Sam Altman on OpenAI, ChatGPT Origins & AI Hardware
.webp)
Eric Walker · 11, July 2025
“We got very close to not starting OpenAI.”
In a recent Y Combinator interview, Sam Altman, CEO of OpenAI, shared his journey, the origins of ChatGPT, and his vision for the future of artificial intelligence. Conducted in front of an audience of 3,000 aspiring entrepreneurs, the discussion offered a candid look at the challenges and opportunities of building transformative AI technologies.
The Genesis of OpenAI: A Leap of Faith
Deciding to Pursue AGI
Altman reflected on the audacious decision to found OpenAI in 2015, a time when artificial general intelligence (AGI) seemed like a distant dream. “We got very close to not starting OpenAI,” he admitted, noting that the idea of pursuing AGI was considered “crazy” by 99% of the world. At the time, AI research was nascent—focused on rudimentary tasks like playing video games or solving a Rubik’s cube with a robotic hand. DeepMind, a competitor, appeared far ahead, and OpenAI had no clear path to revenue or products. Yet, a small group of believers, including Altman, looked at each other in the eye and decided to take the plunge. This moment of conviction, he emphasized, is critical for ambitious ventures. “When in doubt, lean into those moments,” he advised, underscoring the importance of resilience in the face of skepticism.
Rallying Talent for a Moonshot
Despite the odds, OpenAI attracted top talent by offering a unique mission. “If you’re doing a one-of-one thing, you have a really nice tailwind,” Altman explained. The 1% of the world that resonated with OpenAI’s vision included some of the brightest minds in AI, drawn to a mission that was both improbable and profoundly impactful. This ability to concentrate talent around a singular, ambitious goal gave OpenAI an early edge.
The Evolution of ChatGPT: From Sci-Fi to Reality
Early Days of Uncertainty
OpenAI didn’t start as the juggernaut it is today. In its infancy, it was “eight people in a room,” then 20, grappling with how to write impactful research papers. The idea of ChatGPT—a conversational AI capable of human-like interactions—was pure science fiction in 2015. Altman recalled the era’s “vibes” about AI: primitive language models, no viable products, and widespread doubt about scaling laws, which posited that larger models trained on more data would yield better performance. These laws, now a cornerstone of AI development, were contrarian at the time but proved pivotal. “It’s hard to explain now how improbable it seemed,” Altman said, reflecting on how ChatGPT’s success made the impossible seem inevitable.
Surprising Use Cases and Memory
Fast forward to 2025, and ChatGPT’s capabilities have soared. Altman highlighted the “product overhang”—a gap between what current models like GPT-3 and GPT-4 can do and the products yet to be built. He noted that recent models, like OpenAI’s o3, have unlocked new possibilities, particularly in reasoning and memory. “Memory is my favorite feature we’ve launched this year,” he said, describing how it makes ChatGPT feel like a companion that “knows” the user. This shift toward persistent, proactive AI—capable of running in the background, connecting to personal data, and acting on a user’s behalf—hints at a future where AI feels like an extension of oneself.
Free ChatGPT available on GlobalGPT, an all-in-one AI platform.
OpenAI's API available at OpenAI developer platform.
The API Revolution and Cost Declines
Altman also discussed the transformative potential of OpenAI’s API, which has seen dramatic cost reductions. “Last week, o3 cost five times as much as it does this week,” he revealed, predicting further price drops. An upcoming open-source model promises to run powerful AI locally, opening new doors for developers. This aligns with a 2025 TechCrunch report that OpenAI’s API cost reductions have spurred a wave of AI startups. Altman sees this as a golden moment for entrepreneurs to build on these advancements, leveraging reasoning models to create novel applications.
The Future of AI: Convergence and Embodiment
Multimodal Models and AGI
Looking ahead, Altman envisions a convergence of reasoning and multimodal capabilities—models that can reason deeply, generate real-time video, and write code on the fly. “That will feel like a real new kind of computer interface,” he said. While GPT-5 won’t fully realize this vision, it’s a step toward an integrated model that combines vision, speech, and reasoning. This trajectory points to embodied AI—robots powered by these models. Altman speculated about a future where a high-tier ChatGPT subscription might include a free humanoid robot, a nod to OpenAI’s robotics ambitions. A 2024 Forbes piece highlights OpenAI’s exploration of robotics, suggesting partnerships with hardware firms are on the horizon (source: Forbes, “OpenAI’s Next Frontier: Robotics,” 2024).
AI as a New Interface
Altman drew inspiration from sci-fi, imagining interfaces that “melt away.” Unlike today’s notification-heavy smartphones, future AI could handle tasks autonomously, surfacing information only when needed. He cited collaboration with designer Jony Ive as a step toward reimagining human-computer interaction. “AI opens the playing field for something completely new,” he said, hinting at a forthcoming device that could redefine how we engage with technology. This vision aligns with a 2025 The Verge report on OpenAI’s hardware ambitions, noting Ive’s involvement in crafting an AI-driven device.
Building Defensible AI Startups
Avoiding the OpenAI Trap
Altman addressed a common concern: how to build an AI startup that doesn’t get “run over” by OpenAI. “We don’t want to run you over,” he reassured, encouraging founders to avoid replicating ChatGPT and instead explore untapped opportunities. He suggested OpenAI could become a platform, driving traffic to startups via an app store-like model or “Sign-in with OpenAI” feature. This echoes a 2025 Bloomberg analysis, which notes that OpenAI’s platform strategy is fostering a startup ecosystem (source: Bloomberg, “OpenAI’s Platform Push: A New Era for AI Startups,” 2025).
Contrarian Thinking and Defensibility
Drawing on Peter Thiel’s philosophy of being “contrarian but right,” Altman urged founders to pursue unique ideas. “The most enduring companies are usually not doing the same thing as everybody else,” he said. OpenAI’s own defensibility evolved over time—from being the first mover to building brand recognition and features like memory. For startups, he advised focusing on niche markets with big potential, iterating quickly, and leveraging AI’s rapid evolution to outpace larger competitors.
AI’s Societal Impact: Abundance and Science
A New Era of Productivity
Altman sees AI as a tool for radical abundance, akin to the transistor’s impact on computing. “The whole arc of technology is that one person can do way more than before,” he said, predicting that AI will empower individuals and small teams to achieve unprecedented outcomes. By reducing coordination costs, AI could unlock exponential productivity gains, reshaping how we work and create value. A 2024 Harvard Business Review study supports this, estimating that AI could boost global productivity by 20% by 2030 (source: Harvard Business Review, “AI’s Productivity Promise,” 2024).
AI for Scientific Discovery
Altman’s personal passion lies in AI-driven scientific discovery. “If we could vastly increase the rate of new scientific discovery with AI, that would compound to incredible wonders for everyone’s lives,” he said. He believes that accelerating science, coupled with good governance, is the key to sustainable economic growth. OpenAI’s work on tools like Code Interpreter and Deep Research aligns with this vision, enabling complex workflows that could transform research.
Lessons from the Trenches
Hiring for Slope, Not Y-Intercept
On hiring, Altman emphasized seeking driven, curious, and team-oriented individuals with a strong track record of accomplishment. “Hire for slope, not y-intercept,” he said, quoting Y Combinator’s Paul Buchheit. He prioritizes velocity and problem-solving over polished resumes, especially early in a startup’s life. This approach has been key to OpenAI’s ability to attract talent, even when its mission seemed far-fetched.
The Hardship of Leadership
As CEO, Altman faces relentless challenges, from managing OpenAI’s massive infrastructure to navigating competition. “The number of things we have to do at the same time is more context than I thought possible,” he admitted. Yet, he remains optimistic, driven by the belief that AI’s potential outweighs the difficulties.
Advice to Aspiring Founders
Altman closed with a rallying cry for the 3,000 attendees: “This is the best time ever in the history of technology to start a company.” He urged them to embrace the ground-shaking changes AI brings, iterate faster than incumbents, and build for a future where intelligence and energy are abundant. Reflecting on his own journey, he wished he’d better understood the importance of conviction and resilience. “The good parts are better than you think, and the hard parts are shockingly harder,” he said, likening entrepreneurship to parenthood.
Shaping the Future
Sam Altman’s journey with OpenAI—from a risky bet on AGI to the creation of ChatGPT—illustrates the power of vision, resilience, and contrarian thinking. As AI evolves, from reasoning models to embodied robots, the opportunities for innovation are vast. For the next generation of founders, Altman’s message is clear: seize this moment, build something unique, and contribute to a future of abundance. Whether you’re a scientist, engineer, or dreamer, the tools to shape the future are here—now it’s up to you to act.