红杉资本播客 Training Data 开播：与 AI 建造者探讨世纪主题

It's been 18 months since Generative AI's Big Bang moment, and the AI universe has been expanding quickly. We're talking trillions of tokens, billion-dollar valuations, millions of GPUs, and hundreds of thousands of models and counting. The space is changing so fast it's hard to keep up, and every day brings new questions. Are LLMs showing emergent properties of reasoning? Are we approaching the limits of scaling laws? Is more inference compute the antidote? And can we ever actually understand AI, or will it always be a black box? No one really knows the answers to these important questions, but if there's any set of people who can help us triangulate towards a clearer picture, it's the guest on our new show, Training Data. I'm Sonia Huang, a partner at Sequoia Capital. Together with my co-host Pat Grady, we'll invite founders, builders, and researchers at the forefront of AI onto our new show and ask them the most pressing questions on our minds. We'll soak up their knowledge and perspective about the latest developments and opportunities, and we'll use these conversations to train our own neural nets about where the space may be going, the challenges and roadblocks along the way, and how those might be overcome. We're at the cusp of the biggest technological leap of our generation, and perhaps in history. Join us as we wrap our heads around AI's implications for technology, business, and society with the people best positioned to help us understand them. Training Data, coming soon, wherever you listen to podcasts. .