Demis HassabisEvery podcast appearance, updated as new ones drop

When Demis Hassabis pitched DeepMind to a few venture capitalists back in 2010, the business plan was almost comically audacious. “Step one: Solve intelligence. Step two: Use it to solve everything else,” he recalls in a conversation at Stanford Graduate School of Business with Stanford University President Jonathan Levin. “And people were quite confused. But we really meant it.” Sixteen years later, the “broad arcs” of that plan have gone “unbelievably well,” says Hassabis, a chess prodigy turned video game developer turned neuroscientist turned Nobel Prize-winning AI pioneer. Today he’s on a mission to create “the ultimate tool for science,” building on his decision to give away AlphaFold, the groundbreaking AI system that predicts the structures of proteins. The future, Hassabis says, is just around the corner: “Ten years from now, I think we’ll realize that we were standing in the foothills of the singularity now.” AI@GSB, the Dean's Applied AI initiative at the Stanford Graduate School of Business (GSB), and Stanford Medical School hosted a conversation with Demis Hassabis, Co-founder and CEO of Google DeepMind, on the frontier of artificial intelligence and what it means for how we live, work, and flourish. See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info .

Demis Hassabis says we’re in the ‘foothills of the singularity’ In this exclusive conversation, Rowan Cheung (@ rowancheung ) sat down with DeepMind CEO Demis Hassabis (@ demishassabis ) to break down: Why AGI is coming in 2030, plus or minus a year How AI will compress drug discovery from 10 years to weeks Why glasses are the killer app form factor AI was waiting for What Demis will work on after AGI (and what's left for human meaning) __ Join our daily AI newsletter: https://www.therundown.ai/subscribe Learn AI hands-on with our AI University: https://rundown.ai/ai-university/

Demis Hassabis is an artificial intelligence researcher, scientist, and entrepreneur. In 2010, he co-founded DeepMind, an AI research lab which is now part of Google. In 2024, Hassabis won a Nobel Prize for using AI to predict the 3D structure of proteins, critical for disease understanding and drug discovery. He was also awarded a knighthood that year by King Charles III. On April 20, 2026, Sir Demis Hassabis came to the Sydney Goldstein Theater in San Francisco to talk with author Sebastian Mallaby, who recently published a book about Hassabis’s work, The Infinity Machine: Demis Hassabis, DeepMind, and the Quest for Superintelligence . The two were interviewed on stage by journalist Emily Chang.

Demis Hassabis：AGI 還缺什麼，Agent 到底行不行，下一個科學突破長什麼樣 Demis Hassabis 是 Google DeepMind 的 CEO，也是 Isomorphic Labs 的 CEO。他在棋手神童和遊戲開發者的身份之外，拿了認知神經科學的博士學位，研究海馬體和記憶的工作方式。2024 年，他因為 AlphaFold 的工作獲得諾貝爾化學獎。 這次他做客 Y Combinator 的 How to Build the Future 直播，和 YC CEO Garry Tan 聊了四十分鐘。幾個核心話題：當前 AI 範式距離 AGI 還差什麼、Agent 的真實水平、AI 在科學領域的突破模式，以及給深科技創業者的建議。 原始影片：https://www.youtube.com/watch?v=JNyuX1zoOgU 原始標題：Demis Hassabis: Agents, AGI & The Next Big Scientific Breakthrough 要點速覽 Hassabis 認為當前範式（預訓練 + RLHF + 思維鏈）會是 AGI 架構的一部分，但有 50% 的機率還需要一兩個尚未發現的關鍵突破，持續學習、長程推理和記憶是三個未解問題。 百萬 token 上下文視窗聽起來很大，但處理即時影片時只夠錄 20 分鐘，當前把所有東西塞進上下文視窗的做法是「用膠帶糊住的臨時方案」。 AlphaGo 和 AlphaZero 時代的技術（蒙地卡羅樹搜尋等）正在被重新引入當代基礎模型，Hassabis 認為未來幾年的進步將大量來自這些舊想法的規模化應用。 他用下棋來測試 Gemini 的推理能力，發現模型會識別出一步是錯棋，找不到更好選擇後又回去走那步錯棋，這種「缺乏自省」是當前推理系統的核心缺陷。 創造力的真正測試是能否從一段高層描述中發明圍棋這個遊戲本身，AlphaGo 下出 Move 37 級別的創造力還遠遠不夠。 完整虛擬細胞大約還需要 10 年，關鍵瓶頸是無法在不殺死細胞的情況下對活細胞進行奈米級解析度成像。 他給創業者的建議：如果你的 AGI 時間線是 2030 年，深科技創業通常需要 10 年，那 AGI 會在你旅程的中途出現，你的商業計畫必須把這個因素算進去。 【1】AGI 還缺一兩塊拼圖，機率 50/50 Garry Tan 開場問：當前的 AI 範式，大規模預訓練、RLHF、思維鏈，這些東西裡已經包含了多少 AGI 的最終架構？還有什麼根本性的缺失？ Hassabis 的回答比較謹慎。他說當前這些組件「幾乎可以確定」會是 AGI 最終架構的一部分，走到今天這一步已經證明了太多東西，不可能突然發現這是一條死路。但在已有的東西之上，可能還需要一兩個大想法。 他列出了三個未解問題：持續學習（continual learning，即模型在部署後持續從新經驗中學習的能力）、長程推理，以及記憶的某些方面。這些問題也許能靠現有技術的漸進式創新解決，也許需要全新的方法。 他給出了一個有意思的機率判斷：50/50。一半機率是現有技術足夠，另一半機率是還缺一兩個關鍵突破。Google DeepMind 兩邊都在押注。 【2】記憶：百萬 token 上下文其實不夠用 話題自然轉到了記憶和上下文視窗。Garry Tan 提到現在的系統每次處理都是無狀態的，持續學習缺失的情況下，大家都在用「夢境循環」（定期批次更新）這類臨時方案。 Hassabis 對這個話題有獨特的發言權。他的博士研究就是海馬體如何將新知識優雅地整合進已有的知識庫。大腦在睡眠（特別是 REM 快速動眼期）中回放重要的經歷片段來鞏固學習，DeepMind 最早的 Atari 遊戲 AI 程式 DQN 就借鑒了這個機制，用「經驗回放」（experience replay）反覆重放成功的遊戲軌跡來加速學習。 我們現在的做法有點像用膠帶糊住，就是把所有東西都塞進上下文視窗。 （“We're kind of using duct tape right now—shove it all in the context window.”） 他接著解釋為什麼這個方案不夠好。百萬 token 上下文視窗聽起來很大，人類的工作記憶平均只有 7 個數字左右，而 AI 有百萬甚至千萬級別的上下文。但問題是，我們把所有東西都扔進去了，不管重要不重要、對不對。更關鍵的是，如果你要處理即時影片流，天真地錄入所有 token 的話，百萬 token 其實只夠 20 分鐘。如果你想讓系統理解你一兩個月的生活，遠遠不夠。 即使儲存空間無限，找到當下決策真正需要的那條資訊，這個搜尋成本也是不可忽視的。Hassabis 認為記憶領域還有很大的創新空間。 【3】AlphaGo 的技術遺產正在復活 Garry Tan 追問 DeepMind 在強化學習方面的歷史積累，AlphaGo、AlphaZero、MuZero 這些系統背後的哲學在今天建構 Gemini 時發揮了多大作用。 Hassabis 說強化學習的重要性「在起伏中輪迴」。DeepMind 從創立第一天起就在做 Agent，Atari 遊戲 AI 和 AlphaGo 說到底都是 Agent 系統，能自主設定目標、做決策、制定計畫。當時選擇遊戲領域是為了讓問題可控，然後逐步挑戰更複雜的遊戲，比如 AlphaGo 之後又做了星海爭霸（AlphaStar）。 過去幾年的核心問題是：能否把這些模型從遊戲推廣到語言和世界模型？而今天所有前沿模型的思維模式和思維鏈推理，其實都可以追溯到 AlphaGo 時代開拓的路徑。 他透露了一個值得關注的資訊：Google DeepMind 正在重新審視當年的一些舊想法，包括蒙地卡羅樹搜尋（Monte Carlo tree search）等方法，在當今基礎模型的規模上重新應用。他認為未來幾年 AI 的很多進步將來自於 AlphaGo 和 AlphaZero 時代的想法與現代基礎模型的結合。 【4】小模型在快速變聰明 Garry Tan 觀察到蒸餾技術讓小模型越來越接近前沿模型的能力，Flash 模型大約能達到前沿模型 95% 的水平，成本只有十分之一。他問蒸餾有沒有極限。 Hassabis 說這是 Google DeepMind 的核心優勢之一。他們當然要建最大的模型來推動能力邊界，但快速把這些能力壓縮到更小模型中是他們的強項。Google 有十幾個十億使用者級的產品，搜尋的 AI 概覽和 AI 模式、Gemini 應用、YouTube、Maps，每一個都需要 AI 服務。幾十億使用者需要極快、極高效、低延遲的服務，這種商業壓力反過來成了技術進步的引擎。 關於蒸餾的理論極限，他說目前沒有看到任何資訊密度的硬性天花板。他們的工作假設是：前沿模型發布半年到一年後，同等能力就會出現在邊緣級小模型上。 他還提到了一個架構設想：未來可能是高效的本地模型處理日常任務（比如音訊和影片流），只在特定情況下才呼叫雲端的前沿模型。這種「本地 + 雲端」的分層架構對隱私和安全特別有意義，尤其是考慮到家用機器人等場景。 【5】Gemini 下棋暴露的推理缺陷 Garry Tan 接著問推理能力：模型能做出很厲害的思維鏈推理，但在聰明大學生不會犯的錯誤上翻車。 Hassabis 認為當前的思維範式還很粗糙，有很大的創新空間。比如可以監控思維鏈的進展、在推理過程中途介入糾正。他經常覺得這些系統在「過度思考」，陷入某種循環。 他舉了一個具體的例子。他有時會用 Gemini 下棋，所有前沿基礎模型在遊戲上都表現很差，但這恰好提供了一個有趣的觀察視窗。因為棋局的規則是確定的，他能很快判斷模型的思維鏈是否在走彎路。 他觀察到的現象是：模型考慮某一步，意識到這步是臭棋，但找不到更好的，於是繞了一圈又回到那步棋，然後走了出去。 在一個真正精確的推理系統裡，你不應該看到這種情況。 （“You just shouldn't be seeing that happening in a very precise reasoning system.”） 這就是他所說的「鋸齒狀智慧」（jagged intelligence）：一方面能解國際數學奧林匹亞（IMO）金牌級別的問題，另一方面換個提問方式就會犯基本的算術錯誤。在他看來，這種不一致說明系統缺少某種對自身思維過程的「自省」能力。但他也補充說，修復這種缺陷可能只需要一兩個關鍵調整。 【6】Agent：實驗階段，投入產出比還沒對上 Garry Tan 問 Agent 是炒作還是剛剛開始。Hassabis 的回答是：剛剛開始，但還在實驗階段。 他的論點是：要達到 AGI，你必須有一個能主動解決問題的系統，Agent 就是通向 AGI 的路徑。但目前，Agent 在「完整任務」上還不夠好，主要是因為它們不能在具體使用環境中持續學習和適應。缺乏持續學習是 Agent 無法做到「交付後不管」（fire and forget）的根本原因。 他還提到了 a 個耐人尋味的觀察： 我看到很多人啟動幾十個 Agent 跑 40 個小時，但我不確定產出能匹配這種級別的投入。 （“I see a lot of…

📝 本期播客简介 本期我们克隆的是 Y Combinator 官方播客的一期深度对谈。 Demis Hassabis: Agents, AGI & The Next Big Scientific Breakthrough 主持人 Gary 是 YC 的 CEO，嘉宾 Demis Hassabis 是 DeepMind 的联合创始人兼 CEO，他因破解生物学上长达五十年的蛋白质结构预测难题，在去年获得了诺贝尔化学奖。 在这期节目里，你将听到 Demis Hassabis 从国际象棋神童、游戏设计师到诺奖得主的传奇经历，以及他对通用人工智能的终极思考。对话深入探讨了当前 AI 系统缺失的关键组件——持续学习、长期推理和记忆；他独家披露了 DeepMind 如何将其在 AlphaGo 上验证过的强化学习和搜索哲学，融入当今最先进的 Gemini 大模型。此外，他还分享了小模型的“蒸馏”艺术、智能体的真实进展，以及 AI 将在未来五年如何彻底变革材料科学、药物发现等基础科学领域。对于每一位正在科技前沿探索的创始人，Demis 给出了一条至关重要的建议：在 AGI 可能于途中降临的时代，你该如何预判技术走向，构建真正具有防御性的深度科技公司。 👨‍⚕️ 本期嘉宾 Demis Hassabis，Google DeepMind 联合创始人兼 CEO，2024年诺贝尔化学奖得主。他从小是国际象棋神童，17岁便设计了畅销游戏《主题公园》，后来获得认知神经科学博士学位，并于2010年创立 DeepMind，致力于“解决智能问题”。他领导的团队开发了击败世界围棋冠军的 AlphaGo 和破解蛋白质结构预测难题的 AlphaFold，后者已被全球超过三百万研究人员使用，被誉为 AI 加速科学发现的里程碑。目前，他正带领团队打造 Gemini 模型，并继续朝着通用人工智能的宏大目标前进。 ⏱️ 时间戳 开场与嘉宾传奇 00:00 开场：Y Combinator播客简介与Demis Hassabis的非凡成就 03:30 Demis的职业生涯回顾：从棋坛神童到认知神经科学博士，再到DeepMind创立 05:00 AlphaGo与AlphaFold：两个改变世界对AI认知的里程碑 06:30 诺奖背后：免费开放AlphaFold，赋能全球每一位科学家 AGI架构的未来拼图 07:15 当前范式的局限：大模型还缺什么？持续学习、长期推理与记忆 09:45 “梦境循环”与海马体：神经科学启发下的经验重放技术 12:30 上下文窗口是终极方案吗？工作记忆的蛮力模拟与信息检索成本 15:00 生物大脑不是机器：完美记忆的承诺与逻辑成本困境 从AlphaGo到Gemini：强化学习的回归 17:45 智能体的原始基因：Atari游戏与AlphaGo如何定义自主系统 19:45 “想太多”的模型：在思维链中如何避免循环错误 22:00 强化学习被低估了吗？将游戏策略泛化到世界模型 24:00 AlphaZero的旧思想与当今基础模型的新结合 超高效的小模型：蒸馏的艺术 26:30 从庞大前沿模型到轻量级Flash：蒸馏技术的极限在哪？ 28:30 为何必须极致高效：服务数十亿用户的谷歌生态与低延迟刚需 30:30 速度优于绝对能力：迭代效率如何弥补5%的能力差距 32:00 隐私与安全：设备端小模型的战略意义 智能体时代的黎明 35:00 智能体真实能力：到底是炒作还是真正的起步？ 37:00 人机协作：为什么还没出现“AI造出的爆款游戏”？ 38:45 失踪的创造火花：一个能发明“围棋”的系统何时到来？ 40:15 工具的灵魂：人类品味与创造力的不可或缺 多模态、开源与设备端模型 43:30 从Gemini到Gemma：开放科学基因与开源模型战略 46:00 为何开放边缘模型？部署风险与安卓、机器人的全球布局 48:00 多模态先见：如何让AI理解物理世界并遥遥领先 AI与基础科学的下一个突破口 51:00 AlphaFold的范式：组合搜索空间、清晰目标函数与合成数据 53:30 迈向虚拟细胞：我们需要什么样的活细胞成像技术？ 56:00 未来五年最具变革潜力的科学领域：材料、气候与数学 58:00 根节点问题：如何用AI解锁全新科学发现的分支 给深科技创始人的忠告 01:00:30 预测AI走向与跨学科结合：如何构建不被基础模型吞没的护城河 01:03:00 拥抱深度科技：真正有价值的事从不简单，相信你的另类视角 01:06:30 为自己热爱的事业而战：即使技术未成，你也会找到继续的路 01:08:30 终极建议：在AGI终点途中启动你的深科技征程 终极挑战：科学推理与AI的创造力 01:10:30 系统能否自己提出“黎曼假设”？超越模式匹配的科学推理 01:13:00 “爱因斯坦测试”：训练截止1901年的模型，它会发现狭义相对论吗？ 01:16:00 通用工具与专用系统的未来：为何AGI不应是一个巨无霸大脑 🌟 精彩内容 💡 AGI的最终架构：还缺哪几块拼图？ Demis 明确指出，尽管当前的大规模预训练、RLHF和思维链范式已非常强大，但要实现完全的通用智能，我们还必须攻克持续学习、长期推理和更高效稳定的记忆机制。“我觉得现有的组件会是AGI最终架构的一部分，但可能还需要一两个重大的想法去突破。” 🧠 从神经科学借来的AI灵感 Demis 结合其认知神经科学的博士背景，解释了DeepMind早期突破的核心概念——“经验重放”如何源于对大脑海马体在睡眠中巩固记忆的研究。这一在2013年被用于Atari游戏AI DQN的 “远古时期”突破，至今仍对克服模型的“无状态”难题具有深刻启发。 🚀 小模型的大智慧与蒸馏的极限 当被问及小模型的聪明程度是否有极限时，Demis 乐观地表示目前远未触及信息密度的天花板。“我们的一款前沿模型发布半年到一年后，你就能在那种非常小、几乎能跑在设备端的模型里看到同样的能力。” 这不仅关乎成本，更关乎速度与隐私，他认为设备端高效模型加云端强大模型协调将是理想的终局。 ♟️ 当AI“想太多”：从国际象棋的错误说开去 Demis 分享了与Gemini下棋的有趣观察：模型有时会意识到某步是臭棋，但因找不到更好的选择而依然走出那一步。“在一个精确的推理系统里，你根本不应该看到这种事……我总觉得它对自己的思考过程缺少一种内省。” 这反映了当前系统推理

Demis Hassabis has had one of the most extraordinary careers in tech. He started as a chess prodigy and video game designer at 17 before getting a PhD in neuroscience and going on to found DeepMind. His lab cracked Go, solved protein structure prediction with AlphaFold, and then gave it away free to every scientist on earth. That work won him the 2024 Nobel Prize in Chemistry. Today he leads Google DeepMind, pushing toward the same goal he set as a teenager: AGI. On this special live episode of How to Build the Future, he sat down with YC's Garry Tan to talk about what still needs to happen to get us to AGI, his advice for founders on how to stay ahead of the curve and what the next big scientific breakthroughs might be. Chapters:00:00 — Intro00:46 — Demis Hassabis: From Chess Prodigy to DeepMind01:48 — What’s Missing Before We Get To AGI?03:36 — Why Memory Is Still Unsolved06:14 — How AlphaGo Shaped Gemini08:06 — Why Smaller Models Are Getting So Powerful10:46 — The 1000x Engineer12:40 — Continual Learning and the Future of Agents13:32 — Why AI Still Fails at Basic Reasoning15:33 — Are Agents Overhyped or Just Getting Started?18:31 — Can AI Become Truly Creative?20:26 — Open Models, Gemma, and Local AI22:26 — Why Gemini Was Built Multimodal24:08 — What Happens When Inference Gets Cheap?25:24 — From AlphaFold to the Virtual Cells28:24 — AI as the Ultimate Tool for Science30:43 — Advice for Founders33:30 — The AlphaFold Breakthrough Pattern35:20 — Can AI Make Real Scientific Discoveries?37:59 — What to Build Before AGI ArrivesApply to Y Combinator: https://www.ycombinator.com/applyWork at a startup: https://www.ycombinator.com/jobs

Demis Hassabis – CEO and co-founder of Google DeepMind – is one of the world’s most visionary technologists. A child chess prodigy from North London, Hassabis was awarded the 2024 Nobel Prize in Chemistry for using artificial intelligence to predict the complex structures of nearly all known proteins. His company DeepMind, now owned by Google, is at the forefront of the pursuit to build artificial general intelligence, and considered Google’s engine room of AI innovation. Sebastian Mallaby – former FT contributing editor, two-time Pulitzer Prize finalist and author of numerous books – has, for the past 3 years, explored the moral questions at the heart of AI and AGI, through the story of Demis Hassabis. With extensive access to DeepMind and its key players, Mallaby has conducted hundreds of hours of interviews with Hassabis and his inner circle as well as detractors and rivals at other companies. No other journalist has had such a closeup view of the opportunities, hype and threats AI could pose for us all. In April 2026 Hassabis and Mallaby came together for an intimate exploration of The Infinity Machine, Mallaby’s definitive account of Hassabis’ life and career. They discussed how he came to lead the world’s most ambitious AI lab, what the pursuit of AGI might cost as well as what it might unlock, and what the story of Hassabis and DeepMind can tell us about humanity’s innate drive to develop new technologies. --- If you'd like to become a Member and get access to all our full ad free conversations, plus all of our Members-only content, just visit intelligencesquared.com/membership to find out more. For £4.99 per month you'll also receive: - Full-length and ad-free Intelligence Squared episodes, wherever you get your podcasts - Bonus Intelligence Squared podcasts, curated feeds and members exclusive series - 15% discount on livestreams and in-person tickets for all Intelligence Squared events ... Or Subscribe on Apple for £4.99: - Full-length and ad-free Intelligence Squared podcasts - Bonus Intelligence Squared podcasts, curated feeds and members exclusive series … Already a subscriber? Thank you for supporting our mission to foster honest debate and compelling conversations! Visit intelligencesquared.com to explore all your benefits including ad-free podcasts, exclusive bonus content and early access. … Subscribe to our newsletter here to hear about our latest events, discounts and much more. Learn more about your ad choices. Visit podcastchoices.com/adchoices

Demis Hassabis – CEO and co-founder of Google DeepMind – is one of the world’s most visionary technologists. A child chess prodigy from North London, Hassabis was awarded the 2024 Nobel Prize in Chemistry for using artificial intelligence to predict the complex structures of nearly all known proteins. His company DeepMind, now owned by Google, is at the forefront of the pursuit to build artificial general intelligence, and considered Google’s engine room of AI innovation. Sebastian Mallaby – former FT contributing editor, two-time Pulitzer Prize finalist and author of numerous books – has, for the past 3 years, explored the moral questions at the heart of AI and AGI, through the story of Demis Hassabis. With extensive access to DeepMind and its key players, Mallaby has conducted hundreds of hours of interviews with Hassabis and his inner circle as well as detractors and rivals at other companies. No other journalist has had such a closeup view of the opportunities, hype and threats AI could pose for us all. In April 2026 Hassabis and Mallaby came together for an intimate exploration of The Infinity Machine, Mallaby’s definitive account of Hassabis’ life and career. They discussed how he came to lead the world’s most ambitious AI lab, what the pursuit of AGI might cost as well as what it might unlock, and what the story of Hassabis and DeepMind can tell us about humanity’s innate drive to develop new technologies. Learn more about your ad choices. Visit podcastchoices.com/adchoices

Google DeepMind CEO、诺贝尔奖得主 Demis Hassabis 的最新访谈：The Hardest Problem AI Ever Solved, with Google DeepMind CEO（视频链接：https://www.youtube.com/watch?v=C0gErQtnNFE&list=PLF-HhhjMki5mV1OrDe5YkVkS8UIi4lY7m） 在这场别开生面的访谈中,主持人 Cleo Abram 与 Demis Hassabis 用叠叠乐积木作为道具,展开了一次关于 AI 未来的深度对话。每一块积木代表一个改变世界的 AI 项目——从获得诺贝尔奖的 AlphaFold 到击败围棋世界冠军的 AlphaGo,从消费级产品 Gemini 到前沿科学项目 Alpha Genome。 Demis 分享了他如何用 AI 破解"蛋白质折叠问题"这一生... 去小宇宙查看完整单集简介 在小宇宙查看该单集文稿

Demis Hassabis is the Co-Founder & CEO of Google DeepMind - working on AGI, responsible for AI breakthroughs such as AlphaGo, the first program to beat the world champion at the game of Go; and AlphaFold, which cracked the 50-year grand challenge of protein structure prediction and was recognised with the 2024 Nobel Prize in Chemistry. Demis is revolutionising drug discovery at Isomorphic Labs. Ultimately, trying to understand the fundamental nature of reality. AGENDA: 00:04:00 — What Actually Counts as AGI; and Where Are We Today? 00:05:00 — What Are the Biggest Bottlenecks Holding AI Back Today? 00:06:00 — Have We Hit the Limits of Scaling Laws? 00:07:00 — Where Is AI Ahead of Expectations; and What's Still Missing? 00:07:30 — Why Can't AI Systems Learn Continuously Like Humans? 00:08:30 — How Did DeepMind Go from Behind to Leading the Pack? 00:11:00 — Are We Heading Toward Model Commoditization; or Winner-Takes-All? 00:12:00 — What Does the Future of Open Source Really Look Like? 00:13:00 — What Does a Post LLM World Look Like? 00:14:45 — Can AI Really Fix Drug Discovery—and Cut the 10-Year Timeline? 00:17:00 — What Does "Good" AI Regulation Actually Look Like? 00:18:00 — Who Should Be the Ultimate Arbiter of Truth in an AI World? 00:19:30 — If Demis Had One Shot to Fix AI Safety, What Would He Do? 00:21:00 — Is This Time Different for Jobs; or Will History Repeat Itself? 00:22:00 — Is AGI Bigger Than the Industrial Revolution; and Faster? 00:23:00 — Are We Underestimating AI Despite All the Hype? 00:23:30 — Does AI Lead to Massive Inequality; or Universal Prosperity? 00:24:30 — How Do We Solve the Energy Crisis Created by AI? 00:26:00 — Why Stay in the UK Instead of Moving to Silicon Valley? 00:28:00 — Will Europe Ever Build a Trillion-Dollar Tech Giant? 00:29:30 — Meeting Elon Musk for the First Time? 00:31:00 — What Big Questions About AI Is No One Talking About? 00:31:30 — What Does Demis Want His Legacy to Be?

AI is front and center in Davos this year, as world leaders and tech executives debate how quickly the technology is reshaping the economy and workforce. Demis Hassabis, co-founder and CEO of Google DeepMind, sits down with CNBC’s Andrew Ross Sorkin at the World Economic Forum. The two discuss Gemini’s position in the AI race, the evolution of artificial general intelligence (AGI), and what it all means for jobs. In this episode: Demis Hassabis, @demihassabis Andrew Ross Sorkin, @andrewrsorkin Cameron Costa, @CameronCostaNY Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

In this episode of ET@Davos, ET’s Sruthijith KK speaks to Demis Hassabis, CEO of Google DeepMind and Nobel Laureate 2024, on the future of AI. The chess prodigy-turned scientist-turned-AI pioneer explains how DeepMind balances frontier research with a billion-user scale. Hassabis says Google’s Apple partnership followed direct model comparisons where Gemini prevailed; China is now only months behind the West but lacks frontier breakthroughs; and AGI could arrive within a decade, triggering “post-scarcity” abundance. He defends AI’s energy demands, citing AI-designed fusion and grid optimisation. From Transformers to AlphaFold, Hassabis argues Google pioneered modern AI but moved too slowly. His bottom line: within 5–10 years, machines will be doing original science. The stakes couldn’t be higher. You can follow Sruthijith K.K. on his social media: X and Linkedin Check out other interesting episodes like: When Grinch Almost Stole Gig Workers' Christmas , How Will a Volatile ₹ Impact You in 2026? , How Quick Commerce is Triggering a Health Crisis for Gen Z , India’s Labour Law Reboot , Viral to Valuation: Building Women’s Cricket as a Brand and much more . Catch the latest episode of ‘The Morning Brief’ on The Economic Times Online , Spotify , Apple Podcasts , JioSaavn , Amazon Music and Youtube . See omnystudio.com/listener for privacy information.

📝 本期播客简介 本期我们克隆了：达沃斯论坛现场对话 《Demis Hassabis on an AI Shift Bigger Than Industrial Age》 站在 AI 浪潮之巅的 Demis Hassabis 怎么看当前的竞争？在这场深度对话中，这位谷歌 DeepMind 的掌舵人、诺贝尔奖得主，首次详尽披露了谷歌在 Gemini 研发背后的紧迫感。他不仅回应了关于“红色警报”的传闻，还给出了他对于通用人工智能（AGI）降临的最新时间表：2030年。 Demis 认为，我们正在经历一场广度和深度都将是工业革命 100 倍的技术变革。他分享了 AI 在物理世界（机器人）的突破节点、对中国 AI 竞争力的冷静观察，以及在“后稀缺”时代，当 AI 解决掉能源和材料问题后，人类该如何寻找生存的意义。这不仅是一场关于技术的硬核对谈，更是一位顶级思想家对人类文明走向的深刻预判。 👨‍⚕️ 本期嘉宾 Demis Hassabis，谷歌 DeepMind 首席执行官，DeepMind 联合创始人。他是神经科学家、人工智能研究者、国际象棋大师，并因在蛋白质结构预测方面的贡献荣获诺贝尔化学奖。他被誉为“AI 界的爱因斯坦”，致力于通过“解决智能”来“解决一切问题”。 ⏱️ 时间戳 00:00 开场 & 播客简介 谷歌的“红色警报”与回归 02:05 找回状态：Gemini 系列与谷歌的创业公司冲劲 02:57 核心优势：从 Transformer 到 TPU 的全栈能力 04:36 极限工作流：每周100小时，凌晨一点的思考时刻 AI 的物理版图与国际竞争 05:31 物理智能的“AlphaFold 时刻”：未来18-24个月的突破 06:41 机器人的挑战：为什么人类的手极难被超越 07:14 冷静看中国：DeepSeek 证明了追赶速度，但原创性仍待观察 AGI 的时间表与技术路径 08:27 2030 预判：AGI 必须具备人类所有的认知能力 09:38 “锯齿状智能”：为什么 AI 还没能完全取代白领工作 14:39 路径之争：Transformer 是死胡同吗？ 15:11 缺失的拼图：世界模型、推理能力与持续学习 安全、监管与协作 12:39 理想主义：建立 AI 领域的“国际欧洲核子研究中心（CERN）” 13:47 巨头间的默契：谷歌与 Anthropic、OpenAI 合作的可能性 19:13 信任基石：为什么 Google 的“科学公司”基因至关重要 后稀缺时代的终极思考 10:24 丰裕世界：核聚变、新材料与“后稀缺”时代 18:04 科学工具的终极版：AI 独立发现能拿诺贝尔奖吗？ 21:49 终极谜团：利用 AI 探索物理极限、费米悖论与意识本质 23:23 给下一代的建议：在剧变时代，唯一重要的技能是“学会如何学习” 🌟 精彩内容 💡 100 倍于工业革命的变革 Demis 强调，AI 带来的变革速度是工业革命的 10 倍，规模也是 10 倍，综合影响是 100 倍。他认为人类必须利用这种“超能力”去解决能源（如核聚变）和材料科学的根本问题，从而进入一个极度丰裕的社会。 🤖 机器人的“十八个月”窗口期 虽然大语言模型已经很成熟，但 Demis 认为物理世界的突破还需要 18 到 24 个月。他特别提到了与 Boston Dynamics 的合作，并感叹人类双手的精妙结构是目前 AI 和硬件最难攻克的堡垒。 🧠 AGI 的“2030 门槛” Demis 坚持 2030 年实现 AGI 的预测，但他对 AGI 的定义非常严苛。他认为目前的 AI 存在“锯齿状智能”，在某些领域极强但在常识和稳定性上极弱。要达到 AGI，还需要在世界模型、逻辑推理和长期规划上实现 1 到 5 个关键技术突破。 🔬 AI 是科学研究的“终极显微镜” 作为诺奖得主，Demis 坚信 AI 的最高使命是加速科学发现。他将 AI 比作“科学工具的终极版本”，就像更高级的望远镜或显微镜。在可预见的未来，科学发现仍将是顶尖科学家与 AI 协作的成果，人类负责提出假设，AI 负责穷尽探索。 🎨 寻找“后工作时代”的意义 如果未来大家都不需要为了生存而工作，人类该怎么办？Demis 坦言他更担心意义感缺失而非经济问题。他建议大家现在就开始培养“学会如何学习”的能力，并鼓励通过艺术、极限运动或深层科学探索来重构人生的目标感。 🌐 播客信息补充 本播客采用原有人声声线进行播客音频制作，也可能会有一些地方听起来怪怪的 使用 AI 进行翻译，因此可能会有一些地方不通顺； 如果有后续想要听中文版的其他外文播客，也欢迎联系微信：iEvenight 在小宇宙查看该单集文稿

Demis Hassabis is the CEO of Google DeepMind. Hassabis joins Big Technology Podcast to discuss where AI progress really stands today, where the next breakthroughs might come from, and whether we’ve hit AGI already. Tune in for a deep discussion covering the latest in AI research, from continual learning to world models. We also dig into product, discussing Google’s big bet on AI glasses, its advertising plans, and AI coding. We also cover what AI means for knowledge work and scientific discovery. Hit play for a wide-ranging, high-signal conversation about where AI is headed next from one of the leaders driving it forward. --- Enjoying Big Technology Podcast? Please rate us five stars ⭐⭐⭐⭐⭐ in your podcast app of choice. Want a discount for Big Technology on Substack + Discord? Here’s 25% off for the first year: https://www.bigtechnology.com/subscribe?coupon=0843016b Learn more about your ad choices. Visit megaphone.fm/adchoices

Entrevista a Demis Hassabis, CEO de Google DeepMind y premio Nobel por sus aportaciones al diseño de proteínas con IA, quien analiza el progreso actual y los desafíos futuros de la inteligencia artificial general (AGI). El texto destaca hitos científicos como AlphaFold y el desarrollo de Gemini 3, subrayando el potencial de la IA para revolucionar la ciencia de materiales, la fusión nuclear y la medicina. Hassabis explica la importancia de los modelos de mundo y la simulación para otorgar a las máquinas una comprensión física y espacial más profunda. Asimismo, aborda temas críticos como la seguridad, la necesidad de consistencia en el razonamiento y el impacto socioeconómico que podría traer una revolución tecnológica diez veces más rápida que la industrial. Finalmente, reflexiona sobre la consciencia, la naturaleza computable del universo y su misión personal de guiar esta tecnología hacia un beneficio global y responsable.

Demis Hassabis is the CEO of Google DeepMind. He joined Big Technology Podcast in early 2025 discuss the cutting edge of AI and where the research is heading. In this conversation, we cover the path to artificial general intelligence, how long it will take to get there, how to build world models, whether AIs can be creative, and how AIs are trying to deceive researchers. Stay tuned for the second half where we discuss Google's plan for smart glasses and Hassabis's vision for a virtual cell. Hit play for a fascinating discussion with an AI pioneer that will both break news and leave you deeply informed about the state of AI and its promising future. --- Enjoying Big Technology Podcast? Please rate us five stars ⭐⭐⭐⭐⭐ in your podcast app of choice. Want a discount for Big Technology on Substack + Discord? Here’s 25% off for the first year: https://www.bigtechnology.com/subscribe?coupon=0843016b Questions? Feedback? Write to: bigtechnologypodcast@gmail.com --- Wealthfront.com/bigtech. If eligible for the overall boosted 3.90% rate offered with this promo, your boosted rate is subject to change if the 3.25% base rate decreases during the 3-month promo period. The Cash Account, which is not a deposit account, is offered by Wealthfront Brokerage LLC ("Wealthfront Brokerage"), Member FINRA/SIPC, not a bank. The Annual Percentage Yield ("APY") on cash deposits as of 12/19/25, is representative, requires no minimum, and may change at any time. The APY reflects the weighted average of deposit balances at participating Program Banks, which are not allocated equally. Wealthfront Brokerage sweeps cash balances to Program Banks, where they earn the variable base APY. Instant withdrawals are subject to certain conditions and processing times may vary. Learn more about your ad choices. Visit megaphone.fm/adchoices

Welcome to episode 146 of the AI for Career Success podcast from Curt Robbins. This educational content is designed to give working professionals who leverage AI as a tool for efficiency and productivity a competitive edge. In this episode, hosts Daphne and Fred unpack a December 16, 2025 interview with Google DeepMind CEO Demis Hassabis. Hassabis discusses the transition from large language models to agentic AI and the pursuit of Artificial General Intelligence (AGI). He emphasizes that reaching AGI requires a balanced approach of scaling and innovation, with a particular focus on world models that allow AI to understand physical dynamics through simulation. Hassabis highlights the potential for AI to solve "root node" problems in science and energy, such as achieving viable fusion and discovering new materials. He also addresses the risks of the current commercial race, advocating for a scientific persona in AI to combat hallucinations and misinformation. Ultimately, Hassabis views AI as a tool to explore the limits of computability and better understand the unique qualities of the human mind. _________________________________ "It will not be AI that takes away the job of a technical writer, but rather another technical writer with deep AI skills," said Robbins. I am currently taking on new clients. I enjoy helping companies with their documentation and communications strategy and implementation. Contact me to learn about my reasonable rates and fast turnaround. — Curt _________________________________ >> Read the Robbins article "The Global AI Race: America vs. China": https://tinyurl.com/2uckj7wy >> Read the Robbins article "Understanding AI Hallucinations in Technical Writing": https://tinyurl.com/bdeyd64t >> Read the Robbins article "Yale Study: Impact of AI on the Job Market": https://tinyurl.com/f3cuvvxn >> Read the Robbins article "Why Large Language Models are Changing the World": https://tinyurl.com/bdfv63ca >> Read the Robbins article "Understanding Anthropic: Rising Star in AI": https://tinyurl.com/46btw22z >> Read the Robbins article "Comparing ChatGPT, Gemini, Copilot, & Grok": https://tinyurl.com/3zwttxhk >> Read the Robbins article "AI Job Replacement Fears Are Good. Here's Why.": https://tinyurl.com/p5t27t7d >> Join the LinkedIn group Technical Writing Success: https://tinyurl.com/mr28u7td >> Subscribe to the Technical Writing Success podcast: https://tinyurl.com/uu9hpyzt

📝 本期播客简介 本期节目，主持人 Hanna Fry 教授与 Google DeepMind 联合创始人兼 CEO Demis Hassabis 展开了一场关于人工智能未来走向的深度对话。Demis Hassabis 作为AI领域的领军人物，将带我们跳出日常新闻，思考AI的终极目标和发展路径。他分享了对通用人工智能（AGI）的看法，探讨了AI在科学、商业和社会层面的深远影响，以及如何应对随之而来的挑战。 翻译克隆自： The future of intelligence | Demis Hassabis (Co-founder and CEO of DeepMind) 👨‍⚕️ 本期嘉宾 Demis Hassabis，Google DeepMind 联合创始人兼 CEO。他是一位神经科学家、人工智能研究员、视频游戏设计师和企业家，被广泛认为是全球领先的AI思想家之一。 ⏱️ 时间戳 开场与播客简介 00:00 欢迎收听：跨国串门计划与本期节目介绍 02:03 Demis Hassabis 精彩语录：AGI、心智与计算极限 AI 领域的最新进展与未来愿景 02:57 2023年AI回顾：从语言模型到Agent AI的重心转移 03:51 AI的飞速发展：Gemini 3与世界模型带来的惊喜 04:25 “根节点问题”：AI如何解锁科学与医学的下游效益 04:46 探索前沿：材料科学、核聚变与量子计算的突破 05:59 核聚变的深远影响：清洁能源与气候问题解决方案 通用人工智能的挑战与思考 07:15 AI的“参差不齐”智能：数学奥赛金牌与低级错误并存 08:03 AGI的缺失环节：推理一致性与持续学习能力 09:15 AlphaGo与AlphaZero：从学习人类知识到自我发现 11:07 科学研究与商业竞赛的平衡：AlphaFold与聊天机器人的路径选择 13:15 AI竞赛的加速效应：更多资源与技术普及 14:10 规模化瓶颈与合成数据：AI发展“撞墙”了吗？ 15:21 DeepMind的优势：研究优先、世界级工程与基础设施 16:19 AGI之路：规模化与创新并重 16:49 AI“幻觉”问题：AlphaFold的置信度机制能否借鉴？ 18:37 世界模型的重要性：语言模型无法捕捉的物理世界动态 19:57 模拟世界的构建与验证：Genie和VEO的逼真生成能力 21:31 AI Agent的演化实验：Simmer与Genie的互动循环 23:23 确保模拟世界的真实性：物理基准测试与幻觉控制 25:30 意识的起源：模拟Agent演化实验的设想 27:10 涌现属性的风险：在安全沙箱中运行模拟 AI 对社会与人类的深远影响 28:00 AI泡沫与长期价值：短期高估，长期低估 29:56 避免社交媒体覆辙：构建以用户为中心的负责任AI 31:34 AI人格的科学：平衡支持与挑战不合逻辑观点 33:35 AGI的融合之路：语言模型、图像模型与世界模型的结合 34:46 工业革命的启示：AI带来的社会冲击将更剧烈、更迅速 36:52 后AGI时代：经济体系与人类意义的重构 38:53 国际合作与AI安全：地缘政治挑战与潜在的“警钟” 42:56 人类与机器的界限：图灵机的极限与意识的本质 46:17 顶尖AI研究者的心路历程：兴奋、责任与挑战 48:40 AI领域的竞争与合作：超越商业成败的更高 stakes 49:39 未来十年：Agent系统的风险与期待 51:20 终极使命：安全引导AGI到来后的“学术假” 🌐 播客信息补充 本播客采用原有人声声线进行播客音频制作，也可能会有一些地方听起来怪怪的 使用 AI 进行翻译，因此可能会有一些地方不通顺； 如果有后续想要听中文版的其他外文播客，也欢迎联系微信：iEvenight 在小宇宙查看该单集文稿

The provided text offers excerpts from a Google DeepMind podcast transcript featuring an interview with Demis Hassabis , the CEO and co-founder, who discusses the current landscape and future trajectory of Artificial Intelligence. Hassabis explains that DeepMind is currently focusing on both scaling and innovation to achieve Artificial General Intelligence (AGI), with research efforts also geared toward solving major scientific problems such as fusion energy and advanced material science using AI. The discussion highlights the recent rapid advances in large language models, the development of agent-based and world models like Gemini and Genie for better understanding physical dynamics, and the challenges remaining, such as ensuring consistency and reducing hallucinations in current systems. Furthermore, Hassabis addresses the crucial societal and economic implications of AGI, including the need for new models to manage the disruption that will likely occur faster and on a larger scale than the Industrial Revolution.

The provided text offers excerpts from a Google DeepMind podcast transcript featuring an interview with Demis Hassabis , the CEO and co-founder, who discusses the current landscape and future trajectory of Artificial Intelligence. Hassabis explains that DeepMind is currently focusing on both scaling and innovation to achieve Artificial General Intelligence (AGI), with research efforts also geared toward solving major scientific problems such as fusion energy and advanced material science using AI. The discussion highlights the recent rapid advances in large language models, the development of agent-based and world models like Gemini and Genie for better understanding physical dynamics, and the challenges remaining, such as ensuring consistency and reducing hallucinations in current systems. Furthermore, Hassabis addresses the crucial societal and economic implications of AGI, including the need for new models to manage the disruption that will likely occur faster and on a larger scale than the Industrial Revolution.

(0:00) Introducing Sir Demis Hassabis, reflecting on his Nobel Prize win (2:39) What is Google DeepMind? How does it interact with Google and Alphabet? (4:01) Genie 3 world model (9:21) State of robotics models, form factors, and more (14:42) AI science breakthroughs, measuring AGI (20:49) Nano-Banana and the future of creative tools, democratization of creativity (24:44) Isomorphic Labs, probabilistic vs deterministic, scaling compute, a golden age of science Thanks to our partners for making this happen! Solana - Solana is the high performance network powering internet capital markets, payments, and crypto applications. Connect with investors, crypto founders, and entrepreneurs at Solana's global flagship event during Abu Dhabi Finance Week & F1: solana.com/breakpoint. https://solana.com/ OKX - The new way to build your crypto portfolio and use it in daily life. We call it the new money app. https://www.okx.com/ Google Cloud - The next generation of unicorns is building on Google Cloud's industry-leading, fully integrated AI stack: infrastructure, platform, models, agents, and data. https://cloud.google.com/ IREN - IREN AI Cloud, powered by NVIDIA GPUs, provides the scale, performance, and reliability to accelerate your AI journey. https://iren.com/ Oracle - Step into the future of enterprise productivity at Oracle AI Experience Live. https://www.oracle.com/ Circle - The America-based company behind USDC — a fully-reserved, enterprise-grade stablecoin at the core of the emerging internet financial system. https://www.circle.com/ BVNK - Building stablecoin-powered financial infrastructure that helps businesses send, store, and spend value instantly, anywhere in the world. https://www.bvnk.com/ Polymarket: https://www.polymarket.com/ Athletic Brewing: https://athleticbrewing.com/ Follow Demis: https://x.com/demishassabis Follow the besties: https://x.com/chamath https://x.com/Jason https://x.com/DavidSacks https://x.com/friedberg Follow on X: https://x.com/theallinpod Follow on Instagram: https://www.instagram.com/theallinpod Follow on TikTok: https://www.tiktok.com/@theallinpod

Demis Hassabis, a pioneer in artificial intelligence, is shaping the future of humanity. As the CEO of Google DeepMind, he was first interviewed by correspondent Scott Pelley in 2023, during a time when chatbots marked the beginning of a new technological era. Since that interview, Hassabis has made headlines for his innovative work, including using an AI model to predict the structure of proteins, which earned him a Nobel Prize. Pelley returns to DeepMind’s headquarters in London to discuss what’s next for Hassabis, particularly his leadership in the effort to develop artificial general intelligence (AGI) – a type of AI that has the potential to match the versatility and creativity of the human brain. Fertility rates in the United States are currently near historic lows, largely because fewer women are having children in their 20s. As women delay starting families, many are opting for egg freezing, the process of retrieving and freezing unfertilized eggs, to preserve their fertility for the future. Does egg freezing provide women with a way to pause their biological clock? Correspondent Lesley Stahl interviews women who have decided to freeze their eggs and explores what the process entails physically, emotionally and financially. She also speaks with fertility specialists and an ethicist about success rates, equity issues and the increasing market potential of egg freezing. This is a double-length segment.

L'intelligence artificielle est-elle sur le point de franchir un cap historique ? À l’occasion du sommet pour l’action sur l’IA, j’ai assisté à une rencontre exceptionnelle entre deux figures majeures du secteur : Demis Hassabis, cofondateur et PDG de DeepMind, et James Manyika, vice-président de Google chargé de la recherche. Ensemble, ils ont partagé leur vision des opportunités et des risques liés à l’essor de l’IA, en particulier sur la route vers l’intelligence artificielle générale (AGI). Rediffusion du 14/02/2025 Dans cet échange captivant organisé par Google France, les deux intervenants reviennent sur les bénéfices actuels de l’IA, notamment dans le diagnostic médical dans les pays en développement, et sur la promesse d’un assistant numérique universel. Ils abordent également la perspective de systèmes intelligents capables d’exécuter des tâches complexes, et l'impact à venir sur le marché du travail. Mais cette évolution s'accompagne aussi de sérieux défis : sécurité, gouvernance, dérives possibles, éthique... Demis Hassabis insiste sur la nécessité de mettre en place des garde-fous et de s'assurer que les systèmes d'IA intègrent les bonnes valeurs. James Manyika, lui, appelle à anticiper dès maintenant les effets sur la société et à investir dans la formation. Hébergé par Audiomeans. Visitez audiomeans.fr/politique-de-confidentialite pour plus d'informations.

📝 本期播客简介 本期，Lex Fridman 再次邀请到谷歌 DeepMind 的联合创始人兼 CEO、新晋诺贝尔奖得主 Demis Hassabis，进行了一场关于人工智能、宇宙奥秘和人类未来的深度对话。Demis 分享了他在诺贝尔奖演讲中提出的一个宏大猜想：自然界中所有由演化和物理规律塑造的模式，原则上都可以被经典 AI 高效学习。从蛋白质折叠到星系形成，宇宙的结构本身或许就是可计算的。对话深入探讨了 AI 的惊人涌现能力，例如视频模型 Veo 如何在未被明确教导的情况下发展出“直觉物理学”，这不仅挑战了我们对“理解”的定义，也预示了通往真正“世界模型”的道路。Demis 还以一位资深游戏设计师的视角，描绘了由 AI 驱动的、拥有无限可能性的动态游戏世界的未来。这不仅是一场关于 AGI 何时到来、如何定义的思辨，更是一次对 AI 如何解决能源危机、带领人类进入“激进富足”时代、甚至帮助我们理解生命起源和意识本质的宏大展望。 👨‍🔬 本期嘉宾 Demis Hassabis 博士，谷歌 DeepMind 的联合创始人兼 CEO，人工智能领域的领军人物，因其在蛋白质结构预测方面的开创性工作而荣获诺贝尔化学奖。他的研究横跨神经科学、计算机科学和人工智能，致力于构建通用人工智能（AGI）以加速科学发现，解决人类面临的最重大挑战。 📒 文字版精华 见 微信公众号（点击跳转） 🌟 精彩内容 🌌 自然的终极密码：为何宇宙万物皆可被 AI 高效学习？ Demis 阐述其诺贝尔演讲中的核心猜想：自然并非随机，无论是生物演化还是物理风化，其产物都具有内在结构。这种“稳定者生存”的原则使得宇宙万物（从蛋白质到山脉）的模式可以被神经网络有效学习和建模，从而将许多看似无法处理的组合爆炸问题变得 tractable。 “如果真是这样，那么就应该存在某种可以被你反向学习的模式。实际上，这就像一个‘流形’，能帮助你搜索到正确的解…因为它不是一个随机的模式…我认为神经网络就有可能学会它。” 🎥 AI 的“直觉物理学”：为何视频模型 Veo 能在不“理解”的情况下模拟现实？ 一个令人惊讶的发现是，像 Veo 这样的视频生成模型，仅通过观看海量视频，就能相当准确地模拟液体、光照和材质等复杂物理现象。Demis 认为，这表明 AI 发展出了一种“直觉物理”模型，挑战了“必须通过与世界互动才能理解物理”的传统观念，也暗示了现实世界可能存在可被学习的低维结构。 “它至少有某种直觉物理的概念，一种关于事物应该如何运作的直觉理解。可能就像一个人类小孩理解物理的方式，而不是一个博士生能真正解构所有方程式的那种理解。” 🎮 终极开放世界：AI 将如何创造为你量身定制、无限可能的动态游戏？ 作为资深游戏设计师（曾主导《主题公园》、《黑与白》），Demis 展望了 AI 将彻底改变电子游戏。未来的开放世界将不再是预设脚本，而是由 AI 实时生成内容、动态调整故事线，为每个玩家创造独一无二、真正由其选择驱动的沉浸式体验。 “现在，在未来五到十年内，我们或许正处于一个临界点，即将拥有能够真正围绕你的想象力进行创造的 AI 系统。它们可以动态地改变故事，围绕你的选择来叙述…” 💡 何为真正的 AGI？“第 37 手”式的灯塔时刻与终极图灵测试 Demis 预测 AGI 可能在 2030 年左右出现，但他对其标准极高。真正的 AGI 不仅要在各项任务中表现优异，更要具备人类顶尖的“研究品味”和创造力。衡量标准将是那些“灯塔时刻”：它能否像 AlphaGo 走出“第 37 手”一样提出全新的科学假说，甚至发明一个像围棋一样深邃优雅的新游戏？ “提出一个好的猜想，比解决它更难…一个系统能提出一个值得研究的猜想吗？…那是一种远为困难的创造力。我们真的不知道。” ✨ 激进富足的时代：AI 如何解决能源危机，带领人类走向星辰？ Demis 认为，AI 最重要的应用之一是解决能源（如核聚变）、新材料等根本性科学难题。一旦成功，人类将进入一个资源不再稀缺的“激进富足”时代。届时，水资源、太空旅行成本等问题将迎刃而解，为人类文明走向星辰、实现卡尔·萨根的宇宙梦想铺平道路。 “这将是人类历史上第一次，我们不再受资源限制。我认为那对人类来说可能是一个全新的、了不起的时代，一个不再是零和博弈的时代。” 🌐 播客信息补充 翻译克隆自： #475 – Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games 本播客采用原有人声声线进行播客音频制作，也可能会有一些地方听起来怪怪的 使用 AI 进行翻译，因此可能会有一些地方不通顺； 如果有后续想要听中文版的其他外文播客，也欢迎联系微信：iEvenight 在小宇宙查看该单集文稿

Demis Hassabis is the CEO of Google DeepMind and Nobel Prize winner for his groundbreaking work in protein structure prediction using AI. Thank you for listening ❤ Check out our sponsors: https://lexfridman.com/sponsors/ep475-sc See below for timestamps, transcript, and to give feedback, submit questions, contact Lex, etc. Transcript: https://lexfridman.com/demis-hassabis-2-transcript CONTACT LEX: Feedback – give feedback to Lex: https://lexfridman.com/survey AMA – submit questions, videos or call-in: https://lexfridman.com/ama Hiring – join our team: https://lexfridman.com/hiring Other – other ways to get in touch: https://lexfridman.com/contact EPISODE LINKS: Demis’s X: https://x.com/demishassabis DeepMind’s X: https://x.com/GoogleDeepMind DeepMind’s Instagram: https://instagram.com/GoogleDeepMind DeepMind’s Website: https://deepmind.google/ Gemini’s Website: https://gemini.google.com/ Isomorphic Labs: https://isomorphiclabs.com/ The MANIAC (book): https://amzn.to/4lOXJ81 Life Ascending (book): https://amzn.to/3AhUP7z SPONSORS: To support this podcast, check out our sponsors & get discounts: Hampton: Community for high-growth founders and CEOs. Go to https://joinhampton.com/lex Fin: AI agent for customer service. Go to https://fin.ai/lex Shopify: Sell stuff online. Go to https://shopify.com/lex LMNT: Zero-sugar electrolyte drink mix. Go to https://drinkLMNT.com/lex AG1: All-in-one daily nutrition drink. Go to https://drinkag1.com/lex OUTLINE: (00:00) – Introduction (00:29) – Sponsors, Comments, and Reflections (08:40) – Learnable patterns in nature (12:22) – Computation and P vs NP (21:00) – Veo 3 and understanding reality (25:24) – Video games (37:26) – AlphaEvolve (43:27) – AI research (47:51) – Simulating a biological orga

🎙️ Demis Hassabis sur l’IA et l’AGI Dans cet épisode, on entre dans la tête de Demis Hassabis , CEO de DeepMind , pour comprendre où va l’intelligence artificielle … et ce qui nous attend avec l’ AGI . 🧠 Au menu : L’évolution de Gemini et les ambitions de DeepMind Pourquoi l’ intelligence artificielle générale n’est plus de la science-fiction Comment l’IA peut aider à résoudre des défis scientifiques et sociétaux majeurs Les dangers de la vitesse de développement et les inquiétudes sur la sécurité La complémentarité entre grands modèles généralistes et outils spécialisés L’avenir de l’ éducation, du travail et… de la créativité humaine face à des machines toujours plus capables ✨ Une conversation à la fois lucide, inspirante et vertigineuse.

This week, we take a field trip to Google and report back about everything the company announced at its biggest show of the year, Google I/O. Then, we sit down with Google DeepMind’s chief executive and co-founder, Demis Hassabis, to discuss what his A.I. lab is building, the future of education, and what life could look like in 2030. Guest : Demis Hassabis, co-founder and chief executive of Google DeepMind Additional Reading: At Google I/O, everything is changing and normal and scary and chill Google Unveils A.I. Chatbot, Signaling a New Era for Search Google DeepMind C.E.O. Demis Hassabis on the Path From Chatbots to A.G.I. We want to hear from you. Email us at hardfork@nytimes.com . Find “Hard Fork” on YouTube and TikTok . Subscribe today at nytimes.com/podcasts or on Apple Podcasts and Spotify. You can also subscribe via your favorite podcast app here https://www.nytimes.com/activate-access/audio?source=podcatcher . For more podcasts and narrated articles, download The New York Times app at nytimes.com/app. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

Explore o futuro da Inteligência Artificial com Demis Hassabis, cofundador e CEO da Google DeepMind. Nesta conversa, ele compartilha sua visão sobre a busca pela Inteligência Artificial Geral (AGI), que, segundo ele, pode estar a apenas 5 a 10 anos de distância — uma tecnologia que "vai mudar praticamente tudo sobre a maneira como fazemos as coisas". Hassabis descreve como a IA está avançando em uma “curva exponencial”, com aprendizados emergentes e capacidades cada vez mais impressionantes: sistemas que veem, ouvem e conversam sobre o mundo, como no Projeto Astra, e robôs capazes de raciocinar a partir de instruções vagas. Ele também destaca conquistas revolucionárias da DeepMind, como a decodificação da estrutura das proteínas — trabalho que rendeu à equipe um Prêmio Nobel — e o potencial de acelerar descobertas médicas, quem sabe até curando todas as doenças no futuro e promovendo uma “abundância radical”. Apesar do entusiasmo, Hassabis também expressa preocupações importantes sobre segurança, controle e os riscos associados à corrida global pelo domínio da IA. Uma conversa essencial sobre aquela que pode ser a ferramenta mais poderosa para expandir o conhecimento humano.

Bird flu, which has long been an emerging threat, took a significant turn in 2024 with the discovery that the virus had jumped from a wild bird to a cow. In just over a year, the pathogen has spread through dairy herds and poultry flocks across the United States. It has also infected people, resulting in 70 confirmed cases, including one fatality. Correspondent Bill Whitaker spoke with veterinarians and virologists who warn that, if unchecked, this outbreak could lead to a new pandemic. They also raise concerns about the Biden administration’s slow response in 2024 and now the Trump administration’s decision to lay off over 100 key scientists. Demis Hassabis, a pioneer in artificial intelligence, is shaping the future of humanity. As the CEO of Google DeepMind, he was first interviewed by correspondent Scott Pelley in 2023, during a time when chatbots marked the beginning of a new technological era. Since that interview, Hassabis has made headlines for his innovative work, including using an AI model to predict the structure of proteins, which earned him a Nobel Prize. Pelley returns to DeepMind’s headquarters in London to discuss what’s next for Hassabis, particularly his leadership in the effort to develop artificial general intelligence (AGI) – a type of AI that has the potential to match the versatility and creativity of the human brain. One of the most awe-inspiring and mysterious migrations in the natural world is currently taking place, stretching from Mexico to the United States and Canada. This incredible spectacle involves millions of monarch butterflies embarking on a monumental aerial journey. Correspondent Anderson Cooper reports from the mountains of Mexico, where the monarchs spent the winter months sheltering in trees before emerging from their slumber to take flight.

How can AI help us understand and master deeply complex systems—from the game Go, which has 10 to the power 170 possible positions a player could pursue, or proteins, which, on average, can fold in 10 to the power 300 possible ways? This week, Reid and Aria are joined by Demis Hassabis. Demis is a British artificial intelligence researcher, co-founder, and CEO of the AI company, DeepMind. Under his leadership, DeepMind developed Alpha Go, the first AI to defeat a human world champion in Go and later created AlphaFold, which solved the 50-year-old protein folding problem. He's considered one of the most influential figures in AI. Demis, Reid, and Aria discuss game theory, medicine, multimodality, and the nature of innovation and creativity. For more info on the podcast and transcripts of all the episodes, visit https://www.possible.fm/podcast/ Listen to more from Possible here . Learn more about your ad choices. Visit podcastchoices.com/adchoices