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
not going to do is we're not going to change the hardware because there's about a there's 12 nobel prizes worth of advances inside of these machines they're fantastic what we've done is rip out the brains and we've replaced it with our own brain so a lot of what we've noticed is the hardware actually is already pretty good in the realm of scent and chemistry but the software or the maps that link the different pieces of hardware has been completely missing that's what we built this is kind of like an inner sanctum here um
So this is where we keep every AI-designed molecule that we've made, which is probably a significant fraction of all AI-designed molecules ever. And so this is just one slice of it. So in this room is 10,000, 20,000 molecules that have all been designed by AI. And we have a digital twin of each. So if we need to go back and access it, we know it's fridge one, shelf three, row column two, four.
And the sum total of it kind of smells like a bready radish or something like that. Amazing.
We've brought a lot of people into Osmo that have like truly world-class noses. And so I can say definitively that like I'm not world-class. So this is kind of the Rolls-Royce machine. It does the same thing as the other one, except there's two more things that are interesting. One is you don't have to inject a liquid into this.
You can put like anything into these vials and it will suck the smell from out of what you put in the vials. It will analyze it.
Directly. So what it does is it pumps air into these vials with a needle syringe. So it'll get dropped in here. A needle will be pushed into it. So basically we'll suck the air and we'll concentrate it onto, you know, like Kodak film absorbs light. We have film that absorbs scent.
And they basically concentrate the smell on that thin piece of film, and then you move that needle and you inject it into the spectrometer, and it uses a flash of heat to remove all those molecules. It kind of develops the film. And then the normal machine runs, we analyze the data with AI, and then we can pull back out what the scent actually was.
So this means that we can analyze flowers and vegetables and people and fruits. And so what we did, the first scent that we fully teleported digitally was a fresh summer plum. So it was like kind of the purple plum, you know, like the really good ones have like a snap when you bite into it. It was like one of those fresh ones.
So we sliced it, we put it into one of these vials, we analyzed the smell, and then we actually reprinted the smell on the other side of the lab, which I'll show you. The other thing you can do with this machine, which is really cool, is you can pause the smell at any point in time and you can just sniff molecule by molecule.
So a scent will be like 30 molecules, 100 molecules all blended together, different types. You can smell them one by one by putting your nose on here. It's kind of like a debugger for software. Wow. So this is called a GCO or gas chromatograph olfactometer.
But when we really want to understand the smell and kind of like build our intuition when we're building new protocols, we'll actually sit here and sniff stuff that comes off the machine.
Exactly. And if you can read and write, then you can create this virtuous cycle where you're creating data at every run of the loop. And so if you actually can create new smells and then you can turn those smells into data readings of some kind, you're training AI.
And then if you can tilt that process so the next smells that you create the next day teach the system even more, that's when you're doing what's called active learning. And that's how you get AI systems to get smart really fast. And that's what we do.
There's more nuance to how we do that to create data that can actually be fed into a machine learning system. But that's effectively it, which is like, do these things match? And there's a few tricks that you use to help de-bias people and get reliable data. But like, you're the arbiter, right?
If it's like a smell you're familiar with and I'm trying to recreate a memory that you have, like we either did it or we didn't.
Aromarama.
They only had certain smells that were like whole scenes. And so they didn't have primary odors. They didn't have the ability to create any smell. And so here on this robot, you're obviously not putting this behind a couch cushion yet, but we're going to make this smaller. But the idea here is you need to have all the ingredients together that can be mixed on the fly to create any experience.
not just like eight pre-programmed experiences that's like a slideshow right we want an actual display that can show anything what is the most surprising thing about primary smells we're kind of at the scientific frontier and so like everything that we discover every week every month like pushes back what's known about smell and how to construct it um i think
One thing that I've found in doing science and machine learning and combining these things is problems that people sometimes think are totally intractable, once you just get started, you're like, oh, we're actually making progress.
And so the idea of creating sense with AI and creating those sense in partnership with people and fusing human and machine to work in this very emotional world of sense. People like, you just don't think of it. It's like, oh, that's crazy.
Showing 901–920 of 1,050 · page 46 of 53
← Previous
Next →