ChatGPT launched as a quiet "research preview" whose own staff badly underestimated how popular it would be, and it became one of the fastest-growing apps in history. We tell the seven-year story behind it, from a nonprofit with a big mission through GPT-1, GPT-2, GPT-3, and the training trick that taught a model to follow instructions. Then we explore why the chat interface mattered as much as the technology, and how one launch set off a "code red" at Google and an industry-wide race. This episode includes AI-generated content.
In 2017, eight researchers, most of them at Google, published a paper with a catchy title and a radical idea: throw out the old way of processing language and rely on "attention" instead. We explain in plain language how attention lets an AI weigh every word against every other word, why the design made giant models like GPT possible, and how the "T" in ChatGPT traces back to a paper about translation. Plus the irony of history: Google invented the transformer, but OpenAI turned it into a household name. This episode includes AI-generated content.
OpenAI unveils LifeVal, a benchmark for whether ChatGPT can handle real-life tasks like school enrollment and home searches, and pledges its data centers will cut power use when the grid is strained. Anthropic's updated IPO filing reports about $17 billion in quarterly revenue and warns it may soon cross its own line for AI-accelerated AI research. Plus: The New York Times and three other papers sue Brave over scraped journalism sold to AI companies, and Marvell's $115 billion deal for Lumentum bets on light-speed connections inside AI data centers. This episode includes AI-generated content.
OpenAI is merging ChatGPT's Chat and Work modes and putting ChatGPT and Codex on one metered token allowance, while planning a post-midterm push in Washington and signing government deals in Abu Dhabi. We also look at Nvidia's reported plan to guarantee debt on Anthropic's West Virginia data center, the banks and retailers shutting out personal AI agents like Meta's Muse, and China's AI moves as Apple brings Qwen into Siri and DeepSeek open-sources an always-on agent design. This episode includes AI-generated content.
A roughly 51-minute deep dive into this week's biggest AI news: OpenAI cuts its funding round to $20B at a $1.25T valuation as AI debts soar, the GPT-6.1 Terra price cut and new ChatGPT Pro limits, Anthropic's Nasdaq IPO filing, Google's Gemini 4 Carbon and cloud growth, Nvidia and AMD investing in Meta, and Mistral Large 4 going open weights. This episode includes AI-generated content.