MIT alumnus and entrepreneur Ben Vigoda took his probability-processing technology to market with help from the Institute.
Helping computers navigate ambiguity
Vigoda’s group is creating computer chips that perform inferences and machine learning on uncertain data — data that can be incomplete or contradictory — more efficiently than today’s chips.
“If a normal computer program receives an unanticipated or noisy input, it will ordinarily either give the user an error message, crash the program or even, in some rare cases, crash the machine,” Vigoda says. “With probabilistic processing, the hope is to help the computer directly understand that the world is noisy, ambiguous, or even contradictory, and to be able to cope with that in a more native way.”