Bob McGrew is at the epicenter of the AI revolution.  As the VP of Research at OpenAI, he's instrumental in breakthroughs that are reshaping the world, from building GPT models and launching ChatGPT to overseeing the Dall-E project.  How was GPT-4 trained and what will GPT-5 look like? Why does ChatGPT respond with certain biases and how do they correct for that? What breakthrough led Bob to believe AGI could be achievable?

Bob and I were in Phi Psi together at Stanford and both interned at PayPal in its early days.  We hired Bob as the second engineer at Palantir, where he built and shipped the first products for the intelligence community and went on to lead engineering and help run the company.  Bob joined OpenAI part-time in 2016 and full-time in 2017, where he's been at the forefront of AI innovation.  In this episode, we discuss the early days of Palantir, how he knew AI's moment had arrived, and the most important research projects underway at OpenAI.  We also look ahead to what GPT-8 could unlock for humanity and what's needed for Large Language Models to move beyond mimicking humans to higher forms of intelligence and creativity.

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit blog.joelonsdale.com