Slashdot Mirror
Video Games Are So Realistic That They Can Teach AI What the World Looks Like (vice.com)
Jordan Pearson, reporting for Motherboard:Thanks to the modern gaming industry, we can now spend our evenings wandering around photorealistic game worlds, like the post-apocalyptic Boston of Fallout 4 or Grand Theft Auto V's Los Santos, instead of doing things like "seeing people" and "engaging in human interaction of any kind." Games these days are so realistic, in fact, that artificial intelligence researchers are using them to teach computers how to recognize objects in real life. Not only that, but commercial video games could kick artificial intelligence research into high gear by dramatically lessening the time and money required to train AI. "If you go back to the original Doom, the walls all look exactly the same and it's very easy to predict what a wall looks like, given that data," said Mark Schmidt, a computer science professor at the University of British Columbia (UBC). "But if you go into the real world, where every wall looks different, it might not work anymore." Schmidt works with machine learning, a technique that allows computers to "train" on a large set of labelled data -- photographs of streets, for example -- so that when let loose in the real world, they can recognize, or "predict," what they're looking at. Schmidt and Alireza Shafaei, a PhD student at UBC, recently studied Grand Theft Auto V and found that self-learning software trained on images from the game performed just as well, and in some cases even better, than software trained on real photos from publicly available datasets.
What side do you want?
China
France
Russia
UK
USA
India
Israel
Pakistan
North Korea
Although you are right about the poor quality of what passes for "research" nowadays, people need to realize that there are limits to technology. It doesn't increase in power and get better and better indefinitely. There are limits based on Physics and engineering. Particularly with digital computers we are hitting those limits now. CPUs in particular are only getting marginally faster with each (expensive) generation.
AI doesn't think at all. What we call "AI" is really just a bunch of algorithms.
Good. We need AI that can correctly identify a person as a prostitute and various fictional weaponry and automobiles.
Why is this so strange to you? Virtual worlds have been used for simulating inputs into control systems and the like since at least the times of Apollo (feeding simulated data into the AGC back then).
Ezekiel 23:20
The SR-71 is still the fastest plane ever built.
That doesn't mean that it's not possible to build a faster plane. There's no reason to build one at this point. You don't need a plane capable of outrunning any missile when you can take pictures with a satellite instead. You don't even need a plane that fast for carrying people when we haven't even attempted to handle the legislation involved with supersonic passenger travel. We didn't lose any capability after the SR-71 was retired, the capability had already been replaced by the time the last ones were laid up.
And, there most definitely are faster aircraft than the SR-71. The SR-71 is the fastest manned air-breathing aircraft. There are faster manned aircraft that are not air-breathing (mach 6.7 X-15), and there are faster air-breathing aircraft that are not manned (mach 9.6 X-43).
The SR-71 met a specific goal at a certain time and performed exceedingly well at that mission, but we haven't replaced it with a better plane because we don't need one. That's like saying that naval scientists or shipbuilders have lost their art because the last battleship was launched in 1944. We just don't need them anymore.
"Our two-party system is like a bowl of shit looking at itself in a mirror." - Lewis Black
No, both of you missed the point. Simulating stuff like this is cheaper, faster, simpler, therefore more efficient than working with the physical world. It is the physical alternative that would cause more global warming than this.
Ezekiel 23:20
We tried neural nets back then. Didn't work.
It seems it is you that are stuck in 1960, because connectionist techniques nowadays are nothing like that. And deep-learning has been breaking record after record, even achieving superhuman performance in some tasks. And deep-learning is just a smart math trick over the regular backprop algorithm, allowing more layers without degrading the error gradients. When the models that actually incorporate neuroscientific knowledge (current research) mature, expect even better performances.
And wtf, this "I tried once, I failed, I'll never try again" is surely a loser-talk.
`echo $[0x853204FA81]|tr 0-9 ionbsdeaml`@gmail.com
An interrogatory sentence should end with a question mark.
"Our two-party system is like a bowl of shit looking at itself in a mirror." - Lewis Black
Now the fleets of self driving cars will be sipping "hot coffee" and beating the hookers to collect their money back for services rendered.
They will also know how to rack up a respectable 5 start wanted level and just hide under a bridge till it all goes away.
Oh, and my favorite new autonomous feature? Spawning attack choppers out of thin air to aid in robbing every convenience store in the city while performing the above two tasks.
You are being ripped off every second of every day, so that advertisers can help rip you off even more tomorrow.