People are already using LeelaZero (the #3 ranked in gobot competition and capable of beating 9d professionals) to cheat, but when a human plays drastically better than their rating, the cheating is pretty obvious.
LeelaZero - an open source go bot, has beat 9d professionals and other lower ranked professionals. It is also ranked #3 in the world in gobot competitions, and that was with using half or less of the hardware resources that many of hte competitors had (LeelaZero was using 4 1080 TI GPUs; the competitors had 10 1080 TI GPUs).
It still hasn't reached the level of AlphaZero, but if you'd like to help it do so, you can contribute here.
Note that they benchmarked against LeelaZero, but had it misconfigured - they gave their bot 80,000 playouts, and LeelaZero 50 seconds per move, but left a default where LeelaZero doesn't use all of its time. So often it was moving in 3 seconds. It might well be weaker than LeelaZero on similar hardware when LeelaZero is correctly configured.
I hope the feds or states go after them for antitrust violations. They are leveraging their monopoly in high speed internet access to support their cable subscription business.
We actually have no idea if it is possible. We certainly can't do it with today's technology - but it may be that the connectome and other information (mylenation thickness and extent) preserved when freezing the brain - may be sufficient to do a digital simulation/upload 'in the future'.
A connectome does seem adequate to simulate extremely simple nervous systems.
The gain difference isn't necessarily artificially darkened. It is speculated it was tuned so that it wouldn't be blown out by headlights of oncoming traffic. Without high dynamic range cameras, you have tradeoffs for what lighting conditions your camera is adapted for.
What reason did the Uber car have for going 38 in a 35 zone?
I was doing 38 it a 45 (there is a photo online of the speed limit sign). The report of it being a 35 zone was a misreport/misspeak by the police representative.
Microsoft has 124,000 employees, 25.8% that are women.
So 31,992 women with 40 complaints a year. so.1% of female employees file a complaint each year. I suspect that is probably less than the industry norm.
[quote]On the flip side, it's a good time to learn how to be a robot repair person. These things are going to break, frequently.[/quote]
Training robots to repair robots will probably be quite easy. Usually it will be just swapping out a complete sub assembly. Then the subassembly will be shipped to a central repair location (if it is expensive enough to be worth repairing) where it can be run through advanced diagnostic and repair tools.
There will never be a significant number of 'robot repair' jobs for humans.
Hopefully most stuff made of meat will be made from plant sources in the near future. The energy savings and positive environmental impact could be extraordinary.
If this were feasible, those same businesses would already be operating public transit (buses with drivers) with the same premise. So what's missing? Ah, it's income qualification be make sure the passengers have enough disposable income to make purchases likely. Now THAT I can believe.
We already have something similar called 'validated parking'. Shop and your parking is free.
One part solving extremely complicated math word problems. (algorithm creation) One part extreme proof reading for a grammar nazi. (debugging) One part playing charades or pictionary with the worlds most inept clue giver. (gathering requirements from clients)
It isn't clear if the allegations are true. TFA says the accuser "walked back" some of his allegations, which is a euphemism for "admitted he was lying".
After he wrote the letter he was hired as a 'consultant' for 4.5 million dollars, of which, it is paid out over time and he has thus far recieved 1 million (a significant percentage of the payment is due at the end of the contract). If he testifies that everything in the document is true, he will likely be fired and not get 3.5 million dollars, so he has 3.5 million reasons to say that the document isn't accurate.
It isn't in GANs one model, the generative, tries to fool the other, classifier one, by giving it images itself has generated. There's no incremental task switching from easier problem domains to more difficult ones.
I'm familiar with GANs - what it sounded like (and what they did) is add curriculum learning, but they also did it layerwise as is done with autoencoders, (Also they had some other interesting ideas, but that was the crucial bit). In this case the easy is the lower resolution images and the hard is the higher dimensional images.
From their paper
The idea of growing GANs progressively is related to curriculum GANs (Anonymous), where the idea is to attach multiple discriminators that operate on different spatial resolutions to a single generator, and furthermore adjust the balance between resolutions as a function of training time. That work in turn is motivated by Durugkar et al. (2016) who use one generator and multiple discriminators concurrently, and Ghosh et al. (2017) who do the opposite with multiple generators and one discriminator. In contrast to early work on adaptively growing networks, e.g., growing neural gas (Fritzke, 1995) and neuro evolution of augmenting topologies (Stanley & Miikkulainen, 2002) that grow networks greedily, we simply defer the introduction of pre-configured layers. In that sense our approach resembles layer-wise training of autoencoders (Bengio et al., 2007).
It takes more than a Format command to destroy data on a hard drive (the age of the computer means spinning platters), because the erase head never perfectly aligns with the written data, there is always a little bit missed. A good data recovery company should be able to recover everything.
They techs did a triple degaussing, even the NSA is unlikely to be able to recover anything.
They used character by character training, rather than words tagged with word sense.
More sophisticated 'word sense' tagging can be done which can differentiate between all of the various meanings of gay, but for most words there isn't enough data to train all of the word senses sufficiently.
Tesla Motors was created for the purpose of proving EVs were viable and proliferating them, so this would be a win for Elon. If you haven't you noticed, all their patents are free to use.
Only for those companies that reciprocate and make all of their patents free to use for Tesla.
Fair use is far broader than that - 'transformative use' - (such as using a book as input to a machine learning algorithm) is one of many additional fair use defense.
Sure, there exist a computer that, in a clearly rigged contest (the computer knew games of the player, but not the other way round), defeated a really good Go player.
You have it backwards. AlphaGo knew literally nothing about its opponents and hadn't been trained on any of its opponents games. The opponents (Lee Sedol and the much more recent match versus Ke Jie) had access to a small number of previous Go games played by AlphaGo.
You must have missed the memo reporting that a lot of these ads were plugging Black Lives Matter, Hillary's widespread support, and similar topics.
It matters who those ads were targeted at. If those ads were targeted at Republicans - it would tend to increase turnout for Republican voters who feared Clinton winning.
I studied AI algorithms back in University a decade ago and they really haven't changed much from now till today. The biggest improvements have come from having faster computers, not more efficient or even effective algorithms.
Actually a number of algorithmic improvements were needed to train deep networks, so it isn't just hardware, some major algorithmic improvements have occurred also. Also GAN's certainly weren't around a decade ago.
Slashdot Mirror
People are already using LeelaZero (the #3 ranked in gobot competition and capable of beating 9d professionals) to cheat, but when a human plays drastically better than their rating, the cheating is pretty obvious.
LeelaZero - an open source go bot, has beat 9d professionals and other lower ranked professionals. It is also ranked #3 in the world in gobot competitions, and that was with using half or less of the hardware resources that many of hte competitors had (LeelaZero was using 4 1080 TI GPUs; the competitors had 10 1080 TI GPUs).
It still hasn't reached the level of AlphaZero, but if you'd like to help it do so, you can contribute here.
http://zero.sjeng.org/
Note that they benchmarked against LeelaZero, but had it misconfigured - they gave their bot 80,000 playouts, and LeelaZero 50 seconds per move, but left a default where LeelaZero doesn't use all of its time. So often it was moving in 3 seconds. It might well be weaker than LeelaZero on similar hardware when LeelaZero is correctly configured.
I hope the feds or states go after them for antitrust violations. They are leveraging their monopoly in high speed internet access to support their cable subscription business.
We actually have no idea if it is possible. We certainly can't do it with today's technology - but it may be that the connectome and other information (mylenation thickness and extent) preserved when freezing the brain - may be sufficient to do a digital simulation/upload 'in the future'.
A connectome does seem adequate to simulate extremely simple nervous systems.
The gain difference isn't necessarily artificially darkened. It is speculated it was tuned so that it wouldn't be blown out by headlights of oncoming traffic. Without high dynamic range cameras, you have tradeoffs for what lighting conditions your camera is adapted for.
It was doing, not I... stupid typo
What reason did the Uber car have for going 38 in a 35 zone?
I was doing 38 it a 45 (there is a photo online of the speed limit sign). The report of it being a 35 zone was a misreport/misspeak by the police representative.
238/6 years is 40 a year.
Microsoft has 124,000 employees, 25.8% that are women.
So 31,992 women with 40 complaints a year. so .1% of female employees file a complaint each year. I suspect that is probably less than the industry norm.
[quote]On the flip side, it's a good time to learn how to be a robot repair person. These things are going to break, frequently.[/quote]
Training robots to repair robots will probably be quite easy. Usually it will be just swapping out a complete sub assembly. Then the subassembly will be shipped to a central repair location (if it is expensive enough to be worth repairing) where it can be run through advanced diagnostic and repair tools.
There will never be a significant number of 'robot repair' jobs for humans.
It landed in the water.
Absolutely if it is equal in taste and nutrition.
Impossible Burger and Beyond Burger are growing rapidly, and might well end up capturing huge amounts of the US beef market.
https://www.fooddive.com/news/impossible-burger-making-its-way-to-foodservice-venues/507812/
Eggs will likely be replaced by plant based artificial eggs in most of the food industry.
http://www.collective-evolution.com/2015/11/06/the-worlds-first-plant-based-egg-is-putting-the-egg-industry-in-a-panic/
Hopefully most stuff made of meat will be made from plant sources in the near future. The energy savings and positive environmental impact could be extraordinary.
If this were feasible, those same businesses would already be operating public transit (buses with drivers) with the same premise. So what's missing? Ah, it's income qualification be make sure the passengers have enough disposable income to make purchases likely. Now THAT I can believe.
We already have something similar called 'validated parking'. Shop and your parking is free.
One part solving extremely complicated math word problems. (algorithm creation)
One part extreme proof reading for a grammar nazi. (debugging)
One part playing charades or pictionary with the worlds most inept clue giver. (gathering requirements from clients)
It isn't clear if the allegations are true. TFA says the accuser "walked back" some of his allegations, which is a euphemism for "admitted he was lying".
After he wrote the letter he was hired as a 'consultant' for 4.5 million dollars, of which, it is paid out over time and he has thus far recieved 1 million (a significant percentage of the payment is due at the end of the contract). If he testifies that everything in the document is true, he will likely be fired and not get 3.5 million dollars, so he has 3.5 million reasons to say that the document isn't accurate.
It has long been known that memory of the arbitrary is uncorrelated with IQ.
To be fair - technologists nearly always know a hell of a lot more about whatever the columnist is pretending to be knowledgeable about.
It isn't in GANs one model, the generative, tries to fool the other, classifier one, by giving it images itself has generated. There's no incremental task switching from easier problem domains to more difficult ones.
I'm familiar with GANs - what it sounded like (and what they did) is add curriculum learning, but they also did it layerwise as is done with autoencoders, (Also they had some other interesting ideas, but that was the crucial bit). In this case the easy is the lower resolution images and the hard is the higher dimensional images.
From their paper
The idea of growing GANs progressively is related to curriculum GANs (Anonymous), where the idea is to attach multiple discriminators that operate on different spatial resolutions to a single generator, and furthermore adjust the balance between resolutions as a function of training time. That work in turn is motivated by Durugkar et al. (2016) who use one generator and multiple discriminators concurrently, and Ghosh et al. (2017) who do the opposite with multiple generators and one discriminator. In contrast to early work on adaptively growing networks, e.g., growing neural gas (Fritzke, 1995) and neuro evolution of augmenting topologies (Stanley & Miikkulainen, 2002) that grow networks greedily, we simply defer the introduction of pre-configured layers. In that sense our approach resembles layer-wise training of autoencoders (Bengio et al., 2007).
This sounds like the standard idea of curriculum learning - you teach NNs via progressively more difficult tasks.
It takes more than a Format command to destroy data on a hard drive (the age of the computer means spinning platters), because the erase head never perfectly aligns with the written data, there is always a little bit missed. A good data recovery company should be able to recover everything.
They techs did a triple degaussing, even the NSA is unlikely to be able to recover anything.
They used character by character training, rather than words tagged with word sense.
More sophisticated 'word sense' tagging can be done which can differentiate between all of the various meanings of gay, but for most words there isn't enough data to train all of the word senses sufficiently.
Tesla Motors was created for the purpose of proving EVs were viable and proliferating them, so this would be a win for Elon. If you haven't you noticed, all their patents are free to use.
Only for those companies that reciprocate and make all of their patents free to use for Tesla.
Fair use is far broader than that - 'transformative use' - (such as using a book as input to a machine learning algorithm) is one of many additional fair use defense.
Sure, there exist a computer that, in a clearly rigged contest (the computer knew games of the player, but not the other way round), defeated a really good Go player.
You have it backwards. AlphaGo knew literally nothing about its opponents and hadn't been trained on any of its opponents games. The opponents (Lee Sedol and the much more recent match versus Ke Jie) had access to a small number of previous Go games played by AlphaGo.
You must have missed the memo reporting that a lot of these ads were plugging Black Lives Matter, Hillary's widespread support, and similar topics.
It matters who those ads were targeted at. If those ads were targeted at Republicans - it would tend to increase turnout for Republican voters who feared Clinton winning.
I studied AI algorithms back in University a decade ago and they really haven't changed much from now till today. The biggest improvements have come from having faster computers, not more efficient or even effective algorithms.
Actually a number of algorithmic improvements were needed to train deep networks, so it isn't just hardware, some major algorithmic improvements have occurred also. Also GAN's certainly weren't around a decade ago.