7 ms·
Part of the problem is... games have explicitly defined rules, start and end points, boundaries, and discrete "win" and "loss" states (and sometimes "draw"). If
by skywhopper 9y ago
Part of the problem is... games have explicitly defined rules, start and end points, boundaries, and discrete "win" and "loss" states (and sometimes "draw"). If the game itself (ie, all the rules including the ability to judge "win", "lose", or "draw") can be easily represented in a simple computer program, we shouldn't be surprised that a complex computer program can master the game.
The real world is not a finite problem with explicit rules, obvious boundaries, well-known start conditions, or any way to judge a specific situation as "win", "lose", or "draw". But, even if you want to argue that specific tasks can be broken down this way, you still have to be able to represent this subset of reality in the computer, before AI magic can even begin to work on the problem.
- dmreedy 9y agoPrecisely so. You can, given enough time, brute force your way to victory in any game of perfect information. This is not the case with reality, as far as we understand it so far. Thus, from a theoretical perspective, the class of problems that contains all games is strictly easier than the class of problems that exist in a less artificially constrained environment.
- Agebor 9y agoPerfect information is not the case when the AI is reading pixel colors off the screen and pressing keys on the keyboard. This is very like the real-world situation. As for goals, they are indeed better defined in games. But we can create artificial goals attached to the real life: - goal: GPS sensor reports specific position - goal: the camera reports a recognised object in a specific position - goal: a button was pressed (signal reaching the AI e.g. through the cloud)
- dmreedy 9y agoEven then, it still reduces to a brute-forceable game with finite, explorable states; 'just' with an extra layer of (granted, quite interesting and technically impressive) parsing. We don't know what the case is for reality. And yes, we can do our best to turn real life in to a game, but all such models leak pretty badly, and the leaks tend to cause much more fundamental instability.
- Agebor 9y agoI think the main problem is with simulating enough games/tries. In the game, you can easily let the AI play against itself countless times. Much, much slower in real life. Until we have good enough simulators/approximators of reality, AI can't learn fast. However, they can already learn driving in GTA, so who knows?
- _emacsomancer_ 9y ago> However, they can already learn driving in GTA, so who knows? Yeah, that sounds like it'll end well.
- deong 9y agoYou'll find pretty quickly that the exact techniques that worked so well in learning to play Atari games very often fail spectacularly when you have to introduce goal-steering. Reinforcement learning turns out to be fantastically clever at finding really stupid solutions if you give it the tiniest opening to do so. You put a camera in a room to provide feedback for an agent to learn to move an object closer to the camera, and it will happily learn to knock the camera over.
- AnIdiotOnTheNet 9y agoTo be fair to our AI brethren, this is true of Humans too. I watch people game metrics every single day, and many of them fail spectacularly when presented with real world problems outside of their experience.
- jacobush 9y agoGaming metrics is the very soul of corporate life
- _emacsomancer_ 9y agoThen which is real and which illusion?
- jacobush 9y agoGood question. Sometimes, the metrics must be gamed to make real progress. (As opposed to the more usual gaming them for personal gain.) Between these two, some self serving bias and delusion, mission statements etc.. what is real?!
- taneq 9y agoIf they meet the spec, they're not stupid solutions, it's a stupid spec.
- candiodari 9y ago
- taeric 9y agoI think you are misusing the term "perfect information." In this case, chess is a real world game where both players have perfect information. That is, neither player can keep a secret of where any of the pieces have moved. So, the complexities of how the AI is taught to interact ultimately don't matter. It may have a lot of effort to parse the visual of the board to get the perfect information, but the game is defined as one of perfect information. Wikipedia calls out that there does not seem to be consensus on what this term means for games with chance or concurrent play. https://en.wikipedia.org/wiki/Perfect_information https://en.wikipedia.org/wiki/Perfect_information
- Agebor 9y agoTrue. I think my point was more about games with non-trivial rules (or many degrees of freedom). For example going from chess/go to turn-based video games like civ, to starcraft. Usually it involves vastly more possibly positions in time and space.
- cortesoft 9y agoYep, this is pretty much word for word what the article is laying out.
- ttflee 9y agoReal world is so damned complicated, full of various mini games. How about using a MMORPG as a naive start point?
- digi_owl 9y agoAnd all too often most of the players will try to bend/break the rules in their favor.
- rollo 9y agoBecause that is indeed a very effective approach. In a game where rules can be broken or warped to suit your goals, doing so is a smart move. Just like in life, where rules don't actually exist.
- _emacsomancer_ 9y agoI've been stuck in a minigame for the past couple of decades.
- tinymollusk 9y agoWhat if closed systems are just that much different than open systems that it's not generalizable? Closed systems don't have Black Swan possibilities. Black swans are sometimes super-linear, jolting results in directions and scale unimagined. To answer briefly (and respectfully): because an MMORPG is a different type of system than reality. Comparing cars to buses.
- tonysdg 9y ago"Greetings, Professor Falken." "Hello, Joshua." "A strange game. The only winning move is not to play. How about a nice game of chess?"
- jacobush 9y agoThe ”fux dat”, or stoic approach.
- FranOntanaya 9y agoI would also bet nobody publishes research on games AI didn't perform well at.
- nerdponx 9y agoIt's a question of generalizability. Today's "AI" algorithms are intensely overfitted to their problem domains, even if they generalize well within those domains.
- Raphmedia 9y ago"AI plays Dwarf Fortress" is something I'd like to see.
- TuringTest 9y agoHow do you define a "win" condition in Dwarf Fortress to train the AI?
- muthdra 9y ago"Avoid dwarven deaths"
- TuringTest 9y agoDoesn't that get you dwarven overpopulation and sad dwarves?
- platz 9y ago"maximize happiness according to Marxist values. Allow workers to control and maximize the means of production"
- ben-schaaf 9y agoAvoid Death doesn't mean reproduce.
- repsilat 9y agoDepends on whether it's the death of the society or the dwarves you're optimising for, I guess? I don't know the game well enough to say, but suicide could be the most reliable way to minimise death. I prefer "maximise happiness" if only because the Repugnant Conclusion is more interesting than the VHEMT. https://en.wikipedia.org/wiki/Mere_addition_paradox https://en.wikipedia.org/wiki/Mere_addition_paradox https://en.wikipedia.org/wiki/Voluntary_Human_Extinction_Movement https://en.wikipedia.org/wiki/Voluntary_Human_Extinction_Mov...
- yogrish 9y agoLimitations of AI reminds me of "Moravec's paradox" https://en.wikipedia.org/wiki/Moravec%27s_paradox https://en.wikipedia.org/wiki/Moravec%27s_paradox As Moravec says, "it is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility."
- mgoetzke 9y agounless AI figures out the way to win life is by figuring out how to survive