Slashdot Mirror
AI Is in a 'Golden Age' and Solving Problems That Were Once Sci-fi, Amazon CEO Says (yahoo.com)
An anonymous reader writes: Artificial intelligence (AI) development has seen an "amazing renaissance" and is beginning to solve problems that were once seen as science fiction, according to Amazon CEO Jeff Bezos. Machine learning, machine vision, and natural language processing are all strands of AI that are being developed by technology giants such as Amazon, Alphabet's Google, and Facebook for various uses. These AI developments were praised by the Amazon founder. "It is a renaissance, it is a golden age," Bezos told an audience at the Internet Association's annual gala last week. "We are now solving problems with machine learning and artificial intelligence that were in the realm of science fiction for the last several decades. And natural language understanding, machine vision problems, it really is an amazing renaissance." Bezos called AI an "enabling layer" that will "improve every business."
Kindly do the needful and rephrase the headline as a question.
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
For the most of us, AI is trouble ahead that needs to be stopped and slowed down.
Perhaps Mr. Bezos should read a bit of Herbert's Dune to know what happens when you let technological progress go unchecked. The end result is worse than what it would be if humanity were included.
Twitter supports and protects racists - by smearing their critics with the "Hate Speech" label.
Once again, someone is talking about soft AI, and the reporter interprets it as hard AI, and mass confusion results. Expect follow-up stories about how AI will take over the world.
"First they came for the slanderers and i said nothing."
This might be the Stone Age, or at the most the Bronze Age, but we are nowhere near the Golden Age or the Renaissance.
The golden age of AI was in the early 1980's when I read all about it Bytes Magazine.
This is just more hype, ignore it.
His assertions stand in stark contrast to what I've heard about developing for Alexa. As I understand, it performs the speech-to-text translation for you, but when it comes to parsing the text and interpreting what it means, you're on your own.
Right now, most of it runs on GPUs, and requires lots of them. GPUs weren't really made for that task and there is potential for efficiency gains.
No, GPUs were made for a very similar task: parallel iamge processing. It is similar in the same way that your visual cortex, used for input, is arranged in much the same way as a GPU is, for output. The potential for efficiency gains from using GPUs better has barely been scratched. Sure, some special purpose code like s-expression evalutation can be accelerated by ASIC, but it would be a serious mistake to assume that is all there is to AI.
When all you have is a hammer, every problem starts to look like a thumb.
I'm very concerned right now; I find myself nodding to the comments of an A/C.
than the same claims of the 70's and 80's....you're only as good as the data you provide and the second was that OOP models to do deep learning basically plateaued always no matter how much compute power you threw at it.
The more current models for deep learning seems to scale much better, but also the sheer amount of data collection (not to mention data storage/cost) is why you are seeing people so jazzed about this.
Here is the rub though...you still need the "right kind" of data to correctly train todays deep learning models. One of the biggest mistakes I have seen people make is they train with the wrong type of data. For instance if you want to do facial recognition, so you just grab 10k random faces from snapchat...well, what if your real life image capture is much lower quality? How about race demographics? Your training your model on data that isn't indicative of real life situations, and this is why lots of startup AI fails.
I was impressed by Apple's phone system.
I said a sentence, and it out me right to the correct place.
Wow, sent an e-mail as suggested when clicking on "use classic" banner, and got a fast response that addressed my msg
Scientific advancement should not be stopped or slowed down because it might put a bunch of low-level functionaries out of a job.
Our economic model needs to adapt to the new technological landscape, and leave the Luddites behind.
I would like to see AI involved in running businesses. ,well now that's some real money!
Replacing 100 low level people saves a fair chunk of the bottom line.
Replacing a CEO earning 500x those low level people
Imagine if AI replaced some of your congress critters, how are the corporates going to buy one of those off - offer some green electricity? Licence maintenance paid for another year?
What is being lauded as A.I. is not the sort of A.I. described in most S.F. books and Bezos knows it. He has a product to sell so that's why he's embiggening A.I.
I'd rather be riding my '63 Triumph T120.
Obviously, he missed the bit about the rise of Sexbots.
Sent from my ASR33 using ASCII
Office Space :
Lawrence: Well, what about you now? What would you do?
Peter Gibbons: Besides two chicks at the same time?
Lawrence: Well, yeah.
Peter Gibbons: Nothing.
Lawrence: Nothing, huh?
Peter Gibbons: I would relax... I would sit on my ass all day... I would do nothing.
Lawrence: Well, you don't need a million dollars to do nothing, man. Take a look at my cousin: he's broke, don't do shit.
Or they put you in a position where you have no choice.. "Take our $10 million or we'll fight you forever"
I think you mean high-level functionaries. It makes very little sense to put low paid people out of the job when you can put high paid ones out of the job.
"Replacing a CEO earning 500x those low level people ,well now that's some real money! "
Considering what these C officers actually do in most organizations I'd think LISA could replace them successfully.
If it is possible at all. There has been zero progress in the direction of "strong" AI (the version that actually has intelligence in the sense of understanding) in the last 30 years or so and that unfortunately includes theory. The theory we have today just does not include anything that can generate intelligence and hence for that question "faster" does not make one bit of difference. Sure, automation is becoming useful in more areas due to speed and memory increases, but qualitatively it has not changed.
For example, when I wen to university, there was an expectation that today the majority of mathematical proofs would be done by machines. That did not happen at all. What happened is that mathematicians how have systems they can explain the details of a proof to and the machine can (mechanically) verify that the steps are correct, fit together and deliver the expected result. Very useful, but the idea has to be supplied by actual intelligence, i.e. by a human being.
That there has not been any progress on strong AI in the last few decades can mean two things though: a) It is actually impossible. This would mean we have a fundamental gap in understanding what intelligence is. That is a possibility, though in what direction actual explanations would go is unclear. Certainly some sort of "magic" would be required. And b), it is possible, but some fundamental idea is missing. In that case we could still end up at something like a), because if the idea needed is way to complex for humans, then we are not going to find it. Alternatively, in the next few 100 years, some smart person may have the right idea.
Personally, I started out thinking strong AI is possible, but likely with the usual drawbacks, i.e. consciousness, personality, not smarter then humans and has to be motivated somehow. Over the years of following AI research, I have mostly moved my expectations to "likely not possible", because a lot of very smart people have not made any progress at all. Something seem to be missing in this picture and the explanation "human brains do it" does not cut it either, because on that front there was also no progress in finding out how intelligence is done. Sure, _automation_ in the brain is beginning to be understood, but that is not intelligence.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
In another posting, people were complaining how hard it is to write skills for the Amazon Echo by trying to write the skill so it can catch every possible phrase spoken. They concluded, "this is not A.I." Well, technically, A.I. is a massively complex decision tree, so, yes, it is A.I.
On that note, Amazon Lex is one of the services that you should be looking for if you want to know more about A.I. on AWS to create artificial conversations without writing every possible phrase.
Kriston