Yoshua Bengio: How AI Is Changing Scientific Discovery in Medicine

In a presentation that was part of the ML for Drug Discovery Summer School from Valence Labs, AI pioneer Yoshua Bengio, a Turing Award-winner, computer science professor at the University of Montreal and one of our scientific advisors, shared insights into how AI has fundamentally changed drug discovery.
Thanks to our ability to run large-scale experiments, Bengio notes, we’re able to collect massive quantities of data. Our own automated wet labs perform 2.2 million experiments per week.
“The amount of data we're collecting… is just too big for human brains to deal with,” Bengio says. “And so we need machines to help us make sense of that data.”
But, he notes, “the space of potential hypotheses to explain data or the space of potential candidates that we could experiment with that could be interesting, is huge… and only a very, very tiny fraction of these are good in some sense.”
This is where machine learning and active learning comes in, Bengio notes. “We can use machine learning in particular what's called amortized inference in order to help decide where to throw darts, whether it's in the space of hypotheses or the space of candidates for experiments. We're gonna take advantage of the ability of machine learning to generalize in order to make good guesses.”
Active learning allows us to continue this process in a constant flywheel of improvement, what we at Recursion call “cycles of virtuous learning.”
Watch the full talk here: • Day 3 - Exploring Mole...
#ai #ml #activelearning #drugdiscovery