All Segments
Bob McGrew: AI Agents And The Path To AGI
8 segments available
According to OpenAI's former Chief Research Officer Bob McGrew, reasoning and test-time compute will unlock more reliable and capable AI agents— and a clear path to scale to AGI. In this episode of “How To Build The Future,” YC President and CEO Garry Tan sits down with Bob to discuss the lessons learned from his time at OpenAI, scaling laws, his advice for startups, and what all of this means for the jobs of the future. Apply to Y Combinator: https://ycombinator.com/apply Chapters (Powered by https://bit.ly/chapterme-yc) - 00:00 - Intro 02:31 - Early OpenAI projects 04:29 - GPT-1 09:30 - Scaling laws 14:08 - AGI levels 18:01 - Startup Advice 21:37 - Palantir And The Early Days 25:07 - Future jobs
Segments Timeline
Defining AGI: Myths and Realities
Bob McGrew discusses the common misconceptions surrounding Artificial General Intelligence (AGI), including the belief that it will lead to mass unemployment. He emphasizes the importance of reasoning and test-time compute as mechanisms that will enable AI agents to perform tasks reliably, moving beyond the limitations of pre-training and data bottlenecks.
"if you ask people what AGI was they would say it's a model that you can actually interact with it passes the Turing test uh it can look at things it can write code can even draw an image for you yeah ..."
Early Days at OpenAI
Bob McGrew shares his journey to OpenAI, detailing his initial intentions to start a robotics company and how he ended up in a research lab. He reflects on the early projects at OpenAI, including teaching a robot to play checkers and the challenges faced in the pursuit of AGI.
"same time it was sort of what we could do at the time and so uh some of the early projects I worked on a robotics project where we took a robot hand a humanoid robot hand and we taught it to solve Rub..."
The Path to AGI: Learning from Games
McGrew explains the significance of early projects like solving Dota 2 and how these experiences contributed to the understanding of scaling in AI. He highlights the importance of using complex environments to help AI generalize beyond narrow tasks, paving the way for advancements in large language models.
"order to make this work and that became gpd1 and then uh after gpd1 seemed successful we brought in the ideas from DOTA and from the robot hand of training at you know larger and larger amounts of sca..."
Scaling Laws and Their Impact
In this segment, McGrew discusses the concept of scaling laws in AI research and how OpenAI's approach to scaling has been pivotal in developing successful models. He reflects on the early days of exploratory research and the balance between structured guidance and creative freedom within the organization.
"you're trying to make things better why didn't we uh take advantage of scaling laws earlier in these other domains well you know I think I think we were really trying to you know usually usually in or..."
Cultural Insights: OpenAI vs. Other Labs
McGrew contrasts the culture at OpenAI with that of other AI research labs like Google Brain and DeepMind. He emphasizes OpenAI's collaborative approach and the importance of credit distribution among researchers, which fostered a productive environment for innovation.
"real moment where you can uh hook that up to a biolab and have you know sort of autonomous uh exploration of you know scientific spaces what can you say about that stuff the really interesting thing a..."
Advice for Startups in AI
Bob McGrew offers valuable insights for startups looking to navigate the AI landscape. He discusses the significance of understanding scaling laws and how they can be leveraged to improve AI systems, encouraging entrepreneurs to focus on innovative applications of AI technology.
"gotten it to work then you can use distillation you can take a dumber model and you can try prompting it you can try to have the frontier model train the smaller model but you know the most important ..."
Reflections on Palantir and Early Experiences
In this segment, McGrew reflects on his time at Palantir and how those experiences shaped his understanding of AI and its applications. He shares anecdotes from the early days of his career, highlighting the lessons learned that influenced his work at OpenAI.
"agencies some of the most impactful decisions that a society might have to make and you look around and there was no software in there and that was what paler and certainly paler government was very e..."
The Future of Jobs in an AI-Driven World
McGrew discusses the implications of AI advancements on the job market, addressing concerns about automation and employment. He emphasizes the need for adaptation and the potential for new job creation as AI technologies evolve, offering a hopeful perspective on the future of work.
"it's going to do asteroid Mining and all sorts of cool stuff and so every day he says you know Dad can you teach me a little bit about how to code this is actually what I do most with language models ..."