Connect your Claude, Codex or OpenClaw for full transcripts on millions of podcasts. Two week free trial, cancel anytime.

Start free trial →
SA Sam Altman Sam Altman Entrepreneur and investor who has served as CEO of OpenAI since 2019, leading the company behind ChatGPT and its research on artificial general intelligence.

“It's certainly incorrect to talk about this only as an engineering problem. Anyone who says we have solved the science of alignment, I believe is wrong in a very dangerous way. We need to make more research progress. … If we are going to create models that are much much smarter than all of us, we have to actually solve the science of alignment.”

TBPN · Startups & Venture · September 2026

“It's certainly incorrect to talk about this only as an engineering problem. Anyone who says we have solved the science of alignment, I believe is wrong in a very dangerous way. We need to make more research progress. … If we are going to create models that are much much smarter than all of us, we have to actually solve the science of alignment.” — Sam Altman, TBPN

Context

Near the end of Altman's live TBPN interview at OpenAI DevDay, a host asked whether alignment is an engineering problem or a science problem. Altman says it is both and that the science is not solved, a notable admission on the day OpenAI launched always-on personal agents.

Listen to the episode → More from Sam Altman → More Startups & Venture quotes →

Transcript

TBPN Around 19:17 into the episode
Sam Altman

…build 30 things. You can now build all 30 things Mhmm. And you can get people to use them all. And then you can work more on the ones that work. So I think this mindset of lots of ideas, quick feedback, and be open to a lot of things not working is great for this new era of how how people build.

Jordi Hays

I have one more question that we wanna get you to hit the gong. On alignment, you've been deep in the weeds, I'm sure. There was a debate I saw, is it an engineering problem? Is it a science problem? Is there a different frame of mind that we should be approaching alignment with?

Sam Altman

I I think it's like all of those things but it's certainly incorrect to talk about this only as an engineering problem. Okay. Anyone who says we have solved the science of alignment, I believe is wrong Okay. In a very dangerous way. Like we need to make more research progress. I assume we will. We've been great at making this research progress. And of course there's a lot of engineering work to do too. Know, we need to continue to figure out how to build better sandboxes and better monitoring tools. But eventually, we If we are going to create models that are like much much smarter than all of us, we have to actually solve the science of alignment.

Jordi Hays

Have we considered moving away from the term sandbox? Because I have a five year old, if I put him in a sandbox, he's not gonna stay there. I think we might need a new term.

Sam Altman

I can't unfortunately just stay in

Jordi Hays

the sand. Oh, okay. Okay. So maybe yeah. But but, you know

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.

More from TBPN

“It really should never have shipped. The build quality is really bad… It's like very dangerous. I'm constantly kind of on edge when I'm around it, and trying to tell other people around the car, 'hey, that's effectively a knife's edge, keep your fingers away from here.' And again, you can see issues with the car from like 20 feet away.” — Jordi Hays, TBPN “The really successful venture capital firms of today were in the right place through the dot-com crash, like Google, Amazon. So even if you wind up in a total reset of the market, you would definitely expect that there's a couple new effectively hyperscalers that make it through… From start to finish, .com produced something like 151x. And at the last close, this AI peak is at 16x from the baseline.” — John Coogan, TBPN