Sam Altman on the Past, Present, and Unimaginable Future of AI

OpenAI’s CEO reflects on the origins of ChatGPT, the coming wave of AI agents, and why this is the best time in history to start a company.

The "Crazy" Bet on AGI

In 2015, the idea of pursuing Artificial General Intelligence (AGI) was considered a pipe dream by the vast majority of the tech world. Sam Altman and a small group of co-founders nearly didn't start OpenAI. The challenges were immense: DeepMind seemed impossibly far ahead, there were no clear products or revenue models, and the entire endeavor sounded, in Altman's words, "crazy."

The decision to proceed was a pivotal moment, one that Altman describes as common among highly ambitious projects. "They seem so difficult and there's such good reasons not to do them that it really takes a core of people that sit in a room, look each other in the eye and say, 'All right, let's do this.'" This contrarian bet allowed them to attract the 1% of the world's smartest talent who resonated with the mission, concentrating expertise that had nowhere else to go.

From Research Lab to Global Phenomenon

The early days of OpenAI were a far cry from its current status. The team of about eight, then twenty, people had "no ideas for products, no revenue, no really idea that we were ever going to have revenue." They were focused on writing research papers, working on projects like playing video games or a robotic hand that could "sort of barely do a Rubik's cube." The concept of a tool like ChatGPT was pure science fiction.

Altman emphasizes that massive companies never start as such. He quotes investor Vinod Khosla: "There's a very big difference between a $2 million startup and a $10 billion startup but they both have zero dollars of revenue." The key is to pick a market with the potential for massive scale and then put "one dumb foot in front of the other for a long time."

The Current AI Revolution: A Developer's Playground

According to Altman, we are currently in a period of significant "product overhang." The capabilities of the latest models, like GPT-4o, far exceed the products that have been built to utilize them. This gap represents a monumental opportunity for builders.

Key trends fueling this opportunity include:

• Dramatically Falling Costs: The price of API calls is plummeting. "Last week o3 cost five times as much as it did this week and that's going to keep going."
• Powerful Open-Source Models: OpenAI plans to release a new open-source model soon that will "astonish" people with its capability, enabling incredibly powerful models to run locally.
• The Rise of Reasoning Models: Models are evolving from simple query responders to agents that can perform multi-step tasks, akin to a "very junior employee."

This combination creates an "exceptional time to go build a company" that leverages these new foundational capabilities.

The Future Interface: From Chatbot to Companion

Altman's vision for the future of AI interaction is one where the interface "almost melts away." He critiques current phone interfaces as stressful, like "walking down Time Square in New York getting bumped into by people."

The future, enabled by features like Memory, is a proactive, always-on AI companion that knows you, connects to your data, and acts on your behalf without constant interruption. This will be facilitated by new devices, deeper operating system integration, and a persistent model running in the background.

The convergence of multi-modal models (like GPT-4o's vision and audio) with deep reasoning models will create a new kind of computer interface. Altman envisions a single model that can "think super hard, doing some research, writing a bunch of code just in time for a brand new app only for you to use or kind of rendering live video that you can interact with."

The Road to Embodied AI and Robotics

This advanced AI will soon connect to the physical world. OpenAI's strategy is to "nail that first" (the cognitive AI) and then connect it to robots. Altman is excited about a near future where subscribing to a top-tier ChatGPT plan might come with a free humanoid robot.

He speculates on the potential for rapid scaling: "If you make a million humanoid robots the old-fashioned way, can they run the entire supply chain?... and then maybe you actually can get a lot of robots in the world quickly." The demand for such automation, he believes, will vastly outstrip our current ability to produce it.

Advice for Builders: Defensibility and Conviction

Addressing a common fear, Altman states unequivocally, "We don't want to run you over." The biggest mistake founders can make is to build a "version of ChatGPT," a space where OpenAI has a strong headstart. Instead, the massive opportunity lies in the vast, uncharted territory of new applications.

He warns against the "social creature" trap of everyone pursuing the same five popular ideas. "The most enduring companies are usually not doing the same thing as everybody else." True defensibility is built over time by being first, then layering in unique features, network effects, and brand value.

On maintaining conviction, Altman is candid about the difficulty. He recalls a "really mean email" from Elon Musk early on, stating OpenAI had a "0% chance of success." Moments like these are crushing, but resilience is key. "You just like get knocked down and get back up and brush yourself off and try to keep going."

The Long-Term Vision: Intelligence and Energy

Looking ahead 10-20 years, Altman believes that "unless something goes hugely wrong, we'll have like unimaginable super intelligence." He is most excited about AI's potential to accelerate scientific discovery, which he sees as the primary driver of long-term economic growth and improvements in quality of life.

He also reflects on the intertwined future of AI and energy, a connection he didn't fully appreciate before starting OpenAI. He now constantly considers the physical limits of computation on Earth, joking about when we'll "have to put all the GPUs in space" to manage heat.

This vision stands in stark contrast to "de-growth" movements. Altman firmly believes in a future of radical abundance powered by technological progress, a core tenet of the startup and AI communities.

A Message to the Next Generation

Altman's final thoughts are for the entrepreneurs in the audience. He calls this "the best time ever in the history of technology ever, period, to start a company." The rapid pace of change favors those who can iterate quickly.

His advice mirrors the journey of his younger self: develop resilience, trust your instincts, have the courage to work on unfashionable problems, and understand that the highs of entrepreneurship are higher than imaginable and the lows are lower. The key is to keep going.