8 ms·
DeepMind’s work in 2016: a round-up
- merqurio 10y agoThey had a vast presence on EWRL, it's really amazing how many effort goes into research. Really inspiring.
- WhitneyLand 10y agoWhat indeterminate games can AI can beat humans at? Backgammon - AI usually beats the best humans Ms. Pac Man - AI loses to almost any human
- john_reel 10y agoWhat makes Ms. Pac Man so hard for AI?
- tim333 10y agoI'm not sure it's hard in general but deepmind's program didn't do well because it couldn't plan ahead https://www.technologyreview.com/s/535446/googles-ai-masters-space-invaders-but-it-still-stinks-at-pac-man/ https://www.technologyreview.com/s/535446/googles-ai-masters...
- varelse 10y agoIt's also inferring the game state from looking at the screen rather than being spoonfed the data. One could conceivably perform a depth-limited search on the actual game state if it were available and then use an AlphaGo-like DNN to predict what a deeper search would find, no?
- tim333 10y agoDunno really. A human would watch the ghosts behaviour and guess their likely future behaviours based on that. I'm not sure if the software gets that, as it were, or how you'd tweak it to do so.
- WhitneyLand 10y agoIts indeterminate because the ghosts turn in random directions with no preserve patterns. Notice Go, Chess, etc all are determinant. This paper gives an overview of the background: http://www.cse.unsw.edu.au/~mit/Papers/AAAI10a.pdf http://www.cse.unsw.edu.au/~mit/Papers/AAAI10a.pdf
- antognini 10y agoAI tends to perform poorly for many real-time games because they constitute many, many more "moves" than a game like backgammon. A backgammon game will typically take < 100 moves. For a game like Ms. Pac Man the AI will have to make that many decisions in a timespan of about four seconds. When you train a reinforcement learning algorithm, the AI has to update the parameters that were used to make several thousand decisions, and the only information it has to go on is its score at the end of the game. (This is known as the credit assignment problem --- of all the decisions you made in the game, only a few probably had a substantial impact on the outcome. How do you decide which ones were the important ones?) A game like go is hard because there are many possible moves every turn. But the games themselves are not very many turns, so it is easier to find the important moves, because in some sense all the moves were more important. A game like Ms. Pac Man is hard because, while you only have 5 moves each turn, the games themselves are many, many turns, and it is very difficult for the NNs to learn these long-term dependencies.
- WhitneyLand 10y agoSince a child could easily outperform the best AI here, doesn't it imply we need a different approach? Otherwise we need 4 orders of magnitude more compute, and it's just hard to believe their is no better way.
- haimez 10y agoNo, just means that someone needs to bother to make an AI for Ms. Pacman. Handling random ghost movements is only trivially confounding for AI learning. Low value, low yield.
- antognini 10y agoIt's true that someone could pretty easily hand-craft an AI that plays Ms. Pac-Man pretty well. But that's not really the point. Because after all that work, you still only end up with a computer program that can play Ms. Pac-Man. It won't play Pong, it won't play Doom, it won't even be able to play regular Pac-Man! The goal is to build a general system that can learn how to play specific games on its own. So, there is potentially high value to developing a Ms. Pac-Man AI, but only if it's not really a Ms. Pac-Man AI, but actually a more general AI, for which playing Ms. Pac-Man is a specific instance. I agree with the parent comment that some sort of new approach is probably needed. The main issue with reinforcement learning is just that the algorithm gets too little feedback. This ends up discarding an enormous amount of data on each run (or, more accurately, learning from that data extremely inefficiently). There's a huge amount of data present, but the learning from that data has to be unsupervised. A child can do it easily --- they only have to play Ms. Pac-Man once or twice to get the hang of it. Computers currently cannot. Getting them to do that will definitely require something different, but no one is quite sure what.
- sherjilozair 10y agoNote that most of recent work in reinforcement learning focus on creating a model which works on all or most of Atari games without using any game-specific knowledge. If researchers were to focus on only Ms. Pac Man, and ignore the "no game-specific tuning" rule, then Ms. Pac Man, and most Atari games would be solved to near perfection. The challenge that the Atari suite provides is not their individual difficulty, but the difficulty imposed by their diversity, i.e. finding a general algorithm which works well on all Atari games is hard.
- WhitneyLand 10y agoI don't know why so many people are suggesting this is easy. Many other attempts has been made targeted specifically against Ms. Pac Man and none have come anywhere near beating an average human player.
- varelse 10y agoConventional AI can clean up at Pacman if it can do tree search on the game state, source: CS188 out of Berkeley. http://ai.berkeley.edu/lecture_videos.html http://ai.berkeley.edu/lecture_videos.html
- WhitneyLand 10y agoPac Man is easy. Ms. Pacman has been tried and failed by many people who know CS188 material.
- varelse 10y agoDo these attempts make use of the documented idioscynracies of Ms Pacman? http://gameinternals.com/post/2072558330/understanding-pac-man-ghost-behavior http://gameinternals.com/post/2072558330/understanding-pac-m...
- gallerdude 10y agoI'm excited for WaveNet to get faster/more accessible. I'm not very technically inclined when it comes to downloading things of github and hacking them together. My goal is audio books - I'd love to hear them read by my favorite movie characters.
- bobajeff 10y agoI can't wait to have Peter (Geoffrey Francis) Jones read me all Wikipedia articles.
- conistonwater 10y ago> In its ability to identify and share new insights about one of the most contemplated games of all time, AlphaGo offers a promising sign of the value AI may one day provide, and we're looking forward to playing more games in 2017. What I find fascinating is how different AlphaGo's impact was from the impact of early chess engines. Once chess engines became decent (not even good—before even Deep Blue), they identified tons of inaccuracies in published chess literature. These were missed tactics, hard-to-see moves, requiring only relatively shallow calculation but a computer's precision. These inaccuracies were found even in classical annotations of well-known chess games, as well as standard books about openings. I believe John Nunn was known for this kind of work. AlphaGo hasn't achieved nearly the same impact. Have they even tried to identify the same types of inaccuracies in classical Go books? Can you imagine how absolutely cool it would be for a go engine to find errors in Invincible? Maybe they tried, but didn't find any inaccuracies, so now this negative result is sitting in one of their file drawers? I really wish they were more active with this sort of thing.
- Jabbles 10y agoOnce chess engines became decent (not even good—before even Deep Blue), they identified tons of inaccuracies in published chess literature. Within one year of their invention?
- conistonwater 10y agoWell, AlphaGo beat Lee Sedol, so I think it wouldn't be fair to compare timelines so directly. With chess engines, the compute power just wasn't there in the beginning, but it's all already there for go. Also, speaking as an optimist, yeah! why not? To be honest, my suspicion is that they aren't trying.
- sqeaky 10y agoI suspect this is a visibility problem. Go is a more popular game in the East I bet there are a few grad student level problems doing just what you described at Tokyo U and other prestigious Schools far from the West.
- rattray 10y agoI can't believe this one passed under my radar: https://deepmind.com/blog/deepmind-ai-reduces-google-data-centre-cooling-bill-40/ https://deepmind.com/blog/deepmind-ai-reduces-google-data-ce... > Our partnership with Google’s data centre team used AlphaGo-like techniques to discover creative new methods of managing cooling, leading to a remarkable 15% improvement in the buildings’ energy efficiency. Which itself is an understatement of the achievement: > Our machine learning system was able to consistently achieve a 40 percent reduction in the amount of energy used for cooling, which equates to a 15 percent reduction in overall PUE overhead after accounting for electrical losses and other non-cooling inefficiencies. It also produced the lowest PUE the site had ever seen. Basically, they built an AI that was able to tune the "large industrial equipment such as pumps, chillers and cooling towers" to react to the dynamic, nonlinear interactions that vary within and between datacenters (weather, utilization, etc). They also describe this AI as "General": > Because the algorithm is a general-purpose framework to understand complex dynamics, we plan to apply this to other challenges in the data centre environment and beyond in the coming months. They seem to imply that this technique could make almost any industrial process more efficient with minimal oversight/training/customization.
- forgotmyhnacc 10y agoI remember reading the blog post. Did they ever follow up with a research paper explaining what they actually did?
- gwern 10y agoSadly, not yet. Might be too secret sauce even for Google.
- conistonwater 10y agoNonlinear control with neural networks has existed for a long time, though, it's not exactly new. Is there reason to think they did something profoundly new in that specific instance?
- malcolmgreaves 10y agoDoes anyone know of any other companies that are similar to DeepMind, but based in SF?
- wrsh07 10y agoOpenAI? (It's a non-profit, not sure if you personally find that truly disqualifying)
- sherjilozair 10y agoOpenAI (https://openai.com/ https://openai.com/). OpenAI's research vision is at a similarly ambitious scale as DeepMind. They're also relatively more open in their research, and known to release papers, code, and models quickly. Currently, they are not as big though, more like how DeepMind was immediately after publishing their first Atari work in 2013. Give them time, and I expect them to become a very formidable opponent to DeepMind.
- deepnotderp 10y agoAs a bonus, OpenAI is a nonprofit that always allows publishing. They also have some of the top names in DL: Ian Goodfellow, Ilya Sutskever and Zaremba. In DRL (deepmind's arena), they have Abbeel and Schulman, both absolute houses.
- gdahl 10y agoThe Google Brain (g.co/brain) team has people in SF. We are part of the same company as DeepMind so maybe this doesn't quite answer your question. ^_^ We are mostly in SF and Mountain View, but we also have people in a few other locations. Right now, SF and Mountain View are the largest. Disclosure: I work for Google on the Brain team.
- doobwa 10y ago> We’re still a young company early in our mission... What? Why don't they consider themselves part of Google?
- sanxiyn 10y agoI don't know why, but it's very clear they consider themselves somewhat independent of Google. The most clear sign I have seen is that Partnership on AI site has both Google logo and DeepMind logo. https://www.partnershiponai.org/ https://www.partnershiponai.org/
- raverbashing 10y agoProbably by design. I assume Google wants to have them doing their own thing instead of just borg-similating it. Also they have Google brain, which is doing similar things