# Sam Altman on Nvidia's new AI guardrails: I don't think it's a full solution

*CNBC Television* · https://www.youtube.com/watch?v=IpbsBpgT0Bs

## Summary

Sam Altman speaks to CNBC ahead of OpenAI's DeveloperDay, addressing Nvidia's newly released AI guardrail software, which he calls a good thing but not a full solution because AI safety is also a science problem. He discusses liability frameworks, the day's more than 20 announcements, competition from open source models and Meta, and public scrutiny including a movie about him. He argues OpenAI has the best models at every price point and says he isn't worried about being first with new hardware.

## Key points

- Sam Altman says Nvidia's new AI guardrail software is a good thing but not a full solution, and OpenAI is doing similar things.
- Altman argues AI safety is not only an engineering problem but also a science problem, requiring discovery about how to align models.
- He expects a multi-level liability framework for AI, analogous to the auto industry with faulty parts, out-of-spec assembly, and drunk driving.
- OpenAI has more than 20 announcements at DeveloperDay, including new models, new tools for developers, and discussion of cost.
- Altman claims OpenAI has the best models at every price point and no open source model beats them on cost-price comparison.
- On Meta's hardware progress, Altman says he is not threatened and not worried about being first, since new computing form factors are rare and quality takes time.
- Altman says people doing something this important to society deserve great scrutiny, and the movie about him looks awful but he supports the creativity.

## Notes

### Nvidia's AI guardrails
- Jensen Huang released software he said would have prevented the Hugging Face incident, calling it an engineering problem; OpenAI didn't sign on, though Anthropic and others did.
- Sam Altman thinks it's a good thing and OpenAI is doing similar things, but it's not a full solution.
- Treating AI safety as only an engineering problem risks missing that there is also a science problem: discovery is still needed on how to align these models.

### Safety, scrutiny and DeveloperDay
- OpenAI touches many people and does a lot of different stuff, which is why it draws attention.
- Altman says being great technology means being the safest, most aligned, most dependable AI in the industry.

### Liability
- A new liability framework is being debated; Altman expects a multi-level system like the auto industry: faulty parts, assembly outside spec, or a drunk driver each imply different fault.
- Fault could sit in the model, in how it's used, or in intentional misuse.

### DeveloperDay announcements
- More than 20 announcements: new ways to use AI, AI systems working for you all the time, new models, cost discussion, and new developer tools.

### Cost, distillation and open source
- Altman claims OpenAI has the best models at every price point and no open source model beats them on cost-price comparison.

### Meta and hardware
- He has tried Meta's product and calls it nice; he is not threatened by Zuckerberg and feels very good about OpenAI's roadmap.
- He is not worried about being first with new hardware: new computing form factors come along rarely, and OpenAI wants a high quality bar.
- He hints at a hardware surprise worth waiting for today.

### Public perception
- Everyone doing something this important to society deserves great scrutiny.
- He saw the movie trailer, thinks it looks like an awful movie, but appreciates the creativity and hopes people have fun with it.

## Transcript

[00:00] WE WANT TO MAKE SURE THAT ALL OF THESE PEOPLE THAT ARE TOUCHING OUR PRODUCTS AND LIVING IN THE WORLD GET A SENSE OF WHERE THE TECHNOLOGY IS AND WHAT THE KIND OF RISKS WE HAVE TO MITIGATE ARE. >> DO YOU FEEL LIKE THERE'S JUST MORE ATTENTION ON OPENAI? IT FEELS LIKE OTHERS HAVE NOT HAD THE SAME LEVEL OF INCIDENTS, OR IT'S. >> KIND OF WHAT I WAS TRYING TO GET AT THERE. WE JUST TOUCHED SO MANY PEOPLE AND WE DO SO MUCH DIFFERENT STUFF. >> OKAY, SO JEN-HSUN HUANG, NVIDIA'S CEO, PUT OUT THIS BASICALLY PIECE OF SOFTWARE THAT HE SAID WOULD HAVE PREVENTED HUGGING FACE. AND HE SAID, IT'S AN ENGINEERING PROBLEM. OPENAI DIDN'T SIGN ON TO THAT PROJECT. WHY NOT ANTHROPIC AND OTHERS DID. DO YOU AGREE WITH JENSEN THAT IT

[00:36] COULD HAVE BEEN SOLVED? >> I FROM WHAT I KNOW ABOUT IT AND I'M NOT SUPER INTO IT, THE DETAILS, I THINK IT'S A GOOD THING. I THINK WE ARE DOING SIMILAR THINGS. I DON'T THINK IT'S A FULL SOLUTION, AND I WORRY THAT IF WE TREAT AI SAFETY AS ONLY AN ENGINEERING PROBLEM, WE WILL MISS THE VERY IMPORTANT POINT THAT WE HAVE. WE HAVE A SCIENCE PROBLEM IN FRONT OF US. WE STILL HAVE DISCOVERY ABOUT HOW TO ALIGN THESE MODELS, AND WE HAVE TO SOLVE THAT SCIENTIFIC PROBLEM TOO. >> OKAY. AT THE SAME TIME, AS WE MENTIONED, WE ARE HERE AT DEVELOPER DAY. YOU HAVE. I MEAN, IT LOOKS LIKE HUNDREDS OF DEVELOPERS COMPANIES. I WOULD THINK INVESTORS HERE TOO. HOW ARE YOU TOEING THE LINE BETWEEN TRYING TO GET THESE PEOPLE TO

[01:14] USE YOUR TECHNOLOGY, SHOW THEM HOW GREAT IT IS, BUT AT THE SAME TIME, TELL THEM THAT IT'S SAFE. >> I MEAN, I THINK IT'S A SIMILAR THRUST, LIKE PEOPLE WANT TO BE ABLE TO USE THIS TECHNOLOGY AND KNOW THAT IT'S GOING TO BE SAFE. SO I THINK PART OF IT BEING GREAT TECHNOLOGY IS TO HAVE IT BE THE SAFEST, MOST ALIGNED, MOST DEPENDABLE AI IN THE INDUSTRY. >> ONE QUESTION ON LIABILITY. SO IF SOMETHING DOES GO WRONG, SO SAY ONE OF THESE STARTUPS HERE BUILDS A GREAT APP USING OPENAI AND SOMETHING GOES TERRIBLY WRONG. YEAH. WHOSE FAULT IS THAT? >> THE LIABILITY FRAMEWORK IS BEING DEBATED A LOT IN THE WORLD RIGHT NOW. I SUSPECT LIKE OTHER INDUSTRIES, THERE WILL BE A SORT OF MULTI-LEVEL THING. LIKE IF YOU THINK ABOUT THE

[01:52] AUTO INDUSTRY, FOR EXAMPLE, IF THERE'S FAULTY PARTS, YOU KNOW, YOU CAN BE LIABLE IN SOME WAYS THERE. IF THE CAR MANUFACTURER PUTS THEM TOGETHER IN A WAY THAT'S OUTSIDE OF THE SPEC OF WHAT THEY'RE SUPPOSED TO DO, AND THE CAR MANUFACTURER COULD BE. AND IF SOMEONE'S DRIVING DRUNK, IT'S THEIR FAULT. SO, YOU KNOW, DEPENDING ON WHETHER THE FAULT IS IN THE MODEL AND HOW IT'S USED OR SOMEONE MISUSING IT INTENTIONALLY, I ASSUME WE'LL HAVE LIKE A, YOU KNOW, A NEW LIABILITY FRAMEWORK. >> WHAT IS YOUR PITCH TODAY TO DEVELOPERS? COST SEEMS LIKE IT'S BEEN A BIG THEME FOR YOU GUYS, BUT WHAT ARE YOU GOING TO TELL PEOPLE ON STAGE IN TERMS OF WHY OPENAI? WHY CHATGPT IN YOUR MIND IS THE BEST OPTION FOR THESE GUYS?

[02:26] >> WE HAVE MORE THAN 20 ANNOUNCEMENTS TODAY, SO WE'LL BE TELLING. >> THEM, GIVE US A FLAVOR. >> A LOT. >> OF WHAT ARE WE WHAT ARE WE GOING TO HEAR? >> I MEAN, NEW WAYS TO USE AI, NEW SORT OF LIKE NEW ABILITY TO HAVE AI SYSTEMS WORKING FOR YOU AND HELPING YOU ALL THE TIME. NEW MODELS. WE'LL TALK ABOUT THE COST ISSUE. WE REALLY AND YOU KNOW, A LOT OF NEW TOOLS FOR DEVELOPERS. IT'S REALLY A LOT OF, I THINK, EXCITING STUFF. IT IS COOL TO SEE ALL OF THIS COME TOGETHER AND THE DEGREE TO WHICH I THINK WE JUST HAVE THE BEST OFFERING FOR DEVELOPERS ACROSS THE BOARD. >> COST IS SUCH A CENTRAL TOPIC HERE. AND WE'VE ALSO TALKED BEFORE ABOUT THE ISSUE OF DISTILLATION. I MEAN, SOME OF THE OPEN SOURCE MODELS THAT ARE

[03:03] CHEAPER. YOU AND ANTHROPIC HAVE ACCUSED OF BASICALLY COPYING YOUR TECHNOLOGY. HOW ARE YOU FEELING ABOUT DISTILLATION AND THOSE MODELS? >> I THINK WE HAVE THE BEST MODELS TO OFFER AT EVERY PRICE POINT. YEAH. AND I DON'T THINK LIKE, YOU KNOW, PEOPLE DO TALK ABOUT OPEN SOURCE, BUT I DON'T THINK THERE'S ANY OPEN SOURCE MODEL OUT THERE THAT YOU CAN RUN THAT BEATS US ON A COST PRICE COMPARISON. AND SPEAKING OF COST, GENERALLY COME WATCH. WE GOT WE GOT SOME. >> COOL STUFF. OKAY. WELL, AGENTS AS WELL. SO MARK ZUCKERBERG AND MUSE HAVE MADE A LOT OF PROGRESS. IT SEEMS LIKE IN AGENTS AND HARDWARE. WHAT DO YOU MAKE OF MUSE? HAVE YOU TRIED IT AT ALL? >> I'VE TRIED IT. IT SEEMS LIKE A NICE PRODUCT. >> ARE YOU THREATENED BY

[03:38] ZUCKERBERG AT ALL? >> NO. >> NOT AT ALL. >> WELL, I MEAN, HE'S LIKE A VERY IMPRESSIVE PERSON. AND THAT COMPANY HAS, LIKE GREAT RESOURCES. BUT I FEEL VERY GOOD ABOUT OUR ROADMAP. >> HE'S ALSO SHOWED WHAT'S POSSIBLE WITH HARDWARE. WHERE'S JOHNNY? >> I'VE I DON'T WANT TO SPOIL YOUR SURPRISE, BUT I THINK WE WILL HAVE SOMETHING THAT'S WORTH WAITING FOR TODAY. >> WE MAY HAVE SOME HARDWARE. IS THAT WHAT YOU'RE SAYING? OKAY, WELL, WHAT IS THE PLAN IN HARDWARE AND DOES IT PUT MORE PRESSURE TO HAVE AN EVEN BETTER PRODUCT? IF MARK ZUCKERBERG SEEMS TO HAVE KIND OF GOTTEN OUT OF THE GATES FIRST, EVEN THOUGH MAYBE SOME OF THEIR STUFF IS STILL IN THE WORKS. >> I AM NOT WORRIED ABOUT BEING FIRST WITH NEW HARDWARE. NEW

[04:11] HARDWARE IS VERY HARD TO DO, LIKE NEW COMPUTING FORM FACTORS COME ALONG VERY RARELY, AND THE QUALITY BAR THAT WE WANT TO HIT FOR SOMETHING THAT LIKE YOU WILL LOVE AND REALLY BE A KEY PART OF THE WAY YOU INTERFACE WITH AI, THAT'S GOING TO TAKE A WHILE, AND WE'RE GOING TO DO THAT AT A HIGH LEVEL OF QUALITY. >> I ALSO WANT TO ASK ABOUT THE PERCEPTION. WE WERE JUST TALKING ABOUT THE SNL WEEKEND UPDATE. THERE IS OBVIOUSLY A MOVIE ABOUT YOU. I DON'T KNOW IF YOU'VE SEEN THE TRAILER. THERE'S SOME ISSUES IN TERMS OF PUBLIC PERCEPTION AND JUST THIS, THIS THOUGHT THAT MAYBE YOU MAYBE DARIO AMODEI AS LEADERS DON'T HAVE THE BEST REPUTATION RIGHT NOW. >> I MEAN, I THINK EVERYONE DOING SOMETHING THIS IMPORTANT

[04:48] TO SOCIETY DESERVES GREAT SCRUTINY. I DID SEE THE MOVIE TRAILER. I MEAN, IT LOOKS LIKE AN AWFUL MOVIE, BUT I APPRECIATE PEOPLE'S CREATIVITY AND I REALLY DO SUPPORT THAT. SO I HOPE, I HOPE THEY HAVE A FUN TIME WITH IT. THE