Hat tip to my brother Jared for the link.
Puzzled by the intricate structure of the M-PMV retroviral protein… scientists have striven to find its chemical key for ten years now. Each enzyme has millions of possible combination in which it can fold its atom bonds, and determining its precise structure is a very laborious enterprise even for high-end computers with large processing power.As a long-shot University of Washington biologists sent the virtual 3D model of the M-PMV to the online game Foldit, where gamers folded and turned it into a myriad of combinations. Eventually, and remarkably enough, the gamers obtained the optimum one – the state that needed the lowest energy to maintain….
“They actually did it in less than 10 days.”
The reason why computers haven’t been able to do this, despite their evidently superior processing capabilities, is that they’re still far from being capable of having human-like spatial reasoning.
I buy that argument as part of it. I also wonder about the characteristics of the neural super-net that is collaborative human interaction.
Interestingly enough, Foldit records the players’ actions and processes them in an algorithm which will eventually help the AI behind the game to someday be able to compile successful structures on its own.
Well, whether you buy that might depend whether you believe that algorithms can generate strong AI, or not.