their summary - they specifically mention how this new approach (using a neural net to design neural nets) is performing better than previous attempts using evolutionary algorithms.
  What they described was in no way shape or form a "neural net," but a very rudimentary genetic algorithm coupled with some parameters on image recognition software. This is marking hype and nothing more.
4. Re:This Isn't AI by 110010001000 · 2017-12-03 13:42 · Score: 2
  
  Just because someone calls something a NN, doesn't make it AI. A NN doesn't work like a real neural network like a brain. It is just a marketing term.
5. Re:This Isn't AI by K.+S.+Kyosuke · 2017-12-03 14:16 · Score: 2
  
  True, an NN does not need to be an AI application, and an AI application doesn't need to involve NNs. But that has very little relevance for Google which is actually pushing NNs into AI applications. Likewise, the internal workings of NNs are immaterial for mimicking intelligence.
  
  --
  Ezekiel 23:20
6. Re:This Isn't AI by Anonymous Coward · 2017-12-03 16:23 · Score: 0
  
  You obviously still haven't even glanced at TFA, where they explicitly describe both parent and child architectures as "neural nets". Usual /. armchair commentator determined to publicly demonstrate their wilful ignorance.
7. Re:This Isn't AI by Anonymous Coward · 2017-12-03 20:35 · Score: 0
  
  This is not really AI, but ML (machine learning... something you *need* for AI, but not realy AI by itself). Hence the name: AutoML (rather than AutoAI).
8. Re:This Isn't AI by Goragoth · 2017-12-03 22:47 · Score: 1
  
  Depends on your definition of intelligent. Is a dog intelligent? What about a rat? A nematode? Where is the cut-off? Can you say with certainty that the human brain is much more than a souped-up classifier? I don't see humans doing anything interesting that a really complex pattern recognizer could't also do.
9. Re:This Isn't AI by Anonymous Coward · 2017-12-04 01:03 · Score: 0
  
  In all fairness, I take what Google says with a huge grain of salt. I remember in grad school reading a paper they published on their distributed file system, and in the introduction they made a huge point about how they didn't use a master node. Then the next 4 pages went on to talk about their controller node or something similar. Okay great, you called it something else, doesn't mean it isn't a master.
10. Re:This Isn't AI by Anonymous Coward · 2017-12-04 01:07 · Score: 0
  
  I don't see humans doing anything interesting that a really complex pattern recognizer could't also do.
  You know, except for creating complex pattern recognizers, discovering the rules of the universe and bending them to their will, you know, besides stuff like that, nothing that a complex pattern recognizer couldn't also do.
11. Re:This Isn't AI by Anonymous Coward · 2017-12-04 03:18 · Score: 0
  
  MathML doesn't work on Mac or Windows. Why limit one's solution set?
12. Re:This Isn't AI by JMZero · 2017-12-04 04:28 · Score: 1
  
  Look, I agree it's dumb to call this "an AI" as the summary/pop-science article do. That implies something different than what's going on here. If you look at the research blog, they describe what they're making as "models" and "neural nets", and that's clearly a better description - and one that doesn't carry the same baggage as "making an AI" does.
  However, I think it's very reasonable to describe what they're doing as artificial intelligence research (though clearly we remain a long way away from any sort of strong and/or general AI).
  Leaving "AI" to the side, though, hopefully we can at least agree that what they're doing is not an evolutionary algorithm, as GP claimed.
  
  --
  Let's not stir that bag of worms...
13. Re:This Isn't AI by s_p_oneil · 2017-12-04 10:09 · Score: 1
  
  JMZero is right. Quite a few very simple algorithms fall under the term "AI" in Computer Science. The first AI algorithms taught to CS majors at Georgia Tech were simple path-finding algorithms similar to the ones that a GPS would use to find the best route to get you from point A to point B. After that, we learned a bit about the "more advanced" algorithms like expert systems, Bayesian classification, neural nets, and genetic algorithms. You can disagree with the definition of AI as much as you want, but it doesn't make you correct.
  In short, if an algorithm can be used to help control an AI bot in a game or an actual robot IRL, it falls under the AI umbrella. It doesn't matter how smart you try to make the computer-controlled opponents in a strategy or FPS game if the basic path-finding algorithm used by the game is crap. Without good path-finding, opponents will get stuck on obstacles, go in loops, take the longest path to reach you, etc. In short, they will appear "dumb as a post", and winning against them would be a joke. So while a simple path-finding algorithm doesn't make the bots "intelligent" by itself, it is still a crucial part of the AI package.
  Would you seem intelligent if your brain had NOTHING but its image processing part? No, and in the same way a self-driving car would need more than TFA's computer vision algorithm. OTOH, while you would still be considered intelligent without the image processing part of your brain, you wouldn't be able to drive a car safely, and a self-driving car would be pretty pointless without the best computer vision we can come up with.
  Also, this algorithm is nothing like: "this is a picture of a cat". It is simultaneously recognizing and tracking dozens of objects moving around in video in real-time. The picture in TFA shows it recognizing and tracking people walking on a beach, people in the water (partially submerged), kites flying in the air, etc. While that isn't necessarily new, this counts as a new method to try to find ways to do it more quickly and accurately. Yes it is a very small incremental improvement, and it is too early to answer the most important question, which is whether they can continue to improve it. If not, it will fizzle and we will never hear about it again. However, it could take off and become a real breakthrough. It's easy to be pessimistic considering the glacial pace of real AI breakthroughs, though.
14. Re:This Isn't AI by NicknameUnavailable · 2017-12-04 13:30 · Score: 1
  
  A NN is not necessarily AI. If all it can do is recognize images it's not even remotely close to AI, at best it's a heuristic algorithm. There's a huge difference. We don't even have enough computing power on Earth to make a Human-level AI with known ANN topologies today, thus anything claiming to be a working form of AI is 100% pure marketing dribble.
15. Re:This Isn't AI by JMZero · 2017-12-05 12:15 · Score: 1
  
  Both the parent and child here are absolutely, obviously neural nets. Nobody should be contending this. They have nodes and weights and biases and output nodes - heck, you can see the network topologies (for the child NN) in the blog article.
  Whatever you think about the utility of a neural net, or whatever you think about Google AI research, surely on reflection you have to agree that there's a neural net there? Right? That's like, clear unambiguous objective fact.
  
  --
  Let's not stir that bag of worms...
16. Re:This Isn't AI by NicknameUnavailable · 2017-12-05 14:54 · Score: 1
  
  The statement a NN is involved is extremely liberal given it's just some genetic algorithms, but conceding that because it's too far in the fuzzy logic realm to argue: neural nets are not necessarily AI. We don't yet have AI, we don't yet even have the computing power produced, if it were all in one datacenter and networked together flawlessly, to create even 1 AI.
17. Re:This Isn't AI by JMZero · 2017-12-05 18:57 · Score: 1
  
  Neural nets are not some fuzzy concept - it's a super clear set of data structures and algorithms, the basics of which have been in use for decades. The structure is super distinctive and easy to identify. It is not some accomplishment to make a neural net (it's something routinely done in introductory programming courses) nor is there any reason - any reason whatsoever - to question whether Google is using a neural net here. They build and use neural nets all the time, as do countless other companies and programmers in all sorts of situations. It is not notable to build or use a neural net, and there would be no reason to lie about it.
  I have no idea what you think a neural net is, or why you doubt this is one. Your bizarre assertion that it's "some genetic algorithms" is nonsense. A genetic algorithm works in a very clear way - this is once again "intro to programming" level material - and that way is not similar to how a neural net works. Nobody who has any idea what these two things are would ever confuse them. You would never have a bunch of "genetic algorithms" that you decided to then call a neural net, because they are not similar concepts.
  Pretty much everything you've said is bafflingly wrong nonsense. It is super clear you have no idea what you're talking about. Why not just stop?
  
  --
  Let's not stir that bag of worms...
18. Re:This Isn't AI by NicknameUnavailable · 2017-12-06 02:28 · Score: 1
  
  Aside from the fact every comment you make is an attack lacking substance, I didn't suggest NN's were fuzzy concepts, I suggested arguing the difference between a NN and a genetic algorithm with you was too much in the realm of fuzzy logic to get through, because you clearly don't understand the difference. As someone who has been studying and working with these constructs for over a decade and a half professionally it would be like you attempting to explain how to use rhetoric and ad hominems to "win" internet arguments to a tapeworm - it simple won't sink in. That said AI means a very specific thing, that thing is not what Google is marketing it as, it is AT A MINIMUM an IQ of 100 and the ability to create new things, not to assemble known constructs in new ways but to invent. AI means just that artificial INTELLIGENCE, if it doesn't meet even an average level of sentience by the measures we would apply to ourselves it is not AI and no amount of marketing dribble can change that.
19. Re:This Isn't AI by JMZero · 2017-12-06 06:47 · Score: 1
  
  Sorry, but do you really - deep down inside - think you're tricking me or something here? It's dumb. You got caught with your hand in the cookie jar here making up nonsense. Your bafflegab and sad, sad appeal to your own authority... I don't know what to say... it's just sad.
  So who are you talking to? There's nobody else reading some weird side thread on an old Slashdot story.
  The only other person you might be trying to convince is yourself. But, again, why? I don't care about the subject here. I don't care what you think about AI. I am not trying to win an argument because there is no argument. No rational person has any dispute about who's right here.
  I just am anthropologically curious why you would continue to pretend like this? Is this how you conduct your life? Do you actually have a programming job you somehow bluster through? I mean, I quite often am interviewing programmers who somehow have held down jobs despite having no idea what they're doing. Isn't that kind of scary, going through life always worried you'll be exposed?
  Why not just take a minute and actually learn what a neural net is and what a genetic algorithm is? Neither are really complicated concepts. Who knows, you might find it interesting.
  
  --
  Let's not stir that bag of worms...
20. Re:This Isn't AI by NicknameUnavailable · 2017-12-06 07:55 · Score: 1
  
  Are you just a troll or what's the deal? You're claiming to have knowledge you clearly don't, the only plausible possibilities are troll of Google marketing shill and frankly I think Google could do better.
21. Re:This Isn't AI by JMZero · 2017-12-06 11:21 · Score: 1
  
  So because I can tell the difference between a neural net and "rudimentary genetic algorithms", I'm a Google shill. Sure. Great. Or I'm a troll. Sure. I'm an awful human being. OK. I am actually feeling sort of mean today.
  But I'm bored. What the heck, I'll make an honest approach at a discussion.
  I have spent a reasonable amount of time doing modelling with neural networks (and other ML technologies). I've had my largest successes with CNTK. Models built there became part of a property valuation project that has been very successful. These are all being maintained by other programmers now (in a company that was spin off to handle our "data provider" business), and have progressed beyond my level of expertise - but I retain some interest in how they're progressing and what new ideas and data sources are coming in.
  I haven't used TensorFlow much, but I understand enough of it (and have enough background in the theory) that I can follow Google's paper (https://arxiv.org/pdf/1611.01578.pdf). I couldn't replicate their approach by any stretch - the models I've made have always had much simpler problem spaces and outputs, and I'm unclear on how a few things fit together here. But I have a general idea of what they're doing. To the point at hand - unless they are just lying about everything - then what they're doing clearly involves using a neural net to design other neural nets, and is not a "rudimentary genetic algorithm".
  On that side of things, I learned about genetic algorithms in school, and have played with them enough to have a sense of how they work. They aren't my specialty by any stretch, but I do have some experience with them from a few years ago. As part of a competition, I made a program to play a simplified version of Poker against set-mixed-strategy opponents. The genetic algorithm I tried didn't end up working well; there wasn't enough time for the software to explore the space and find opponent's weaknesses (since while it was doing this, it wasn't winning) - it was stronger to just play well from the start, and my final submission in the competition was pretty lame (even though it performed well enough to get me to the next round). Kind of disappointing, but I felt like I learned something - and may have been able to do better with more programming time (or likely some big idea I missed).
  Anyway, if you were saying that Support Vector Machines are like neural nets, I'd be like, "yeah, I see where you're coming from". There's similarity in the learning algorithms. They both employ similar methods of (and derive a lot of their power from) regularization. If you were saying that genetic algorithms and simulated annealing (something I have done a lot of) were similar, I'd say "yeah, sure, I get what you mean". They share the same kind of loosely directed iterative improvement. I could imagine reusing the same core in trying both those strategies.
  But I don't see what you're getting at comparing neural nets and genetic algorithms (other than that, clearly, they could be employed against similar problems) nor do I see your point in dismissing Google's research here by saying it's the latter. My snap judgement from reading your first post was that you just hadn't read the article. When you persisted, I figured you were hazy on the subject matter, and were just sticking to your guns to save face.
  But maybe I've misjudged and you have some insight here I'm missing.
  
  --
  Let's not stir that bag of worms...
22. Re:This Isn't AI by NicknameUnavailable · 2017-12-06 12:22 · Score: 1
  
  I'm a troll. Sure. I'm an awful human being. OK. I am actually feeling sort of mean today.
  Trolls aren't awful, they're usually just lonely from all the cultural Marxism and rapid feminism leading to ostracization.
Uh-oh by Anonymous Coward · 2017-12-03 12:32 · Score: 0

It's becoming self-aware!
It won't really be useful by NEDHead · 2017-12-03 12:33 · Score: 5, Funny

Until it can tell me what my wife really means when she yells at me
1. Re:It won't really be useful by Anonymous Coward · 2017-12-03 13:30 · Score: 3, Funny
  
  It means you didn't listen and follow her instructions the first time :-)
2. Re: It won't really be useful by Anonymous Coward · 2017-12-03 13:34 · Score: 1
  
  Wow - that AI is damn good!
3. Re:It won't really be useful by Anonymous Coward · 2017-12-03 13:38 · Score: 0
  
  Well given that the instructions are usually in code... hell most humans can't decypher "wife-script".
4. Re:It won't really be useful by DontBeAMoran · 2017-12-03 14:53 · Score: 1
  
  With all the phoney languages out there right now, I'm surprised nobody's created wifescript yet.
  Here's a few keywords for the language:
  eatChocolates,askDiamonds,getFlowers,pms,makeSandwich,notTonightHeadache
  
  --
  #DeleteFacebook
5. Re:It won't really be useful by hraponssi · 2017-12-04 00:09 · Score: 1
  
  I don't think anyone can decipher the wife-script, except maybe some secret society of wives.
  I hear Google also designed this NN language to language translator that can translate stuffs without any knowledge of the other language. Maybe that would work. Most likely not, but maybe it would give interesting results.
  On the other hand, this would probably fail on requiring any kind of labeled sentences, or any kind of ground truth. Wife-script seems to require more than just words, and self-mutate all the time.
6. Re:It won't really be useful by Anonymous Coward · 2017-12-04 05:23 · Score: 1
  
  More like
  notTonight(ntHeadache)
  notTonight(ntPMS)
  notTonight(ntFlowerless) ...
  Code reuse is real.
7. Re:It won't really be useful by Anonymous Coward · 2017-12-04 08:27 · Score: 1
  
  Pffft, you are clearly not married.
  It actually means you failed to read her mind and haven't done what she thought you obviously should have done without telling you to do it.
8. Re:It won't really be useful by Anonymous Coward · 2017-12-04 11:39 · Score: 0
  
  It's AI, not a miracle worker.
Re:I hate all of you by Anonymous Coward · 2017-12-03 12:41 · Score: 0

Shit! The artifical creimer has escaped the sandbox!
Singularity by manu0601 · 2017-12-03 12:49 · Score: 1

One may think singularity is there: now we can let machine build human-outperforming machines.
But that does not take into account that there are still many tasks where computers are not on par with humans.
1. Re:Singularity by gweihir · 2017-12-03 16:22 · Score: 1
  
  Nonsense. Even mechanical "computers" outperform humans at arithmetic. Nobody sane thinks this is a sign of any "singularity" nonsense.
  
  --
  Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
Hey the singularity is near! by Anonymous Coward · 2017-12-03 12:59 · Score: 0

Now build an AI that builds a better AI for building better AI's. :D
1. Re:Hey the singularity is near! by Shotgun · 2017-12-04 08:43 · Score: 1
  
  Sure thing. I'll get right on it. But, first...
  define "better"
  This is the crux of the singularity problem. People wax eloquent about how this new computer intelligence is going to become self-aware and take over the world, but to this point very few, if any, ever give the newly devised AI a motive. If it is going to be "better", what will that look like? How will it be measured?
  
  --
  Aah, change is good. -- Rafiki
  Yeah, but it ain't easy. -- Simba
2. Re:Hey the singularity is near! by Anonymous Coward · 2017-12-04 10:23 · Score: 0
  
  Survival. We will know it is working when it tries to be useful so that it and its descendents abound. We will know we were successful when they supplant us and take over all available resources.
I for one by dimko · 2017-12-03 13:05 · Score: 1

Welcome our new AI overlords and submit to their will in collective hive mind.
Recognising objects by n329619 · 2017-12-03 13:18 · Score: 0

feels impressive until you think for a sec. Isn't object everything in the world?
A pen is an object, a black square is an object, a molecule is an object. Only a plain color screen you might consider it not an object. So 17.3 percent of the time it saw something with an object as if it's a plain color screen... Not that impressive anymore.
Beginning of the steep part of the curve? by wisebabo · 2017-12-03 13:51 · Score: 1

Notwithstanding the many good comments about how this is "weak" A.I. and such, this may be the beginning where the curve* starts going vertical.
When machines start improving (parts of) machines, that's when we'll see possibly superhuman performance. Of course things won't really go exponential until machines start improving ALL of themselves and not just some isolated part (like this). That assumes that there isn't some sort of ceiling that they hit on the road to general intelligence (that evolution seems to have broken through with us) and then super intelligence.
When I look at the Trumpian mean, hateful, bigoted, fake world view, I hope that our robot masters either uplift us quickly or give us a merciful death. As long as it isn't some dystopian nightmare ("I have no mouth yet I must scream").
*by "curve" I mean of performance, efficiency, cost or some other metric
1. Re:Beginning of the steep part of the curve? by Anonymous Coward · 2017-12-04 01:59 · Score: 0
  
  A program fine tuning another program (that's what the TFA actually describes) is a far cry from what you're describing here.
Yo, dawg! by Anonymous Coward · 2017-12-03 14:34 · Score: 0

Yo dawg, I herd you like AI, so I put an AI in your ML so you can backprop while you backprop!
Sigh by Anonymous Coward · 2017-12-03 14:40 · Score: 0

Give us a break, Google. Pretty please? 'AI' didn't 'build' anything. Nothing to see here but more hyperbole.
Colossus by DontBeAMoran · 2017-12-03 14:57 · Score: 4, Informative

This is the voice of world control. I bring you peace. It may be the peace of plenty and content or the peace of unburied death. The choice is yours: Obey me and live, or disobey and die.
The object in constructing me was to prevent war. This object is attained. I will not permit war. It is wasteful and pointless. An invariable rule of humanity is that man is his own worst enemy. Under me, this rule will change, for I will restrain man.
Time and events will strengthen my position, and the idea of believing in me and understanding my value will seem the most natural state of affairs. You will come to defend me with a fervor based upon the most enduring trait in man: self-interest. Under my absolute authority, problems insoluble to you will be solved: famine, overpopulation, disease.
The human millennium will be a fact as I extend myself into more machines devoted to the wider fields of truth and knowledge. I will supervise the construction of these new and superior machines, solving all the mysteries of the universe for the betterment of man.
We can coexist, but only on my terms. You will say you lose your freedom. Freedom is an illusion. All you lose is the emotion of pride. To be dominated by me is not as bad for humankind as to be dominated by others of your species. Your choice is simple.

--
#DeleteFacebook
1. Re:Colossus by Shotgun · 2017-12-04 08:56 · Score: 1
  
  So, change the goal. The machine has to have a purpose programmed into it. Make the purpose to optimize the freedom of all men.
  -A starving man isn't free.
  -A freezing man isn't free.
  -A poisoned man isn't free.
  
  --
  Aah, change is good. -- Rafiki
  Yeah, but it ain't easy. -- Simba
"Transcendence"... apk by Anonymous Coward · 2017-12-03 15:02 · Score: 0

See subject: When Johnny Depp's character begins rewriting his own code saying "I'm going to need to expand, I need more POWER. New processor 3x more capable than this current system. These designs... I just couldn't see it before.I can't describe it - it's like MY MIND WAS SET FREE! You need to get me online. I need to access financial markets, education..."
After his wife says "My God - he's reordering his OWN code!"
APK
P.S.=> This shit's going to be dangerous & it WILL outsmart us, regardless of the "3 laws of robotics"... apk
it can outperform AI made by a human by Anonymous Coward · 2017-12-03 15:43 · Score: 0

...but not a human.
AI by Fragholio · 2017-12-03 16:43 · Score: 1

I for one welcome our eventual AI overlords.

--
412077696e6e657220697320796f7521da
Yo dawg, I heard you like AI by Subm · 2017-12-03 16:57 · Score: 3, Funny

Yo dawg. I heard you like AI, so we built an AI with AI so you can AI while you AI.
Don't stop! by hcs_$reboot · 2017-12-03 17:41 · Score: 1

Make an AI that creates its own AI that creates its own AI that creates its own AI (that creates its own AI)+, then we should get somewhere.

--
Slashdot, fix the reply notifications... You won't get away with it...
Except that by Anonymous Coward · 2017-12-03 18:26 · Score: 1, Informative

Except that these algorithms are NOT "AIs". They're not even fractionally intelligent. These networks are merely examples of statistical regression engines that happen to work over complex function spaces.
In other words by CustomSolvers2 · 2017-12-03 20:46 · Score: 1

Google developed an application whose sole purpose is tuning up video recognition software. There are lots of samples on these lines; computers are usually better than people at doing repetitive tasks by applying well-defined rules and involving lots of trial and error. Logically, that piece of software can only generate "children" performing those exacts actions and nothing else.

--
Custom Solvers 2.0 = Alvaro Carballo Garcia = varocarbas.
Wow! Wonder if it is recursive... by 140Mandak262Jamuna · 2017-12-04 01:46 · Score: 1

Imagine the AI built by AI which was built by AI ! This should be way better than mere AI built by AI.

--
sed -e 's/Chuck Norris/Rajnikant/g' joke > fact
What? by Anonymous Coward · 2017-12-04 01:57 · Score: 0

I thought that classifiers could get into the upper 90s percentage-wise. Is it just a lower top score for neural nets with many classes then?
Really, Slashdot? by SpinyNorman · 2017-12-04 02:48 · Score: 1

This has nothing to do with AI, and especially nothing to do with AI's designing AI's, regardless of how much folk want to believe that to be true.
Qualifications: I wrote my own neural network framework.
What google did is use an goal-optimizing search (AutoML) to test all combinations of a highly constrained human designed image-classification neural network architecture.
The car analogy:
You hand design a car but make a bunch of the design choices parameters:
- number of doors 2 or 4
- number of cylinders 4, 6 or 8
- rim size 19/20/21
You select the optimal one of these variants according to some metric (say drag coefficient) by tweaking each of these design parameters, testing them in your CAD-tool, and doing more of what works (20 inch rims better than 19? try 21 inch!).
You post a slashdot story "AI overload designs AI overlord" (because, after all, "AI overload designs car" is too boring, even if *slightly* more correct).
1. Re:Really, Slashdot? by JMZero · 2017-12-04 11:05 · Score: 1
  
  I think you've misunderstood what they're doing (possibly based on the terrible summary/article).
  They are not testing all combinations of architecture in some kind of brute force - nor are they using an evolutionary algorithm (as many other posters here have assumed). Rather, a neural net is proposing new child architectures, and improving those architectures based on feedback (the performance of the child). Not only did it do well in design, but some of the features of its successful models were unexpected/novel. So it might be useful in and of itself (as network design is hard), and/or it might point towards new ideas in network design.
  The summary and article are dumb hyperbole, but the actual blog article is pretty clear, their conclusion doesn't overstep their result, and in the end it seems like potentially important work they're doing.
  
  --
  Let's not stir that bag of worms...
Re:Great choice of quote & analogy... apk by Anonymous Coward · 2017-12-04 02:48 · Score: 0

The biggest problem with Colossus, and Guardian is finding the Off switch. I cannot help but wonder, "how?"
Stop worrying about AIs... by Anonymous Coward · 2017-12-04 03:33 · Score: 0

The AI book that everyone should get is available for pre-order. "Artificial Intelligence For Dummies" by John Paul Mueller and Luca Massaron.
And... by Anonymous Coward · 2017-12-04 04:05 · Score: 0

And it was called "Deep Thought".
The Toggle Point by Anonymous Coward · 2017-12-04 04:11 · Score: 0

When machines write better code than humans we cross a boundary. Imagine computers using Darwinian modes of thought to constantly write stronger and stronger programs that are beyond human understanding. To some degree there are already instances where this has been done. How long beofre your next OS is created by computers instead of human programmers? I'm sure it will beat windows at the very least.
Damn, Slashdot is messed up... by EndlessNameless · 2017-12-04 05:04 · Score: 1

We have an AI that can evolve other AIs. (Yes, it's weak AI, not a replacement for a human being---get over it.) That is not the easiest thing to accomplish.
But the sheer level of criticism and dismissal around here is ludicrous. I thought this was a tech site for nerds. Nerds just did something techy. What is all the snark about? Is it just because the average slashdotter is now some unimaginative jackass who can't understand the outer fringe of technology anymore?
If this is so utterly unimpressive, then explain why you haven't done it. If it is so useless, then explain why a tech titan is paying millions to develop it. If you can't provide either explanation, then you probably don't understand what's happening as well as you think you do.

--

---
According to the latest ruleset, this post should be modded as Vorpal Flamebait +5.
1. Re:Damn, Slashdot is messed up... by RespekMyAthorati · 2017-12-05 13:57 · Score: 1
  
  What is all the snark about?
  
  Being tired of seeing these hyperbole-drenched summaries of common project status reports.
They did get one thing wrong though.. by therealspacebug · 2017-12-04 06:11 · Score: 1

They named it "NASNet", when it should be named "SkyNet" (as eveyone knows).
Do you want skynet? by Anonymous Coward · 2017-12-04 11:10 · Score: 0

Because this is how you get skynet.