You wont hear me saying AI is just around the corner. Machine learning is really cool in what it can do, but it has some severe limitations when not used just right.
So once the machine breaks down an image as a number of objects, it can them magically know which objects are sheep and which objects are grass? How does it know if an object is a sheep versus a fence post? Why didn't it work in this case? Did the researchers not do it right?
Most of the machine learning work I've done (minimal but I got a good grade in a graduate course so I think I'm minimally qualified to speak on the topic), training is done by giving the machine learning algorithm an image and an indication if any sheep are in the image (how many sheep would be better but more complicated). The algorithm then looks at thousands or millions of images with sheep and without sheep and finds what is in images with sheep.
I don't know enough about their training to say why they did it wrong. I would guess that grass and mountains are appearing in too many images with sheep. So when told all these images have sheep, the algorithm finds the sheep and finds a lot of other things that are common in images with sheep. For example the algorithm might not separating sheep from grass and sheep from mountains but instead "learning" there is a sheep-grass type of sheep that is white and fuzzy with green stems, and a sheep-mountain type of sheep that is white and fuzzy with grey colored flat things.
If you can separate all of the objects in the image, then tell the machine learning algorithm there is a sheep somewhere in the image, it will hopefully only identify the sheep object as the sheep, rather than combining things from the image together, that are common in all images with sheep but are not sheep, into what the algorithm identifies as sheep.
With all of these, I think video would do better in training recognition of an object, but takes a long longer to train. Humans have it a little easier as identifying what makes up an object (who cares what that object is) can be fairly easy for us... until we start looking at clouds and seeing faces.
So do you need to train an AI to recognize every object separately? There are many billions of different objects. How long is this going to take? They seem to have a hard time training it to recognize sheep. When is someone going to work on that?
You don't need to train the AI to recognize every object, but identify what is an object even if the machine learning can't recognize what that object is. If it can determine the sheep is something, the grass is something and the mountain is something, the machine learning can then identify one of those somethings is a sheep and it doesn't care about the others. Now it no longer correlates grass and sheep as a variant on sheep and mountain, instead sees a sheep among a bunch of unknown objects.
Are neural networks new? The concept of neural networks was invented in the 1940s. Why can't they recognize sheep yet?
Again, the whole process is trash. Rather than instructing the algorithm at all about what a sheep is, it is provided two folders of images. One is labelled "has sheep" and the other is labelled "no sheep" and, with no prior perspective on what a sheep is, the algorithm finds some sort of pattern that is present in the "has sheep" folder but not in the "no sheep" folder. Because the scenes are not controlled to have identical situations other than the existence or non-existence of sheep, there will be other correlations that line up with the folders. Like the old airplane identification training that instead of analyzing the objects in the pictures that the researches wanted it to analyze ended up simply evaluating the brightness level of the picture because all the "has plane" images were taken on bright days and the "no plane" images were grey and overcast. Humans try to make more perfectly random sets, but with sufficiently complicated documents (and that's all a computer sees with pictures, yet another kind of document), there will be unexpected correlations regardless of how well a human tries to filter it.
That is a bit simplistic, if not inaccurate. One of the things I learned with machine learning is false correlations are really bad. So accurately training means (preferably) nothing in the image can be correlated, except for what the machine learning program is supposed to be trained on. This fails miserably if there are too many other correlations (fields and mountain sides) in the training images. Humans are some what the same way, give us a picture of a fish swimming through long fields of grass and our brain will scream bloody murder, though we are still pretty good at identifying that fish.
You got to remember the algorithms are still relatively primitive. My guess is that in that pictures were geo-tagged in a region known for sheep. It saw the tubes coming out the ground as legs. In the other photo it saw the white rocks in the creek bed as wool with shadows.
The training overall matters, if the location is part of it, that can lead to false positives. Also if the neural net does not try to separate unique objects and then identify them, it might identify the grass as "part" of the sheep. Machine learning is still only as good as the data it is being trained on, if it is trained with data with a false correlation, it cannot filter it out without additional training on data without the false correlation.
Status. They're fashionable. New dark blue jeans are associated with low status people, and nobody wants to be mistaken for a low status person. Imitating high status people is very deeply ingrained in humans, for good reason.
Yeah, the low status people that own new stuff. So buying shiny new cars also represents low status?
I've never seen ability to code as being what makes a good developer, at least among workers at least capable of writing a good program. In the environments I've worked in, the programs are all too large for one person to work on so the only way to progress the program is to work with the other people responsible for the rest of the program. Are employers just figuring this out?
I commute to work from my town, to a city about 50KM away. Pickup is across the street from my building, and drop-off is a 5 minute walk from my office. The US just has terrible public transport.
Not sure what I'm more jealous of, the public transport, or the fact you can use metric...
Mass transit is of limited use. It is a pain when you have to do a transfer or your destination is a long ways from a stop. I can easily see Uber which offers door to door service pulling people off from a mass transit system that doesn't really go where they need it to.
Boston also has a special problem of the north commuter rail system not being connected to the south one. So if you have to cross this boundary it forces a transfer onto the subway. Subway and commuter rail are separate systems and require two fares. When you add this up, an Uber Pool is definitely price competitive.
Yeah, last I checked mass transit in my area (fairly populated suburbs), I was looking at taking two buses to travel 15 miles to work, with a 1 hour wait and 1(2?) mile walk between them.
If we're headed for self-driving cars this seemingly trivial problem should be closer to 100% not 78%.
Nothing trivial about this kind of image recognition. There have been plenty of documented cases where a sophisticated algorithm can be made worthless by a simple sticker. Some day we should get closer to 99% but a computer system may never reach 100% when dealing with corner cases.
Need a crypto coin that is unlocked by doing processing for cancer genetic research. Probably no way to make it work but it would be nice if all this crypto coin mining processing went to something worthwhile.
Utter nonsense. Get a cheap bicycle and rid down to a local park. Drop by some basketball courts and meet some new people and get some exercise. Go down to the city library and find some interesting books to read. Go to some wacky community event involving art or music. There's all manner of things that can be done for free and even more on top of that which can be done for $10 or less if you're willing to look around a bit.
I think the real truth is that the 18-24 crowd is too absorbed in Facebook, Twitter, and other social media to want to get outside. If John Calhoun were still alive he'd be yelling about behavioral sink right about now.
Interestingly, most of those inexpensive methods of getting out to have fun use little to no energy, particularly if you use a bicycle to get there.
I wonder if 1/50 is truly face blind or all fall into related conditions. I don't think I'm face blind, I can pick out my wife from pictures pretty well. For me, I quickly forget facial features of people I don't see every day and tend to see familiar faces a lot with an initial glance, but on looking closer I realize it is someone else. I suspect linking faces to familiar memories is fairly common for everyone.
For me, lack of Disney content will kill my viewing of Disney products. Eventually I may give up on Netflix but when that happens, I wont be changing to any other providers, unless their combined cost is less than Netflix.
You are missing the point. Since we can create AI programs that can play Go and Chess, we will soon have AI systems that can replace doctors, lawyers and everything in between. It is inevitable, especially with computers getting faster and faster every year.
Go and chess is a little different from doctors and lawyers. People who go to doctors and lawyers tend not to care how human the doctor or lawyer is (lawyers were questionable long before AI), they just want the job done. People who watch a go or chess tournament are interested in entertainment and watching who plays the most interesting strategy, not necessarily the most efficient.
Motorsports has started looking into AI driven race cars. This will be interesting but I'm not sure how successful it will be. Computer games have had AI players for a long time but some players insist on playing with others online, rather than against the AI. Given enough time, perhaps the AI personality will be convincing enough to feel the same as a human, just feels like that is a fairly long way off.
It has already happened. With deep learning neural networks which think like the human brain, we already have systems that can replace a large swath of professions. Go ask a Go or Chess master how many job offers they get nowadays!
Maybe if America stopped being such a global dick, it wouldn't have to worry about hostile nations. Maybe try not being a dick? Not bombing the shit out of countries? You'd be surprised how angry and hostile people get when American drones are killing innocent civilians in the pursuit of terrorists that American policies created in the first place. I'm just saying, maybe give it a try.
So Syria and Russia aren't an issue? Plus they don't break down which countries in the international coalition actually did the killing. The US leads to coalition but other countries are involved.
I always try to have the distance to the car ahead of me set so I never have to hit the brakes and when the traffic does slow up, not having more than the usual 2 second gap just before it speeds up again. I feel like this results in me maintaining the highest speed possible. Of course people usually cut in front of me, so I have to slow down more than I would otherwise.
This is only for freeway traffic, city streets and inconsistent stoplights are a whole other ballgame.
Seven hundred a month for a 20 year mortgage?? Houses in your area must be cheaaaap.
This was a somewhat low end house in 2005, just before the market plunge. When I sold it and moved into a slightly larger house the mortgage would have been $450 a month, or less, for 30 year. Tiny houses or fixer uppers can be significantly cheaper.
Slashdot Mirror
Yep, just like a brain works, like, a brain that makes a gnat look like Einstein. That they work as well as they do is rather impressive.
You wont hear me saying AI is just around the corner. Machine learning is really cool in what it can do, but it has some severe limitations when not used just right.
So once the machine breaks down an image as a number of objects, it can them magically know which objects are sheep and which objects are grass? How does it know if an object is a sheep versus a fence post? Why didn't it work in this case? Did the researchers not do it right?
Most of the machine learning work I've done (minimal but I got a good grade in a graduate course so I think I'm minimally qualified to speak on the topic), training is done by giving the machine learning algorithm an image and an indication if any sheep are in the image (how many sheep would be better but more complicated). The algorithm then looks at thousands or millions of images with sheep and without sheep and finds what is in images with sheep.
I don't know enough about their training to say why they did it wrong. I would guess that grass and mountains are appearing in too many images with sheep. So when told all these images have sheep, the algorithm finds the sheep and finds a lot of other things that are common in images with sheep. For example the algorithm might not separating sheep from grass and sheep from mountains but instead "learning" there is a sheep-grass type of sheep that is white and fuzzy with green stems, and a sheep-mountain type of sheep that is white and fuzzy with grey colored flat things.
If you can separate all of the objects in the image, then tell the machine learning algorithm there is a sheep somewhere in the image, it will hopefully only identify the sheep object as the sheep, rather than combining things from the image together, that are common in all images with sheep but are not sheep, into what the algorithm identifies as sheep.
With all of these, I think video would do better in training recognition of an object, but takes a long longer to train. Humans have it a little easier as identifying what makes up an object (who cares what that object is) can be fairly easy for us... until we start looking at clouds and seeing faces.
So do you need to train an AI to recognize every object separately? There are many billions of different objects. How long is this going to take? They seem to have a hard time training it to recognize sheep. When is someone going to work on that?
You don't need to train the AI to recognize every object, but identify what is an object even if the machine learning can't recognize what that object is. If it can determine the sheep is something, the grass is something and the mountain is something, the machine learning can then identify one of those somethings is a sheep and it doesn't care about the others. Now it no longer correlates grass and sheep as a variant on sheep and mountain, instead sees a sheep among a bunch of unknown objects.
Why are the algorithms so primitive?
Because the whole model used is trash.
Are neural networks new? The concept of neural networks was invented in the 1940s. Why can't they recognize sheep yet?
Again, the whole process is trash. Rather than instructing the algorithm at all about what a sheep is, it is provided two folders of images. One is labelled "has sheep" and the other is labelled "no sheep" and, with no prior perspective on what a sheep is, the algorithm finds some sort of pattern that is present in the "has sheep" folder but not in the "no sheep" folder.
Because the scenes are not controlled to have identical situations other than the existence or non-existence of sheep, there will be other correlations that line up with the folders. Like the old airplane identification training that instead of analyzing the objects in the pictures that the researches wanted it to analyze ended up simply evaluating the brightness level of the picture because all the "has plane" images were taken on bright days and the "no plane" images were grey and overcast. Humans try to make more perfectly random sets, but with sufficiently complicated documents (and that's all a computer sees with pictures, yet another kind of document), there will be unexpected correlations regardless of how well a human tries to filter it.
That is a bit simplistic, if not inaccurate. One of the things I learned with machine learning is false correlations are really bad. So accurately training means (preferably) nothing in the image can be correlated, except for what the machine learning program is supposed to be trained on. This fails miserably if there are too many other correlations (fields and mountain sides) in the training images. Humans are some what the same way, give us a picture of a fish swimming through long fields of grass and our brain will scream bloody murder, though we are still pretty good at identifying that fish.
You got to remember the algorithms are still relatively primitive. My guess is that in that pictures were geo-tagged in a region known for sheep. It saw the tubes coming out the ground as legs. In the other photo it saw the white rocks in the creek bed as wool with shadows.
The training overall matters, if the location is part of it, that can lead to false positives. Also if the neural net does not try to separate unique objects and then identify them, it might identify the grass as "part" of the sheep. Machine learning is still only as good as the data it is being trained on, if it is trained with data with a false correlation, it cannot filter it out without additional training on data without the false correlation.
Status. They're fashionable. New dark blue jeans are associated with low status people, and nobody wants to be mistaken for a low status person. Imitating high status people is very deeply ingrained in humans, for good reason.
Yeah, the low status people that own new stuff. So buying shiny new cars also represents low status?
I've never seen ability to code as being what makes a good developer, at least among workers at least capable of writing a good program. In the environments I've worked in, the programs are all too large for one person to work on so the only way to progress the program is to work with the other people responsible for the rest of the program. Are employers just figuring this out?
I commute to work from my town, to a city about 50KM away. Pickup is across the street from my building, and drop-off is a 5 minute walk from my office. The US just has terrible public transport.
Not sure what I'm more jealous of, the public transport, or the fact you can use metric...
Mass transit is of limited use. It is a pain when you have to do a transfer or your destination is a long ways from a stop. I can easily see Uber which offers door to door service pulling people off from a mass transit system that doesn't really go where they need it to.
Boston also has a special problem of the north commuter rail system not being connected to the south one. So if you have to cross this boundary it forces a transfer onto the subway. Subway and commuter rail are separate systems and require two fares. When you add this up, an Uber Pool is definitely price competitive.
Yeah, last I checked mass transit in my area (fairly populated suburbs), I was looking at taking two buses to travel 15 miles to work, with a 1 hour wait and 1(2?) mile walk between them.
If we're headed for self-driving cars this seemingly trivial problem should be closer to 100% not 78%.
Nothing trivial about this kind of image recognition. There have been plenty of documented cases where a sophisticated algorithm can be made worthless by a simple sticker. Some day we should get closer to 99% but a computer system may never reach 100% when dealing with corner cases.
Sorry, this was just too good of news not to...
Let me help with that: https://www.youtube.com/watch?...
Need a crypto coin that is unlocked by doing processing for cancer genetic research. Probably no way to make it work but it would be nice if all this crypto coin mining processing went to something worthwhile.
Step 1: Drive somewhere without cloud cover.
Step 2: Realize you no longer have a job due to mandatory meetings today.
Step 3: Sigh
Meetings are easily moved. However since I am currently in the Netherlands it's that "somewhere without cloud" that is woefully impractical :)
We have lake effect snow creating cloud cover anywhere within a 4 hour or so drive which kind of impacts the practicality of relocating here too.
Step 1: Drive somewhere without cloud cover.
Step 2: Realize you no longer have a job due to mandatory meetings today.
Step 3: Sigh
Utter nonsense. Get a cheap bicycle and rid down to a local park. Drop by some basketball courts and meet some new people and get some exercise. Go down to the city library and find some interesting books to read. Go to some wacky community event involving art or music. There's all manner of things that can be done for free and even more on top of that which can be done for $10 or less if you're willing to look around a bit.
I think the real truth is that the 18-24 crowd is too absorbed in Facebook, Twitter, and other social media to want to get outside. If John Calhoun were still alive he'd be yelling about behavioral sink right about now.
Interestingly, most of those inexpensive methods of getting out to have fun use little to no energy, particularly if you use a bicycle to get there.
I wonder if 1/50 is truly face blind or all fall into related conditions. I don't think I'm face blind, I can pick out my wife from pictures pretty well. For me, I quickly forget facial features of people I don't see every day and tend to see familiar faces a lot with an initial glance, but on looking closer I realize it is someone else. I suspect linking faces to familiar memories is fairly common for everyone.
For me, lack of Disney content will kill my viewing of Disney products. Eventually I may give up on Netflix but when that happens, I wont be changing to any other providers, unless their combined cost is less than Netflix.
You are missing the point. Since we can create AI programs that can play Go and Chess, we will soon have AI systems that can replace doctors, lawyers and everything in between. It is inevitable, especially with computers getting faster and faster every year.
Go and chess is a little different from doctors and lawyers. People who go to doctors and lawyers tend not to care how human the doctor or lawyer is (lawyers were questionable long before AI), they just want the job done. People who watch a go or chess tournament are interested in entertainment and watching who plays the most interesting strategy, not necessarily the most efficient.
Motorsports has started looking into AI driven race cars. This will be interesting but I'm not sure how successful it will be. Computer games have had AI players for a long time but some players insist on playing with others online, rather than against the AI. Given enough time, perhaps the AI personality will be convincing enough to feel the same as a human, just feels like that is a fairly long way off.
It has already happened. With deep learning neural networks which think like the human brain, we already have systems that can replace a large swath of professions. Go ask a Go or Chess master how many job offers they get nowadays!
How many job offers did they get 10 years ago?
Maybe if America stopped being such a global dick, it wouldn't have to worry about hostile nations. Maybe try not being a dick? Not bombing the shit out of countries? You'd be surprised how angry and hostile people get when American drones are killing innocent civilians in the pursuit of terrorists that American policies created in the first place. I'm just saying, maybe give it a try.
So Syria and Russia aren't an issue? Plus they don't break down which countries in the international coalition actually did the killing. The US leads to coalition but other countries are involved.
http://www.iamsyria.org/syrian...
I always try to have the distance to the car ahead of me set so I never have to hit the brakes and when the traffic does slow up, not having more than the usual 2 second gap just before it speeds up again. I feel like this results in me maintaining the highest speed possible. Of course people usually cut in front of me, so I have to slow down more than I would otherwise.
This is only for freeway traffic, city streets and inconsistent stoplights are a whole other ballgame.
Seven hundred a month for a 20 year mortgage?? Houses in your area must be cheaaaap.
This was a somewhat low end house in 2005, just before the market plunge. When I sold it and moved into a slightly larger house the mortgage would have been $450 a month, or less, for 30 year. Tiny houses or fixer uppers can be significantly cheaper.
One modern spy sat fly over could probably do more than a year of on the ground mapping.