Artificial Intelligence
Machines that learn. Neural networks and the deep-learning turn, ImageNet and AlphaGo, the transformer, ChatGPT and the models since, the compute race, and the arguments about what any of it means.
Part of Technology
28 moments in this segment.
- - IBM Watson vs Jeopardy Champions. February 14-16, 2011. IBM Watson beats Ken Jennings + Brad Rutter at Jeopardy. The first widely-recognised modern-AI public win.
- - AlphaGo vs Lee Sedol - Move 37. March 2016. Game 2 Move 37, AlphaGo's creative shoulder hit. Commentators thought it was a bug. Lee Sedol left the room. The single most-discussed AI move ever made.
- - Stable Diffusion - Launch Demo. August 2022. Emad Mostaque's Stability AI releases Stable Diffusion. Open-source; runs on a consumer GPU. The first generative-image model anyone could run locally.
- - ChatGPT Launch Tweet. November 30, 2022. Sam Altman: "today we launched ChatGPT. try talking with it here." 100M users in 60 days, fastest consumer-product growth in history.
- - Stable Diffusion. August 22, 2022. Stability AI releases Stable Diffusion, the first generative-image model anyone can run on a consumer GPU. Open-source. Triggers the AI-art era; transforms the entire creative industry within months.
- - GPT-4 Released. March 14, 2023. OpenAI releases GPT-4. Multimodal (text + image input), substantially more capable than GPT-3.5. Bar-exam-passing capability. The first AI model that knowledge workers built daily workflows around.
- - Gemini Fake Demo Controversy. December 6, 2023. Google's Gemini intro video edited to look more real-time than it was. The first major AI-marketing credibility scandal; framed how every subsequent AI launch was scrutinised.
- - Sam Altman Fired (and Rehired). November 17-22, 2023. OpenAI board fires Sam Altman on Friday. By Tuesday, 700+ OpenAI employees threaten to follow him to Microsoft. The board reverses. The five wildest days in 21st-century tech.
- - Sundar Pichai on Bard Launch. February 6, 2023. Sundar Pichai announces Bard, Google's ChatGPT competitor, three months after ChatGPT. Google's stock drops $100B the next day on a factual error in the demo.
- - OpenAI Sora. February 15, 2024. OpenAI demos Sora, minute-long photorealistic video from text prompts. Veo (Google, May 2024) and Runway Gen-3 (June 2024) follow within four months. The first commercial-quality text-to-video models.
- - GPT-4o Multimodal. May 13, 2024. OpenAI releases GPT-4o, natively multimodal across text, image, audio. Real-time voice conversation. The first AI assistant that feels like the science-fiction default.
- - Mira Murati Founds Thinking Machines. October 2024. Former OpenAI CTO Mira Murati founds Thinking Machines Lab, the highest-profile AI-research spinoff of the decade. Raises $1B at $9B valuation within 4 months. Signals AI talent's ongoing diffusion across labs.
- - OpenAI Sora - Tokyo Walk Demo. February 15, 2024. OpenAI demos Sora, first commercially-quality text-to-video model. The "Tokyo Walk" demo went viral globally; reshaped the AI-video debate overnight.
- - GPT-4o Demo - Live Voice. May 13, 2024. OpenAI demos GPT-4o's real-time voice mode. The Spring Update, natural speech, interruptions, emotional inflection. The first AI assistant that felt like sci-fi default.
- - Claude 3.5 Sonnet Demo - Artifacts. June 20, 2024. Anthropic launches Claude 3.5 Sonnet with Artifacts (side-pane previews of generated code/docs). Reframed AI chat as a workspace. Becomes the developer-AI of choice for the next year.
- - Meta Llama 3 Launch. April 18, 2024. Meta releases Llama 3, open-source flagship LLM. Mark Zuckerberg: "open-source is the future of AI." Sets the open-vs-closed-AI battle for the next 18 months.
- - Devin AI - First Software Engineer Agent. March 12, 2024. Cognition Labs unveils Devin, pitched as the first AI software engineer. Demo videos go viral; controversies follow about how doctored some demos were. Defined the agentic-AI hype cycle.
- - Mira Murati Announces Thinking Machines. October 2024. Former OpenAI CTO Mira Murati announces Thinking Machines Lab. Within 4 months: $1B raise at $9B valuation. Most-watched AI-lab founding announcement since OpenAI.
- - Sam Altman on Lex Fridman. March 18, 2024. Sam Altman's second Lex Fridman podcast. 2.5 hours. Covers the November 2023 board firing, Q*, AGI timelines. One of the most-referenced AI interviews of 2024.
- - Claude Computer Use Demo. October 22, 2024. Anthropic ships "Computer Use", Claude can take screenshots and control a mouse + keyboard. The first widely-deployed AI agent that operates a real computer. Defines the 2025 agent-AI roadmap.
- - OpenAI Sora API Open. December 9, 2024. OpenAI ships Sora to ChatGPT Plus + Pro subscribers. First widely-available text-to-video product to consumers.
- - DeepSeek is a Game Changer for AI - Computerphile. In January 2025 the open-source DeepSeek R1 reasoning model rivalled the best proprietary models at a fraction of the cost, upending the AI industry.
- - GPT-4o Image Generation. March 25, 2025. OpenAI releases GPT-4o's native image generation. Within hours: "Ghibli-style" pictures of every famous photograph flood social media. The most-discussed AI release of 2025.
- - Manus AI Agent Demo. March 2025. Chinese startup Manus releases an autonomous AI agent demo that goes viral globally. The agent-AI hype cycle peaks; thousands sign up for invite-only access in 24 hours.
- - Claude Opus 4 Launch. May 2025. Anthropic ships Claude Opus 4, frontier reasoning model. Sustained 7-hour autonomous coding tasks. The release that solidified Anthropic's position alongside OpenAI at the AI frontier.
- - Gemini 2.5 Pro Released. March 2025. Google DeepMind ships Gemini 2.5 Pro, first to top LMSYS leaderboard above OpenAI/Anthropic in all major categories. Google's comeback in the AI race becomes real.
- - GPT-5 Launch Demo. August 2025. OpenAI ships GPT-5, unified model spanning text + image + code + reasoning. The model-class boundary that closed most measurable gaps to human-expert performance.
- - GPT-5. 2025. OpenAI ships GPT-5, extended thinking, agentic workflows, unified text+image+code reasoning. The model class that closes most measurable gaps between AI and human-expert performance.
