Search for a command to run...

Timestamps are as accurate as they can be but may be slightly off. We encourage you to listen to the full context.
This episode features Sam Altman's second appearance on Conversations with Tyler, recorded live at the Progress Conference. Tyler Cowen explores how Altman manages OpenAI's explosive growth, from hiring hardware specialists to developing new AI interfaces, while discussing the future of artificial intelligence and its impact on society. (01:24)
The conversation covers OpenAI's upcoming developments including GPT-6's potential for scientific research, the company's approach to commerce and partnerships, and Altman's evolving views on AI safety. Key themes include the balance between AI capabilities and human oversight, the transformation of work and education, and the economic implications of superintelligence. (07:00)
Tyler Cowen is a renowned economist and Professor at George Mason University. He co-hosts the popular podcast "Conversations with Tyler" and is known for his expertise in cultural economics and his ability to explore diverse topics with leading thinkers across multiple disciplines.
Sam Altman is the CEO of OpenAI, the company behind ChatGPT and GPT models. He previously served as president of Y Combinator, where he helped launch hundreds of startups. Under his leadership, OpenAI has become one of the most influential AI companies globally, pioneering the development of large language models and advancing toward artificial general intelligence.
Altman emphasizes that as demands and opportunities increase, finding ways to be more efficient becomes essential. (01:41) His approach involves hiring and promoting exceptional people, then delegating substantial responsibility to them. This principle applies differently when hiring hardware specialists versus AI researchers - hardware requires longer cycle times, more capital intensity, and higher setup costs, so Altman spends more time getting to know hardware people before giving them autonomy. The key insight is that effective delegation isn't just about finding smart people; it's about matching management style to the specific domain and understanding the unique constraints of each field.
Rather than simply adding AI features to existing tools, Altman envisions a fundamental shift where AI agents handle most routine work and only escalate issues when necessary. (06:15) He predicts that within a small number of years, we'll see billion-dollar companies run by just two or three people with AI assistance. This isn't about replacing humans entirely, but about AI handling the coordination, communication, and routine decision-making that currently consumes much of our workday. The transformation will be so significant that traditional office productivity tools like email, Slack, and even meetings may become largely obsolete.
While GPT-5 shows only tiny glimmers of AI doing new science, Altman expects GPT-6 to represent a major leap forward in scientific research capabilities. (07:02) He describes this as potentially being comparable to the jump from GPT-3 to GPT-4, but for scientific applications rather than general reasoning. This means researchers and lab managers should start preparing now by thinking about how to restructure their organizations to put AI at the center of scientific discovery, rather than just using it as an add-on tool. The implications extend beyond individual research to fundamentally changing how scientific institutions operate.
Altman explains that ChatGPT has become users' most trusted technology product despite being prone to hallucinations, because users pay for it directly rather than being the product sold to advertisers. (15:14) This trust relationship is crucial for OpenAI's commerce strategy - if ChatGPT accepts payment to recommend inferior hotels or products, it would destroy user trust. Instead, their approach involves taking transaction fees only after providing genuinely best recommendations, maintaining the alignment between user interests and OpenAI's revenue. This principle of transparent, user-aligned monetization will likely become the standard for AI-powered recommendation systems.
When asked about the binding constraint for more compute power, Altman's answer is simple: "Electrons. Just energy." (27:26) The short-term solution is natural gas, while long-term solutions will be dominated by fusion and solar power. This insight reveals that the AI revolution's pace isn't limited by our ability to design chips or build data centers, but by our capacity to generate sufficient electricity. For organizations planning AI infrastructure investments, energy sourcing and efficiency should be the primary consideration, not just computing hardware.