AI Project Funded By Google Wants to Teach Computers Regret

Computer researchers at Tel Aviv University are working on a project funded by Google that aims to teach computers how to minimize “regret” or, in other words, to learn from their decisions and make better ones next time.

This foundational research could improve efficiency in many branches of computer science, says Yishay Mansour, a professor at Tel Aviv University’s Blavatnik School of Computer Science. The machines would be able to learn and improve tasks like packet routing, load balancing and prioritizing server resource requests by being able to evaluate all the relevant variables in advance and make the best possible decision.
“We are able to change and influence the decision-making of computers in real-time. Compared to human beings, help systems can much more quickly process all the available information to estimate the future as events unfold — whether it’s a bidding war on an online auction site, a sudden spike of traffic to a media website, or demand for an online product,” says Mansour.
Computers don’t actually feel “regret” — or at least we think they don’t — but they can measure the distance between a desired outcome and the actual outcome, which can be interpreted as “virtual regret.” Mansour has developed an algorithm that minimizes the amount of computer “regret,” adapting to the situation and learning as it’s running. The result, claims Mansour, is “almost as if you knew all the variables in advance.”
It’s easy to see why Google would be interested in such a project. As a company that handles immense amounts of data, any increases in efficiency can be highly beneficial to the search giant. A 20-person team will work on the project together with Google, headed by Mansour and Noam Nisan, a professor at Hebrew University.
Image courtesy of iStockphotokemie

{ 0 comments... read them below or add one }

Post a Comment