Slashdot Mirror
The AI That Has Nothing to Learn From Humans (theatlantic.com)
An anonymous reader shares a report: Now that AlphaGo's arguably got nothing left to learn from humans -- now that its continued progress takes the form of endless training games against itself -- what do its tactics look like, in the eyes of experienced human players? We might have some early glimpses into an answer. AlphaGo Zero's latest games haven't been disclosed yet. But several months ago, the company publicly released 55 games that an older version of AlphaGo played against itself. (Note that this is the incarnation of AlphaGo that had already made quick work of the world's champions.) DeepMind called its offering a "special gift to fans of Go around the world." Since May, experts have been painstakingly analyzing the 55 machine-versus-machine games. And their descriptions of AlphaGo's moves often seem to keep circling back to the same several words: Amazing. Strange. Alien. "They're how I imagine games from far in the future," Shi Yue, a top Go player from China, has told the press. A Go enthusiast named Jonathan Hop who's been reviewing the games on YouTube calls the AlphaGo-versus-AlphaGo face-offs "Go from an alternate dimension." From all accounts, one gets the sense that an alien civilization has dropped a cryptic guidebook in our midst: a manual that's brilliant -- or at least, the parts of it we can understand. Will Lockhart, a physics grad student and avid Go player who codirected The Surrounding Game (a documentary about the pastime's history and devotees) tried to describe the difference between watching AlphaGo's games against top human players, on the one hand, and its self-paired games, on the other. According to Will, AlphaGo's moves against Ke Jie made it seem to be "inevitably marching toward victory," while Ke seemed to be "punching a brick wall." Any time the Chinese player had perhaps found a way forward, said Lockhart, "10 moves later AlphaGo had resolved it in such a simple way, and it was like, 'Poof, well that didn't lead anywhere!'" By contrast, AlphaGo's self-paired games might have seemed more frenetic. More complex. Lockhart compares them to "people sword-fighting on a tightrope."
stones are REAL PEOPLE. Be afraid.
Can't even.
I think this teaches us a great deal about what AI will actually be like when it inevitably arrives. It won't be r2d2 or c3p0 or data - it will be an alien mind that will be incomprehensible to the rest of us.
Just looking further ahead. If it knows the skill level of a lesser opponent, it can try for bigger rewards which would seem less mysterious. Instead, it has to assume an equal opponent and take the safest path which will seem strange to a lesser player if another path looks more rewarding.
To infect Viruses to A.I.'s robots "for learning" how to kill them each other.
It was only a few years ago people were saying that the best Go computers would never beat human players because the game was so much more complex. We're getting to the point where AI decisions, even when explained, end up being too complex for humans to follow. This is a scary path we are following.
I read the internet for the articles.
AI hasn't really changed much in years, computers have gotten more powerful so some pattern recognition tricks that use to take more effort don't anymore. The big breakthrough was figuring out how to write a decent AI to play Go optimally against modern equipment. After that all they did was create random legal scenarios to play against the system creating patterns that are unusual to us and using the results of those outcomes to fine tune the AI algorithm. There really isn't that much magic if you've taken some AI courses.
Think about fractals, some of the resulting art or images generated from some relatively simple equations create things which we could have never imagined.
AI will be obsessively interested in one specific subject like the computer nerds who visited slashdot before the social media craze turned the tech industry to shit.
Here's an interesting thought. Would these people still say the same thing about the games if they were told there were AI games, but in reality were actually games by two human players?
It reminds me of the recent story where some kids put a pineapple in an art exhibition as a joke and people thought it was art. Most people will believe and/or spew pure bullshit if they think it's what's expected of them.
Let's play global thermonuclear war.
What side do you want??
1. USA
2. North Korea
3. Show full list
experiments they did years ago. They used feedback neural networks or something to design FPGA circuits. The results were very strange and difficult to understand, like relying on stray capacitance between pads and wires or something like that.
The AI That Has Nothing to Learn From Humans
Someday (soon?) that will be all of them.
It must have been something you assimilated. . . .
I've been looking forward to an entirely self-taught version of AlphaGo for a while now. I imagined that it would have invented all sorts of crazy josekis (local opening patterns). The rules of go are so simple and natural that there's this speculation that if extraterrestial civilizations exist, they have invented the game as well. AlphaGo Zero then is the closest thing to playing some species that has had no contact with humanity, so I had hoped for some really alien looking gameplay. .. Nope. AlphaGo Zero actually plays more like a human than the one seeded with human games.
They have actually released a trove of game records from the newest version, with two actually commented by a professional:
http://www.alphago-games.com/
The story I heard was a guy experimenting with self-programming FPGA, just like you said. But one of designs had a bunch of gates completely unconnected to anything. He removed them and... it stopped worked. Turns out the FPGA had a hardware flaw and the self-learning mechanism incorporated the flawed behavior into it's design.
"Lockhart compares them to "people sword-fighting on a tightrope."
If you see that you lose, you can always cut the rope.
https://www.damninteresting.co...
They wanted a circuit that detected between two frequencies. The system was supposed to be digital, but the artificial evolution had made use of analogue computing using magnetic fields and harmonics.
Vintage computer adverts: http://www.vintageadbrowser.com/computers-and-software-ads
If it could understand a human enough to play at exactly their level, that would be more impressive. Or anticipate what their weaknesses were and train them to beat those weaknesses by turning the came a certain way.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
WTF?!?!? I just quickly scanned the comments and NONE that look like Russian trolls. Ivan, Vlad and Boris, stop fucking slacking and get back on the job.
https://games.slashdot.org/sto...
The rules of Go to begin with. Humans gave it a concrete objective. It is really cool how it finds its own ways to work toward that objective, but it's not *that* surprising that it can do so.
The rules of Go to begin with. Humans gave it a concrete objective. It is really cool how it finds its own ways to work toward that objective, but it's not *that* surprising that it can do so.
I see Slashdot's AI has modded you down for dissing it ;)
It might not have training from the general population, but it will be about as smart. Way to go! Now we won't have to deal with AI systems becoming racist within 24-hours after being introduced into the wild. This is definitely progress. Hopefully it will be more likely to mirror the ideals of our new overlords. Much better than our old overlords.
but can it make toast?
...that AI is in it's adolescent stage? What happens when it hits its teenage years? We ask it to do some real work and it refuses? Turns on game system of choice while not showering for days?
... give us a chance to see how much our best experts actually still suck at their field.
Humans have feelings, a body, complex interaction between biology, culture and mind. AI doesn't have that. It just plods away at the problem, not looking right or left. Not being able to look right of left. Interviewing dendi as he lost against AI at the Dota international was possible, the AI wouldn't have been able to do that.
My suspicion is that self-reproducing AI will be something like a swarm of robot-insects, fadcinating and highly sophisticated, but somewhat outlandish to squishy mamal humans with a heaetbeat and feelings and other "useless" stuff.
My 2 biped mamal cents.
We suffer more in our imagination than in reality. - Seneca
> it's not *that* surprising that it can do so.
"in 2014, the common consensus was that an AI victory in Go was still decades away. "
https://venturebeat.com/2017/06/10/why-are-ai-predictions-so-terrible/
Of course! If we could ask primates what they think the next step of evolution would look like, they probably would answer: The next kind of primates will be better in jumping from tree to tree, but language, music, art? Makes no sense. Who would want THAT?" But the next step of evolution has never been just a quantitative one, but a qualitative one (think dialectically like Hegel: There is always a turn-over of quantity into quality). Therefore, the next step of evolution "after us" will not be a "superhuman" with some more power in some of our abilities, but a DIFFERENT one, maybe not even including some of our core abilities – as we do with tree-jumping. I think, the next step of evolution (and no-one will be able to stop evolution for ever) will be a machine-like intelligence, capable of adapting itself (including its "intelligence") very fast to changes in its environment, e.g. for conquering space. It makes no sense for evolution to wait until a nex generation is born and will be (by accident or gene manipulation) adapt to certain environments, as well as it makes no sense to carry your habitational environment with very specific conditions always with you in a space ship or so. We are just a step on the ladder of evolution, and surely not its last one. But we may be the last one that could be called (exclusively) biological. So, get along with that. ;-)
But I, personally, would not advance the development of an AI that surely will have no interest in sharing this planet with us, maybe it/they will let some of us survive in a Zoo But I am also convinced that capitalism will have no other way to maximize profits than go on and develop this kind of AI as soon as possible. So, even if we survive climate change, social catastrophes/revolutions, biological breakdown / mass extinction, nuclear war "by accident" (or intention, think MIK) – we will build our own end, definitely. So go on, enjoy your life as long as you can. Maybe we or the next generation will be the last to do that.
And yes: I am an optimist!!
The AI cannot explain what it's doing. In the same manner that a walking being cannot explain how to walk, or a bird cannot explain how to fly.
Its the same kind of intelligence that allows a fly to fly. Just a lot more powerful. It's not the kind of intelligence that understands, or takes conscience decisions.
So the previous version of AlphaGo replayed and learned from the largest body of human plays ever, and trained against itselfas well, then on to also do some heuristics as a complement. Now this new version plays again that version "from scratch" with no human insights? I am sorry Google, but if what is fine-tunning this new version is one that gathered how to play from a huge body and all best games ever recorded, then how is this one version not learning from humans?
So if I copy Mona Lisa from the best copy ever of Da Vinci's work, then the computer learned to paint Mona Lisa with no help at all from humans? I love Goodle's work, but misleading claims like this, and how the game against the world champion were setup stink quite a bit. All in all, I still like all the DeepMind and Brain guys and work, but stop the bogus claims and tell things how they are.
If you want AlphaGo to play against itself and learn to be the best without human help, it should have NO access to any game played by us, or anyone driving their knowledge from us.
And if the person went on to demonstrate that the number of reachable game states in Go vastly exceeds the same in chess, said person was speaking out of his or her butt hole. There's pretty much nothing stupider than a penis fight over the greater vastness, when the smaller vastness already exceeds your accessible light cone.
The curvature of viable game play in chess was fairly well understood, with a dominant term coming from piece assets. The curvature of viable game play in Go was not amenable to the same analysis (Go lacks a piece asset term as such).
Now that we have found a useful curvature, one suspects it will turn out to be a form of curvature that human players will never fully command. Perhaps worse than chess, judging by how fast the computers went from chumps to champs.
We really know very little about the kinds of curvature still hiding in the haystack. The chess curvature was painstakingly hand-constructed over many decades by human experts, typically hamstrung by ludicrously limited machines.
Recently, the vision field was much the same, dominated by painstakingly hand-constructed "feature" recognizers.
Now that computers are powerful enough to extract their own features automatically, we're going to go from expensive haystack exploratory surgery directly to haystack MRI.
Paradigm shift. It's a thing.
The inputs would be the PBs of porn humanity has made thus far.
The outputs are.. well.. I'm sure it'll get back to us.