Alex Wiltschko
speaker
1,050 appearances
3 recordings
3 series
first heard Mar 2025
last heard 21 Apr
Alex Wiltschko’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 2 in all, peaking in Apr 2026 with 1.
Appearances
And I remember the first time I experienced it, whoa, this fragrance changed. Has it gone bad? And then I just sprayed it again and again and again and watched this movie play out for like an afternoon. I was like, no, no, somebody made this. And this is the whole thing. The whole fragrance unfolds over time. And that kind of was the end of it for me. I got just completely hooked.
The way that my brain works, I wanted to understand where it came from, how it was made, how the brain processes it the way that it does. I think if I was born in Southern France, maybe I'd be a perfumer, but I was born to two academics, and so I became a scientist.
I went to school for neuroscience at the University of Michigan, and then realized that there is a subspecialty of neuroscience called olfactory neuroscience, so people who want to figure out how the brain processes smell. And the most people who study that are at Harvard.
So I went to Harvard and realized after many years of doing science there that actually we don't really know how smell works at all. We're making progress. We're learning things. But a simple question, let me draw a molecule on the whiteboard like we're in chemistry class. Can you look at that molecule and tell me what it's going to smell like? Will it smell like apple or cinnamon or anise or what?
So it turns out that's like a hundred year old problem nobody had been able to solve. that really stuck in my craw. I'm like, why don't we know how to do this? I ended up leaving academia. I started and sold two AI companies, one in the biotech space, one in the kind of pure ML as a service space. That ML company was bought by Twitter.
I helped to start their deep learning team with my co-founders and with another company that we were combined with. That's where I really learned like internet scale, artificial intelligence applications. So we... applied AI to their ads platforms and made them a lot of money, and applied AI to their data centers and saved them a lot of money.
And then I was recruited away to Google Brain, which is now called Google DeepMind, which is kind of their Xerox PARC or Bell Labs. And I worked on some internal projects for a bit, but after a year or so, I said, you know what? Let's take a crack at this smell problem again.
And it turned out that in the time between when I left academia and got into tech and entrepreneurship, and when I arrived at Google Brain, some breakthroughs happened. And what happened is AI researchers figured out how to make artificial intelligence work on chemistry. And that maybe doesn't sound too crazy, but up until then, AI systems really liked their inputs to be rectangles, right?
Like images are like grids of pixels and text is like a long thin string of words, but molecules can have any shape. There can be any number of atoms and the bonds can be all rearranged. And there had been a technique that had been really improved and figured out, made to work better, called graph neural networks. And that turned out to be like chocolate and peanut butter for AI and chemistry.
So we didn't figure that out, but a lot of my colleagues, who I was very fortunate to work with at Brain, they figured that out for the world of drug discovery. So the intellectual arbitrage that we did was we said, let's take those techniques and let's apply it to the realm of scent. And I had spent a long time thinking about scent and traveling in that world.
And so I knew where to get the data sets, where to buy them, where to license them, how to treat them. And we fused those two things together. And I was fortunate enough to work with an incredibly talented team of folks at Google Brain. And we made this happen together. And what we're able to do is solve this 100-year-old problem.
It sounds so simple, but why does this molecule with this shape smell the way that it does? And we validated it in a really stringent way. We basically did a double-blind trial where we predicted the smell of hundreds of thousands of molecules. We picked 400 that were very different looking from anything we'd seen before. We kept our predictions secret. We bought or made the molecules.
So some of these had never been made before. We sent them to our collaborator at Monell. Professor Mainland was running this. And he trained a panel of people. And this is kind of like what we do now, but just initially it was at a smaller scale. Train people to smell something and say, okay, This smells fruity and mineral, and that's it. So I'll give it a three out of five fruit.
I'll give it a one out of five mineral, the rest zeros. And that's called rate all that apply. It's just like, that's how we label data. And then what we did is we said, okay, we have our predictions. People have their double blind ratings. Where do our predictions fit within the people? Because the best is the average of the panel. That's how you get really high quality data for AI.
So, were our predictions worse than the worst person, or were they in the pack somehow? And it turned out that our AI predictions of what these smells were going to be were better than the average panelist.
Meaning, if you were going to add one more person to this panel, you'd actually prefer to ask our software what it smells like that doesn't have access to the physical molecule than to train up another person to physically smell it, which is kind of like passing an odor-turing test.
When that happened, it was very clear that Mother Nature was not going to stand in the way of continuing on this journey of actually digitizing the sense. So if you can solve that one problem, it means you can start to ask, okay, great.
Now, what happens if instead of feeding this AI algorithm a pre-digitized molecule, what if I feed it a reading from a sensor, like the data off of a camera, if we were talking about images? And then what if I then ask it to recreate that smell with the ability to mix together different molecules to create a new scent? If you can actually round-trip
a smell, so take a physical smell, put it in one system, and then round trip through the reader this map that we built, this graph neural network-based map, and then write it back out again, and then compare it, and it actually smells like the thing that you put in. It means that you have actually digitized a human sense.
We hit all of our scientific milestones at Google, and we asked ourselves, what's the right way to scale this idea? And that's where Josh Wolf comes in. So we were thinking internally at Google, maybe this should be a company. And I was working with Krishna Yeshwant at GV, who's a very close friend. We'd worked together for five years.
Showing 961–980 of 1,050 · page 49 of 53
← Previous
Next →