You might have read earlier today about AlphaGo Zero
, a new system built by DeepMind that, in just 3 days, taught itself the game of Go well enough to trounce its big brother 100 games to zero. I don’t want to go crazy proclaiming this the biggest thing ever or the advent of the robot revolution (for a dose of skepticism, check out this story on AlphaGo Zero in IEEE Spectrum
), but it is very promising.
I think the ending of the DeepMind blog post on the system nicely sums up why this is:
These moments of creativity give us confidence that AI will be a multiplier for human ingenuity, helping us with our mission to solve some of the most important challenges humanity is facing.
While it is still early days, AlphaGo Zero constitutes a critical step towards this goal. If similar techniques can be applied to other structured problems, such as protein folding, reducing energy consumption or searching for revolutionary new materials, the resulting breakthroughs have the potential to positively impact society.
The key phrase there is structured problems. Because while it’s impressive that AlphaGo Zero was able to teach itself without learning by studying humans, the reality is still that systems like this need some rules to guide their exploration. A computer can’t understand a field like materials science with the same depth that a scientist can, but apparently a computer can do some pretty creative problem-solving given a specific task and some scientific guidelines/constraints.
Supercomputers have been tackling tough scientific problems for years, but they’re very expensive, heavily engineered systems. What makes AlphaGo Zero so inspiring is the speed at which it achieved its feat, as well as its relative efficiency. Whereas the original AlphaGo system ran on 176 GPUs, AlphaGo Zero needed just 4 of Google’s Tensor Processing Units (TPUs).
Business demands will undoubtedly lead companies to try and shove AI into every nook and cranny of their businesses, and some of those efforts will be more successful than others. But it seems plausible that the real economic benefits of AI will actually come from using it to solve problems and help make discoveries in fields like science and medicine, thus opening new opportunities and freeing up budgets in areas that don’t involve “business insights” or replacing humans with machines.