Antscan, AI reconstruction, and why it matters

4:15scanned 2,000 species Sorry, 2,000 specimens in a single week, 800 different ant species. The effort is called antscan. They've got a free public website called antscan.info, and it's kind of like a Google Maps for ant anatomy. The data is also synchronized with genome sequencing projects. >> cool. So, you've got 585 of those scans linking to the genomic data of about 186 different species. I think it's a really cool application of particle physics and like particle accelerators that we don't really think about. It reminds me of a very old story that we did on this podcast about the T-Rex bone. Yeah. Remember that?

4:57>> Yeah. Where it was like they found um blood clotting in a T-Rex bone and showed that, you know, the T-Rex was capable of rapidly healing its bones after some kind of, you know, I think it was a fight with another T-Rex, probably. But it's just it it's a really cool application of particle accelerators that doesn't have anything to do with fundamental physics. It's just leveraging it as a tool to do very high resolution 3D imaging of biological specimens. Which if you asked me before I, you know, we did the original T-Rex story, yeah, which my reaction was like this is unbelievable because we're sort of

5:38repurposing this very expensive Mhm. uh large-scale device Yeah. to become multi-purpose beyond its sort of initial mission or context. >> Yeah, and the scans are like so detailed that they they can be used for further training of machine learning systems to recognize ants in the field. For example, you know, I'm out in the field, I get some specimen, now I have a pre-trained network, a machine learning model that is trained using this really high-fidelity, really high-resolution data, and it can immediately tell me what kind of species it is, or at least what species it is most related to if it's a new species, right? Um it's also something that can be used for like

6:20if Disney or somebody wants to make What is it called? A visual effect. >> Yeah, yeah. Like with like a giant ant. You know, remember that that one movie where like it's like Honey, I Shrunk the Kids? You want to make like a, you know, a giant ant visual effect for that. Now you can be really, really biologically accurate. So if you want to be as precise as Interstellar was in their replication of the physics of a black hole and the amount of science that went into creating so, now particle accelerators can give us animated 3D ants that are giant at very high fidelity. >> Yeah. I kind of is a little bit the the

7:01example you just brought up sounds a little bit like a Pokedex. You can tell if it's a Caterpie Yeah. or a Beedrill or a Weedle Yeah. or any of these, Butterfree. I'm trying to flex my Pokémon knowledge as much as I >> know any of I just know a Pikachu. And um Charizard. >> Charizard, classic. >> Yeah. Classic. Always interesting to see the intersection of different disciplines. >> Yeah. And how again past work can inform new work. And so this was again a methods story. Yeah. Uh fascinating and published in Nature Methods. Uh we have this coming out of the Okinawa Institute of Science and Technology, and the I'll let you you had the good pronunciation on the second institute. >> Oh yeah, that's the Karlsruhe Institute

7:41of >> of technology and Okinawa Institute of Technology of Science and Technology OIST is what we used to call it. I actually did a summer workshop there um for like computational neuroscience. >> That's so sick. It's an amazing like place. That's amazing. >> uh it's on a hill in Okinawa in the middle of the Pacific Ocean. Beautiful, beautiful spot. A lot of bugs, I have to say. >> of bugs. >> of bugs, but you know, it's a tropical island. And you know what does not have