Public transcript

LIVE: OpenAI's Sam Altman speaks at G20 Innovation Ministerial

By: Reuters
Length: 31 min
Language: English
Text from: the video's captions
Transcribed: 13 October 2026

Read the full transcript Open on YouTube

Notes in English.

Summary

Sam Altman speaks live at the G20 Innovation Ministerial about his path from sci-fi-obsessed kid to OpenAI CEO, tracing the company from its 2015 founding through GPT-4 and the coming launch of a new model. He argues that AI adoption is 'non-negotiable' for countries, comparing it to electricity, and predicts the greatest boom in entrepreneurship the world has ever seen. He also flags cybersecurity, biosecurity, and power concentration as risks, and says AI must be made abundant and inexpensive to avoid worsening inequality. The session closes with a handoff to the next guest, Alex Karp.

Key points

Altman was a lifelong sci-fi and AI enthusiast who gave up on AI in school, then returned to it after deep learning started working around 2012.
OpenAI was announced at the very end of 2015, with the first day of work in January 2016, initially training a robot hand to solve a Rubik's cube.
Altman says the consensus moment when AI felt like 'bottling knowledge' came around the time of GPT-4 in March 2023.
He predicts AI will drive the greatest boom in entrepreneurship and small business creation the world has ever seen.
Altman argues refusing AI would be like refusing electricity 100+ years ago, calling adoption 'non-negotiable' for countries.
He describes OpenAI as 'pragmatic centrists' who avoid both blind optimism and doomerism, and says the company is 'on team humanity.'
He cites token usage growing from about 100,000 tokens a month per person to a projected 100 billion tokens a month per capita in another six and a half years.
Key risks he flags are cybersecurity, biosecurity, and too much concentration of power if AI is not made abundant.

Questions it answers

00:00Altman's origin story

How did Sam Altman go from being a sci-fi kid to leading OpenAI?

Altman was a lifelong AI and sci-fi nerd who studied it in school, gave up when nothing worked, turned to entrepreneurship, and returned after deep learning began working around 2012, starting OpenAI at the end of 2015.

Deep learning's 2012 breakthrough drew Altman back to AI.
OpenAI's first day of work was January 2016.

02:17Scaling laws and GPT milestones

What did early OpenAI build and when did it click that AI could capture knowledge?

Early OpenAI knew compute made systems smarter but little else, training robot hands on Rubik's cubes before GPT-1 around 2018 and GPT-3 in 2020; the consensus moment that AI was 'bottling knowledge' came around GPT-4 in March 2023.

GPT-3 in 2020 was the first model over a usability threshold.
GPT-4, held internally for about eight months, convinced skeptics.

05:46AI and entrepreneurship

Why does Altman think AI will supercharge entrepreneurship?

He predicts the greatest boom in entrepreneurship and small business creation ever, recalling a 2023 laundromat owner who ran his business with ChatGPT, and says work that took startups three months is now doable in 17 minutes with Codex.

Coding agents accelerated this in 2025.
Long-term persistent agents are coming next.

13:23OpenAI's identity

How does Altman differentiate OpenAI from other frontier companies?

He calls OpenAI 'pragmatic centrists' that avoid blind optimism and doomerism, want to be a reliable platform respecting sovereignty, don't want to be the only company, and are 'very proudly on team humanity.'

OpenAI sees itself as a platform serving the world.
People should stay at the center and run the future.

18:55AI as electricity

Why does Altman call AI adoption non-negotiable for countries?

He says rejecting AI would be like rejecting electricity 100+ years ago; countries can regulate differently and build or rent data centers, but adoption is 'non-negotiable' because the economic value is too high to ignore.

In a decade, people will expect AI everywhere without thinking about it.
Electricity took time to make safe, and so will AI.

21:51Risks ahead

What could hold AI back over the next five years?

He points to cybersecurity, which he says will go very wrong without urgent action, plus biosecurity and the risk of concentrating power if AI is not made abundant and inexpensive.

Failing to navigate risks could set the technology back greatly.
Abundance is needed to avoid worsening inequality.

24:05Token growth and abundance

How fast is token usage growing and what does it imply for infrastructure?

Per capita token use went from effectively zero to 100,000 tokens a month, with the top user going from 100,000 to hundreds of billions; projecting forward, per capita use reaches about 100 billion tokens a month, requiring vastly more infrastructure despite efficiency gains.

A multi-millionx increase in six and a half years.
AI should become as cheap as an hour of light at night.

28:34Optimism and safety

Is Altman fundamentally optimistic about AI serving the world?

Yes — he compares AI to fire and electricity, where worry and policy made technologies safe, and says things have gone better than doomers predicted five years ago; he cites someone who says they wouldn't be alive without ChatGPT.

Worry is how society anticipates and addresses problems.
He believes in human ingenuity to solve complex problems.

Notes

Sam Altman's path to OpenAI

Altman describes himself as an AI and sci-fi nerd from childhood who studied AI in school, gave up when nothing worked, and pursued entrepreneurship instead.
The 2012 watershed moment when deep learning started working and improved with more compute drew him back; OpenAI was announced at the end of 2015, with work starting January 2016.

Early days and scaling laws

Early OpenAI knew compute could make systems smarter but little else — before language models, projects included training a robot hand on a Rubik's cube.
Progress came with GPT-1 around 2018 and GPT-3 in 2020, the first model over a threshold of usability.
Scaling laws suggested a trillion dollars of compute would produce something amazing; Altman says the consensus 'bottling knowledge' moment came around GPT-4 in March 2023.

AI and entrepreneurship

Altman predicts the greatest boom in entrepreneurship and small business creation ever, citing a 2023 laundromat owner who ran his whole business with ChatGPT.
Work that took a startup three months in his accelerator era is now doable in about 17 minutes with Codex.

OpenAI's positioning

He calls OpenAI 'pragmatic centrists,' avoiding blind optimism and doomerism, wanting to be a reliable platform that respects national sovereignty.
The company is 'very firmly and very proudly on team humanity' and wants the whole global economy to succeed.

Adoption is non-negotiable

Rejecting AI would be like rejecting electricity 100+ years ago; countries can build or rent data centers and set their own rules, but must embrace the technology.

Risks and abundance

Cybersecurity and biosecurity are urgent challenges; too much concentration of power would be a bad outcome.
Token usage per capita went from effectively zero to 100,000 tokens a month, projected to reach 100 billion tokens a month in another six and a half years — requiring massive infrastructure to keep AI abundant and inexpensive.
Like fire and electricity, worry and policy are how society makes technology safe; Altman remains optimistic about human ingenuity.

Text from the video's own captions. Summary and notes written by AI from the transcript, so check anything important against the recording.