We’re making a survey. I hope to write soon about our general methods and plans, so anyone kind enough to criticize them has the chance. Before that though, we have a different request: we want a list of concrete tasks that AI can’t do yet, but may achieve sometime between now and surpassing humans at everything. For instance, ‘beat a top human Go player in a five game match’ would have been a good example until recently. We are going to ask AI researchers to predict a subset of these tasks, to better chart the murky path ahead.
We hope to:
- Include tasks from across the range of AI subfields
- Include tasks from across the range of time (i.e. some things we can nearly do, some things that are really hard)
- Have the tasks relate relatively closely to narrowish AI projects, to make them easier to think about (e.g. winning a 5k bipedal race is fairly close to existing projects, whereas winning an interpretive dance-off would require a broader mixture of skills, so is less good for our purposes)
- Have the tasks relate to specific hard technical problems (e.g. one-shot learning or hierarchical planning)
- Have the tasks relate to large changes in the world (e.g. replacing all drivers would viscerally change things)
Here are some that we have:
- Win a 5km race over rough terrain against the best human 5k runner.
- Physically assemble any LEGO set given the pieces and instructions.
- Be capable of winning an International Mathematics Olympiad Gold Medal (ignoring entry requirements). That is, solve mathematics problems with known solutions that are hard for the best high school students in the world, better than those students can solve them.
- Watch a human play any computer game a small number of times (say 5), then perform as well as human novices at the game without training more on the game. (The system can train on other games).
- Beat the best human players at Starcraft, with a human-like limit on moves per second.
- Translate a new language using unlimited films with subtitles in the new language, but the kind of training data we have now for other languages (e.g. same text in two languages for many languages and films with subtitles in many languages).
- Be about as good as unskilled human translation for most popular languages (including difficult languages like Czech, Chinese and Arabic).
- Answer tech support questions as well as humans can.
- Train to do image classification on half a dataset (say, ImageNet) then take the other half of the images, containing previously unseen objects, and separate them into the correct groupings (without the correct labels of course).
- See a small number of examples of a new object (say 10), then be able to recognize it in novel scenes as well as humans can.
- Reconstruct a 3d scene from a 2d image as reliably as a human can.
- Transcribe human speech with a variety of accents in a quiet environment as well as humans can.
- Routinely and autonomously prove mathematical theorems that are publishable in mathematics journals today.
Can you think of any interesting ones?