125x

Part of Founder interviews · All guides

Sam Altman on GPT-4 and shipping AI: interview summary

Sam Altman: OpenAI CEO on GPT-4, ChatGPT, and the Future of AI | Lex Fridman Podcast #367

Lex Fridman · Mar 25, 2023 · about 144 min

Uploaded Mar 25, 2023 · 6.8M views (as of Oct 30, 2026)

Some value given

OpenAI ships in steps so the public can find strengths and weaknesses quickly.
He wants the degree of alignment to grow faster than the model's capability.
He says a little fear about AI is appropriate, and he's open about it.
He expects many AGIs in the world, with OpenAI contributing one.
On advice, he warns it's too tempting to copy what worked for others.

Why does OpenAI release AI to the public step by step?

Altman says putting models out early and often lets millions of people find both the great uses and the real weaknesses far faster than any internal team could. That gives OpenAI time to fix problems, and gives society time to adjust and give feedback. He says a little fear about AI is appropriate.

Worth watching?Some value given. A thoughtful look at how OpenAI shipped GPT-4 and thinks about risk. It's long and mostly about AI, with a short, honest bit of career advice near the end.

ActionableLessons

Actionable

Ship a small version of your project to real users this week.
Ask users what's broken, then fix the top issue quickly.
Write down which advice you follow only because it worked for someone else.
Give one person on your team more autonomy and higher standards at once.

Lessons

Real-world feedback beats long internal testing.
Being honest about risks builds trust.
Truth seeking beats elegance when the goal is results.
High trust plus high standards lets teams move fast.

Quick answer

Sam Altman sat down with Lex Fridman for this interview, published Mar 25, 2023. Here's our summary in our own words: a verdict, the answer to the big question, 5 takeaways, Actionable and Lessons tabs, then the stories, frameworks and how to use them.

What 125x is: 125x is a web app at 125x.app. You connect your YouTube subscriptions, and it gives you a feed of the latest videos from the channels you follow, each with a summary: a verdict, the 5 key takeaways, and the spoiler, plus Actionable and Lessons tabs with more tips. You read that and decide what's actually worth opening.

Open 125x

Watch the interview

Watch on YouTube: Sam Altman: OpenAI CEO on GPT-4, ChatGPT, and the Future of AI | Lex Fridman Podcast #367

Channel: Lex Fridman. Published Mar 25, 2023. About 144 minutes.

Who is Sam Altman?

Sam Altman is CEO of OpenAI and was president of Y Combinator.

This episode was recorded soon after GPT-4 came out in March 2023.

Key stories from the interview

He explains how GPT-4 went through a lot of testing before release.
Lex reads Altman's 'How to Be Successful' list back to him, and Altman says to be careful copying advice.
He talks about how OpenAI handles pressure from big rivals by sticking to its mission.
He mentions switching from Emacs to VS Code, partly to use Copilot.

Frameworks and ideas

Iterative deployment: release, learn, improve, repeat.
Alignment faster than capability.
High trust, high autonomy, high standards.

In their words (paraphrased)

Putting things out and improving them quickly is really important.
He thinks it's weird when people treat his being a little afraid as a gotcha.
It's way too tempting to take advice from other people.

How to apply this

Set a weekly release cycle for one project.
Write your own success list based on what has worked for you.
Share one risk about your product openly with your users.

A note on this summary

This page reports what was said in the interview, in our own words. Lines under 'In their words' are paraphrased from the transcript, so check the video for exact wording.

How 125x does this for your feed

This page is one sample of what 125x does for every video in your feed. 125x pulls in your YouTube subscriptions automatically, so there's nothing to set up. Each new video gets a watch, skim or skip verdict, 5 takeaways, the answer to the clickbait title, and Actionable and Lessons tabs.

View your YouTube feed at 125x speed. YouTube summaries of every video in your feed.

FAQ: questions people ask

Why does OpenAI release models early?

Altman says public use reveals strengths and problems faster, and gives people time to adjust and give feedback.

Is Sam Altman afraid of AI?

He says a little fear is appropriate and he's open about being both excited and somewhat scared.

What advice does Sam Altman give?

He warns against copying other people's advice too closely, since what worked for him may not work for you.