Grace Huckins @gracehuckins.bsky.social · 21/05/2026Oops, this should obviously say 20th century! 000
Grace Huckins @gracehuckins.bsky.social · 20/05/2026That doesn't mean that this project isn't eliminating any jobs. No one had to scout locations, build sets, or source props for this film. 000
Grace Huckins @gracehuckins.bsky.social · 20/05/2026One thing they aren't doing is any AI-generated performances, though that did mean they had to cast tons of people to fill in all those different locations. 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026One advantage of using AI—Liman didn't have to limit the number of locations in the script. The movie could take place in 150 cities and "make Wolf of Wall Street seem sleepy." 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Liman says that with AI tools available at the start of filming, post-production would have taken 40 years. Given advances that have happened since, timeline is more like 6 months. 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026It looks like a black box (well, gray box) theater production. All the set details will be constructed in post, with AI. 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Now a special broadcast of behind-the-scenes footage from the crypto movie, "Bitcoin." We've been instructed not to take pictures. 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Another part of the work that had to go into it—building strong guardrails around the character. This character either is or was based on NFL player DK Metcalf (I have no idea why), so they had to consider reputational issues—e.g. no trash-talking other players 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026The company has made a film where the viewer can interact with an AI character toward the end. Did that take away the job of a screenwriter? In this case, no—apparently getting the AI character to work required a "novel worth of writing" 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Jed Weintrob, co-founder of Liman's company 30 Ninjas, emphasizes the company's "artist-first, human-first approach to AI storytelling"—the classic AI for art disclaimer 210
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Liman is currently making a film about crypto starring Casey Affleck in which sets/backgrounds will be fully AI generated (www.thewrap.com/creative-con...)thewrap.comInside Doug Liman's $70 Million AI-Made Movie Starring Casey Affleck and Gal Gadot | ExclusiveDoug Liman made a $70 million movie using AI about the creator of bitcoin called Killing Satoshi, starring Gal Gadot, Casey Affleck and Pete Davidson. 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Liman: "Art has always reflected the technology of the time—this [AI] is definitely the technology of the time." 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Last thing from me before my final takes go up on @technologyreview.com later this week—a discussion on AI and creativity with director Doug Liman (The Bourne Identity, Edge of Tomorrow, and...Swingers?) 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026A surprising question now about longevity. Demis says "I don't really worry about that question." Goal of Isomorphic etc. is to cure human disease, and that's the key goal before thinking about aging/longevity. 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026A note for Demis from all the Bay Area tech folks in the audience: "Making sure you have the direction right is often more important than just the speed." 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Ok, very interesting thoughts on scientific progress at the end now. A big motivation for him was looking at the stagnation in progress in physics after the mid-19th century. He had a sense we might be reaching the limitations of the human mind. 211
Grace Huckins @gracehuckins.bsky.social · 20/05/2026(Which it's fair to say he is, along with everything else. But that kind of intellectual fanaticism gives even more scientific cred than the Nobel Prize, I'd say.) 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026One thing Demis asks himself—would he still be working on AI if it hadn't had this explosive moment in 2022? Absolutely yes, he says—it's an obsession for him. More than anything else he's said today, that makes him sound like a scientist. 101
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Says continual learning (i.e. AI that keeps learning after deployment) is one thing DeepMind is focusing on now. Definitely something to watch. 121
Grace Huckins @gracehuckins.bsky.social · 20/05/2026What are the big things to look at going forward? Demis says self-improvement. Coding is a small example of that, but what next? He (very correctly) highlights that coding is a very specialized area that's amenable to AI acceleration. Not clear what he thinks the next SI step is. 121
Grace Huckins @gracehuckins.bsky.social · 20/05/2026(Obviously, certain things—e.g. getting good protein structures—go much faster! But that doesn't necessarily speed up molecular biology in general, much less science writ large.) 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Demis says that the pace of science is picking up in all sorts of fields (implicitly due to AI), but there's a lot of research that science has slowed recently. IMO, whether and how AI will change the rate of scientific progress is a big open question. 101
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Talking about the AlphaGo Lee Sedol match, Demis says it feels simultaneously a century ago and just yesterday. Again, very relatable. Reporting on AI has completely destroyed my sense of time. 120
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Demis says he hates seeing himself on video. Relatable. 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026That database is available for free to anyone and has millions of users. Probably the ultimate example of virtuous AI—an interesting thing to highlight. 111
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Now showing a clip from the documentary The Thinking Game where John Jumper raises the idea to Demis that AlphaFold could just predict (and release) the structure of all know proteins. His response: "why don't we just do that?" 101
Grace Huckins @gracehuckins.bsky.social · 20/05/2026And immediately pivots to AI for science—the virtuous side of AI. My subjective take is that he thinks job loss is very possible but doesn't want to potentiate the backlash. 111
Grace Huckins @gracehuckins.bsky.social · 20/05/2026He won't say clearly whether his hunch is net more jobs or fewer jobs, but he "doesn't see why" there wouldn't be more opportunities. "There's going to be changes, but overall, it will be better." 101
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Demis says he doesn't agree with his peers about this. It's super uncertain whether AI creates more jobs, destroys jobs—the future is yet to be written, he says. (So the implication here seems to be that he's blaming rhetoric from Sam/Dario/etc. for the backlash?) 111
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Allen digging in about the AI backlash now. Interested to hear Demis's thoughts on this. 101
Grace Huckins @gracehuckins.bsky.social · 20/05/2026(This is, recall, a session about AI for science!) 110
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Interestingly, Demis says the big leap forward at Googe in the last year has been on the consumer side, i.e. redesigning its products around AI from the ground up. Perhaps—but a pretty boring point from a Nobel Prize-winner! No mention of AI co-scientist, AlphaEvolve, etc? 110
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Demis's test for AGI, which I admittedly kind of like in principle, is whether a system with a knowledge cutoff of 1901 could re-discover the major physics advancements Einstein made in 1905. In reality, this would never work with current AI—not enough training data before 1901. 121
Grace Huckins @gracehuckins.bsky.social · 20/05/2026His passion has always been "increasing our understanding of the world around us." Says we are on the cusp of doing that with AI now. 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026What's making him feel that way? First think he mentions is his experience working with Google's own coding tools. Nice media training there. 101
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Demis: "When we look back, I think we'll say '26, '27 was when it [i.e. AGI, the singularity] was starting." 111
Grace Huckins @gracehuckins.bsky.social · 20/05/2026(Though I would argue we should all be AI for science nerds!) 200
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Definitely something I'll keep following, because I'm a huge AI for science nerd. But for most AI enthusiasts, there probably isn't much to pay attention to there. 100
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Before Demis starts, my main takeaway from the last session is that there are real AI applications of quantum computing, but they derive more from the fact that both are useful scientific tools that can work effectively together than from any intrinsic connection between quantum and AI. 130
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Meanwhile, it is a zoo outside of Demis’s impending AI for science dialogue. GA line is closed, and the poor line managers are very patiently managing some upset attendees. Everyone is desperate to see Demis in person—the simulcast isn’t cutting it. 121
Grace Huckins @gracehuckins.bsky.social · 20/05/2026One random observation—there are a good number of empty seats in the discussions venue for this talk, which wasn't the case for the Sundar event yesterday. Quantum computing doesn't really seem to be on the top of folks's minds these days. 120
Grace Huckins @gracehuckins.bsky.social · 20/05/2026What is motivating Neven now? To prove correct the "basic thesis" that "quantum AI is a more powerful form of AI" 110
Grace Huckins @gracehuckins.bsky.social · 20/05/2026The key question: “How can quantum help AI?” One connection—quantum computers could generate simulated molecular data that could train AI systems like AlphaFold. Not as glamorous as, say, running superfast networks on quantum chips—but much more realistic, and still important. 111
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Finally we’re getting to quantum AI—why that name? One reason: Neven himself comes from an AI background and did his PhD in self-driving cars. Really unusual background for someone who is now so accomplished in quantum. 111
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Quantum echoes can be used to infer details about molecular structure from experimental data, so there are practical scientific applications. 120
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Ok, we're getting a bit into what quantum computing actually does, and this stuff is pretty cool. Last fall, Google demonstrated quantum computers performing an algorithm (much) faster than classical computers—the first time that happened. It's called quantum echoes. 121
Grace Huckins @gracehuckins.bsky.social · 20/05/2026(Still no mention of AI, and session is half over. Makes one wonder why Google calls their quantum division "Google Quantum AI"—maybe they're hoping the AI sparkle will rub off on quantum) 141
Grace Huckins @gracehuckins.bsky.social · 20/05/2026Manyika says there's a list of ~70 algorithms out there where quantum computers are known to have an advantage over classical computers. (I think this is what he's referring to? quantumalgorithmzoo.org) Sound small, but impact depends on how important those algorithms are.quantumalgorithmzoo.orgQuantum Algorithm ZooA comprehensive list of quantum algorithms. 133
