DeepMind podcast mentions

Quote

research lab called Neocognition that has landed forty million dollars to build agents that actually specialize like humans. So we're gonna get into all of that on the show today. Our first story Is that 10X Science? This is a Stanford spinout of a Nobel laureate Carolyn Bertozi's lab. They just closed a $4.8 million seed, which was led by an Initialized capital. What they're doing that's so fascinating to me is that uh basically there's this problem where models like DeepMind's protein predictor, if they're spitting out thousands of drugs. candidates. So there's t there's just thousands and thousands of these drug candidates. And there's a huge bottleneck in pharma where it's not just about, you know, getting all of these candidates, but it's actually triaging them. It's actually figuring out like Which of all the candidates is worth uh pursuing to try to make medicine or therapeutics? And basically the standard triage tool is just mass spectrometry. So it's very slow. It's very hard to interpret. It's, you know, d domain experts only. 10X Science is basically just building a SAS layer on top of that. And they have deterministic chemistry plus AI agents to try and make the analysis traceable and explainable, which basically matters because regulators don't accept a black box answer on what the molecule does, right? So just because an AI models like, oh my gosh, we discovered…

Quote

of intelligent machines. And I've got lots of questions around the content you've put into there and all the work you've put into there. But before we move into that, tell me and my audience a little bit about So I help companies to become AI first organizations. I've been a future boy of some sort for a long time. I've trained as an engineer, but I've always found orientations in my jobs that allowed me to be thinking about and working on the future. So I worked at Intel for a very long time, when they were a force to be reckoned with. I then moved more recently to Google Deepmind. So DeepMind is Google's AI research lab's base in London. I'm based on the West Coast, but I got to go back to London for a bit and recharge my British So that's what I do. I work with boards, management teams, leaders to help them understand AI, where it's going next, and what they need to do to thrive in an AI. You're definitely in the right spot and the right timing for this sort of thing. The AI is definitely a topic of conversation that is on everyone's mind. So Really excited to dive into that. And I'm curious, let's jump into the book. Now again, the title of the book is The AI Ultimatum Preparing for a World of Intelligent Machines. What led you to write this book. Ultimately what were you trying to accomplish here? Growing their business and they're experts in their field, whether it was m…

Quote

AI operating system, and Morgan Stanley just bumped the 2026 hyperscaler forecast to eight hundred and five billion dollars. I want to break down though is Greg Brockman's uh he went on a podcast recently and he had all these wild quotes, uh including talking about AI. He said, quote, oh uh even over Over the course of December, we went from these agentic coding tools, writing twenty percent of your code to writing eighty percent of your code. This is a four X jump in one single month, and he also said that. Opening I is seventy to eighty percent of the way to AGI. He also agre agrees with Sam and Demise over at Google DeepMind and says that we're maybe two breakthroughs away. Now, of course, this wasn't everybody uh that agreed with him over on X Yan Lacun predictably has been dunking on the AGI claim. He's been basically doing this all week. You also have Andre Carpathy who was a lot more I guess you could say like measured, but basically the thing that I've I've heard him talk a lot about is he says that um sure some of these coding is so and that's kinda where you're getting this twenty to eighty percent jump. But he says the difference is and a harder question is what fraction of all of these tokens, all of this code that's being admitted, um Um are you know gonna be merge commits that a senior engineer would actually approve. And I think a a big point of that is a lot of…

Quote

And as a Google, like what what is the or chart layer? Yeah, yeah, yeah, yeah. That's a great question. Yeah. Um so labs's mission is to build kind of new like Innovative products that the rest of Google isn't well positioned for. Yeah. Like uh riser from an Opel M. Exactly, exactly. Yeah. thing about that's really exciting about labs is we work incredibly closely with DeepMind. Yeah. Right. So all the stuff in terms of the, you know, the c we're building the product, but we work so closely for the mall and you know One of the nice things about being at Google is you have this opportunity to really build an end to end air product, right from Like pixels on the page, through the infrastructure, through, you know, the model and the training and all of that. loops. So Labs is here to build new products and we're really like a product org, but a true A AI product org where we work incredibly closely with with you know uh DME but also you know Other parts of Google uh you know as it makes sense. Um Yeah. Just on the history of AI coding. I had heard that actually Google had an internal version of Copilot. Or something like that that was never released. Is that is that true? What can we say? Yeah, so you know I think there are there are you know Google's published papers in this space uh for a while. Um and so yeah we have you know been good a

Quote

MLST is supported by a cyber fund. I'm Ilya. I spend my days staring at mode. Most of the time they don't do What you expect them to do. So we're trying to fix this. In my previous life I was an academic. And uh I have basically been publishing in both security and machine learning. Then I joined DeepMind where I stayed for two years in the best machine learning security team. Now I left I am very unemployed and am trying to build security tooling for the future to make sure that as we get agentic fleets in integrated into more and more use cases we can actually tell what they're doing, we can impose constraints on them and we can have confidence that tomorrow they're not gonna is a little bit crazy. Now we have machine learning models. We actually know what they do. We know how they think, we can check their state. They're kind of like a resettable human, if you will, right? So suddenly what I'm marking in this work is that actually trusted third parties can exist.

Quote

All right everyone, welcome to another episode of the Twin Mall AI Podcast. I am your host, Sam Sherrington, and today I'm joined by Oliver Wong. Oliver as principal scientist at Google DeepMind and tech lead for Gemini. 2.5 flash image, aka nano banana. I'm sure you've all heard of it. Before we get going, be sure to take a moment to hit that subscribe button wherever you're listening. Thanks for having me. I'm looking forward to digging into the conversation and uh talking with you a little bit about nano banana, what you're seeing with the model and the experiences Uh bringing it to our devices to get Get us started though, I'd love to have you share a little bit about your background and how you came to work in the field. So I've been in this area of uh generative models for image I came in to this field through the media entertainment side actually. So I started um after I got my

Quote

It's a jeu qui a 23 ans, hein. Mais pourquoi un vieux jeu spatial pour entraîner des intelligences artificielles financières? Parce que Evenline, c'est pas un jeu de It's an economy vivante, hypercruel, gérée by design imprévisibles. Faillite. DeepMind va entraîner ses agents à opérer dans une économie chaotique avant de les lâcher dans le vrai monde. Alors justement, le vrai monde. Parce que pendant que Ces modèles obtiennent des cartes de crédit d'entreprise et jouent au loup de Wall Street sur Evenline, les humains qui tenaient ces cartes de crédit bas. On leur indique gentiment la porte. L'entreprise allemande DIA linguistique d'IPL vient d'annoncer le licenciement de 25% de ses effectifs. 250 personnes qui perdent leur emploi. Et quand on écoute le PDG, c'est du jargon d'entreprise purée. Complètement. Il parle du besoin d'équipe, je cite, plus petite et plus percutante parce que l'IA va vite. Il veut moins de niveau hiérarchique, une agilité native de l'IA. Alors qu'est-ce que tout cela signifie vraiment pour nous? Parce que là, on détruit des emplois de col blanc de manière tout à fait. systémique. C'est cynique, mais je vais traduire ce jargon financier pour notre audience. Moins de niveau hiérarchique et agilité, ça veut dire transférer le coût des salaires humains vers la

Quote

or over to Anthropic or OpenAI or some other company. Personally, I don't think it's going to have a huge impact on Google's decision, but it might impact who they're able to recruit in the future. But that's kind of a company by company thing. Many different companies are working with the Pentagon and Anthropics seems to be one of the only ones that isn't really at the moment. So I don't know if even that will have that. much impact in the future. The next thing I want to talk about is that the Commerce Department's Center for AI Standards and Innovation, that's the C A I S I. It's basically basically the Trump administration's renamed successor to the AI Safety Institute, they have signed an agreement with Google DeepMind, Microsoft, and XAI to test all of their frontier models for national safety. security risks before public release. Some people are saying this is fantastic for, you know, AI safety and regulation and everything that we see coming from there. Other people are more concerned about it. They don't like the government being able to, I guess, hold back a model if they don't think it's safe. But operationally how this would actually work. is that all of the Frontier Labs, they're gonna give their newest models to the government with no safety guardrails at all. So this is all gonna be done by the C A I S I and they're gonna be able to probe those models. models a…

Quote

AI operating system and Morgan Stanley just bumped the 2026 hyperscaler forecast to eight hundred and five billion dollars. First thing I want to break down though is Greg Brockman's uh he went on a podcast recently and he had all these wild quotes, uh including talking about AI. He said, quote, oh uh even Over the course of December we went from these agentic coding tools writing twenty percent of your code to writing eighty percent of your code. This is a four X jump in one single month. And he also said that OpenAI is seventy to eighty percent of the way to AGI. He also agre agrees with Sam and Demise over at Google DeepMind and says that we're maybe two breakthroughs away. Now, of course, this wasn't everybody uh that agreed with him over on X Yan Lacun predictably has been dunking on the AGI claim. He's been basically doing this all week. You also have Andre Carpathy who was a lot more I guess you could say like measured, but basically the thing that I've I've heard him talk a lot about is he says that um sure some of these coding writing percentages are like huge, right? Going from twenty to eighty percent. But what he said is if you measure tokens emitted versus um what is so and that's kind of where you're getting this twenty to eighty percent jump. But he says the difference is and a harder question is what fraction of all of these tokens, all of this code that's bein…

Quote

and personalization at 85% adoption rates, operational efficiency at 79%, and fraud prevention at 78%. Real-world implementations demonstrate measurable success. Amazon's personalized recommendation engine leverages deep learning and collaborative filtering to boosts sales and customer satisfaction. General Electric uses predictive maintenance algorithms with sensor data, preventing costly equipment failures and reducing operational Downtime. Google DeepMind's load forecasting system for data setters trimmed cooling energy consumption by up to 40%, cutting costs and carbon footprints. print simultaneously. For organizations considering implementation, focus on behavioral data integration, predictive maintenance applications, and personal data engines aligned with core business functions. Start with clearly defined metrics tied to revenue or cost reduction. Then prioritize edge artificial intelligence and federated learning for data privacy protection. Technical requirements increasingly involve cloud-based platforms and pre-built models that reduce deployment time. Thank you for tuning in to Applied AI Daily. Come back next week for more insights on machine learning and business applications. This has been a Quiet Please production. For more, Check out QuietPlease.ai.

Quote

This podcast is supported by Google. Hi folks, Paige Bailey here from the Google DeepMind DevRel team. For our developers out there, we know there's a constant trade-off between model intelligence, speed, and the model, and the model, and the model And cost. Gemini 2.5 Flash aims right at that challenge. It's got the speed you expect. from flash but with upgraded reasoning power. And crucially we've added controls like setting thinking budgets, so you can decide how much reasoning to apply, optimizing for latency and costs. So try out Gemini 2.5 plush at ai studio.google dot com and let us know what you build. Hello and welcome back to the cognitive revolution. enabled both humans and AI agents to transact seamlessly on the internet with cryptocurrency. The protocol takes its name from the HTTP 402 payment.

Quote

Die Dynamik, die du immer wieder siehst, kann Europa AI bauen? Ja. Und ich glaube, da gibt es viele Case-Studies, wo man das irgendwie klar sehen kann und gute Talente hat, gute Unis, gute Research. Und so weiter. Und dann siehst du, glaube ich, nach so einem initialen Abtick so typischerweise irgendwie einen von drei Wegen, der da irgendwie ist: Stagnation, Stagnation, Glaube, das ist bei alle Fife der Fall. I don't know. No hate, so von meiner Position. Ich glaube, der zweite Weg ist halt irgendwie der Wegzug in die USA, wie so, ne, wie so. So und DeepMind und Google das gemacht haben. Der dritte ist halt so dieser frühe Exit, den man jetzt bei PriorLabs sieht. Glaube ich, dass das deutlich größer hätte werden können. Ja, safe. Ist ein riesiger Markt. Tabellen sind riesengroß. Kundenpulver da. Ich bin mir ziemlich sicher, dass die auch eine sehr, sehr große Finanzierungsrunde. Hätten aufnehmen können. Also ich glaube, da war schon noch ein bisschen Luft nach oben. Ja, auf jeden Fall. Ich meine, es ist definitiv, aber wenn immer eine Lanze für Für das Team und die Gründer da auch zu lächeln. Es war auch keine Kritik. Du sagst es eben, Florian, das war Kritik, so meine ich das gar nicht. Es war einfach nur ein Punkt, den ich da mal reinwerfen wollte. Also nicht falsch verstehen. Ja. Nee, also auch. gar nicht kritick, aber ich meine, ich hätte jetzt mal unter uns gesprochen. Ic…