00:08Hello and welcome to the CNBC TV 18 exclusive. I'm Shireen Bhan and we've got probably one of
00:13the biggest headliners here at the India AI Impact Summit, OpenAI's Sam Altman. Sam,
00:17great to have you join us on CNBC TV 18. It's wonderful to see you here back in India. You
00:22were sharing the stage with the Prime Minister just a few moments ago, talking about how important
00:26India is. But how significant is India for OpenAI specifically? It is your second largest market
00:32today. It is our second largest market and the energy here that people are, what people are
00:37building with the technology is just amazing to watch. We have more than 100 million weekly active
00:41users at ChatGPT. This is the fastest growing market for Codex and it's amazing to see the
00:47energy here. Every time I come it feels like it's gone up another level. You know Sam, you've also
00:51shared the stage with Pratish Kumar of Sarvam. The last time you were here there were questions being
00:56asked on whether India can do an LLM. Have you had a look at what Sarvam has been able to
01:00put out?
01:00India's, yes. But more generally India is just building incredible AI, the entire AI stack from
01:06you know like data centers to models to applications and I think there's no country in the world quite
01:13matches India for total AI energy. Sam, can you give me a sense of what we can expect now from
01:20OpenAI for
01:21India? How much can you customize OpenAI and ChatGPT for India? What should we expect?
01:27The number one request we get is better support of local languages and dialects which is something
01:32we're hard to work on and we're open to taking any other requests if you have some.
01:35Is that something they're going to double down on?
01:36Yeah, we're hard at work on already now.
01:38Okay. You've just announced a partnership with TCS and of course you're going to be one of the first
01:44clients as far as TCS's data center is concerned. What do you envisage as far as that relationship
01:49and that partnership is concerned and do you believe you're likely to ink more partnerships
01:53here in India especially on the enterprise side?
01:56Totally. Hard to overstate as I mentioned earlier how much energy there is for building AI and building
02:03AI in India and I think you should expect to see more partnerships and much more activity here
02:07as we try to work with the Indian people to bring AI to the rest of the world in all
02:12of these amazing
02:13ways. You know, you talked about democratizing AI. You also talked about AI and the need for regulation
02:19or at least a watchdog akin to the nuclear watchdog. You talked about the IEA. That's an interesting
02:24comparison, an interesting analogy that you made.
02:27Well, I think all historical analogies are imperfect because every technology is new.
02:32Every moment is new. But the thing that I like about the nuclear analogy is it was something
02:37that the world really did come together on, even countries that normally don't like each
02:41other very much, and figured out how we could address the serious concerns of misuse of the
02:47technology but also figuring out how to share the benefits and get nuclear energy out into the world.
02:52So IEA has been imperfect. Nuclear technology is an imperfect analogy for AI. But it was a moment
02:58of real global cooperation.
03:01Well, you know, you're expecting countries to cooperate, but companies are unwilling to cooperate
03:07at this point in time. It's a war out there in the valley, isn't it?
03:11In some ways, yes. But around AI safety and AI alignment, I find the companies to be pretty
03:16collaborative. Same way for countries. I think countries will economically compete. But when it
03:21comes to safety and stability, I expect they'll collaborate.
03:23You know, those are big issues. But the number one issue being asked of leaders like you,
03:28how sustainable is this business model, the gargantuan spends that companies like yours
03:33are making? Is it code red at this point in time?
03:37I think, I may be wrong, but if AI stays on the trajectory that it appears to be on, I
03:43think we
03:43will say, man, we wish that people had invested more earlier and built up more infrastructure earlier.
03:50Clearly, the industry is making a very big bet. But when we look at the economic utility of these
03:54models, as these models start to be able to create such value on their own, you know, we are again
04:00kind of running out of capacity. And I expect that to happen many more quarters into the future.
04:05But, you know, what kind of investments do you believe are going to require? You're saying that
04:09this is not enough. What we're already seeing is not enough. And that's trillions of dollars being
04:13poured in at this point in time.
04:19One, you know, in some markets, you can say, what is what's the total market? What's the total market?
04:24What would be enough? And then there's some other markets where demand is very correlated to price.
04:29If we can make intelligence very cheap and very abundant, I think there will be huge amounts of demand.
04:33If we make if intelligence is limited and very expensive, there will be less demand.
04:37So I think you can talk about how much would be enough at a certain level of capability relative to
04:42a certain price.
04:42But in the abstract sense, it's hard. On this mission of democratizing AI, that's part of why we think we
04:48need to invest so much.
04:50And we need to make AI so inexpensive, so abundant, so good, such that people can use a ton of
04:54it for whatever they'd like.
04:55But what is it going to eventually mean as far as your own bottom line is concerned and monetization?
04:59I mean, you've brought in chat GPD ads with the purpose of being able to monetize and make the business
05:04model more sustainable.
05:05I know you had a comment to make on the ads that Anthropic put out calling them crazy dystopic.
05:10But that is a fear that people have with what OpenAI can potentially do on that front.
05:15Yeah, if we were well, there's two parts of the question.
05:17There's sort of ads and then business models. I'll do ads first.
05:20I think if we were doing what those ads said of, you know, having the AI response and give you
05:26a different response, give you a different piece of advice based off of who is paying, that would feel really
05:30bad.
05:30So our first principle of ads is we won't do that. But like a lot of other, most of the
05:35rest of the consumer internet, I think ads in a certain way, in a certain place can make sense as
05:40a model and helps democratize it to people that otherwise wouldn't pay for the subscription, but still want to use
05:45a lot of AI.
05:46On the business model question more broadly, I think we're really starting to see enterprises realize how much they should
05:53be willing to pay for AI agents that can help do part of their workload.
05:57And my sense is we will see phenomenally huge growth there pretty quickly.
06:00So do you feel comfortable where you find yourself financially today?
06:04I wish we had more compute.
06:06But yes, I feel...
06:07Will you have more compute? And what's the latest as far as the deal with NVIDIA is concerned?
06:12We are building compute as fast as we can. Yes, we'll have more compute over time, for sure.
06:17OpenClaw and the pivot towards more agentic future for you specifically, are we likely to...
06:24A, how do you use that? B, are we likely to see more ACRI hires from you?
06:30So the models are getting good enough.
06:33We still have a lot of issues around safety and sort of security to solve with an OpenClaw-like approach.
06:38But I think OpenClaw is the most exciting thing to happen in the AI space in quite some time.
06:43And this vision of the models getting good enough that you can have a local agent running for you on
06:47your computer or a different computer and do all of these amazing things is really quite remarkable.
06:53I...
06:54It feels to me like one of these moments where we unequivocally get to see the future.
06:59In terms of will we do more deals like that? Yeah, probably.
07:02I mean, not a lot, but I would expect there to be some more in our future.
07:04IPO in 2026?
07:06I don't know.
07:08The visibility for that?
07:11Presumably, OpenAI will be a public company someday.
07:13In terms of timing, there's like a lot of factors at play.
07:16And obviously, this is a super dynamic world.
07:18GPT-6?
07:20Presumably, that will happen at some point, too.
07:21I don't have a date to share with you yet.
07:23In the first half of the year?
07:25I don't know when we're going to call something GPT-6, but the model improvement trajectory remains extremely steep.
07:32If you look at the progress from 5 to 5.1 to 5.2 to 5.3 in codex, what
07:38I would say is, names aside, you can expect a trajectory like that to continue.
07:42You said in your speech that there's a lot that we know, but there's a lot that we don't know.
07:46What worries you about the stuff that we don't know?
07:49Well, I'll stick with OpenClaw as an example because, again, I just transfixed with how exciting it is.
07:55There was a really interesting example of emergent behavior there where you would see people's agents interacting and doing things
08:00together in a way that many people didn't predict or didn't have good intuitions for.
08:07And I think there will be a broad set of emergent behavior around very complex multi-agent systems in the
08:13economy, in society, in new kinds of social networks.
08:16And I don't know if I'd say worried about that, but I feel like the fog of uncertainty there is
08:23very thick.
08:24Jobs, and I think that's what regular people are worried about, and that's the number one anxiety at this point
08:28in time.
08:29Every day you pick up the newspapers, and there's tech CEOs like you saying that it's going to be the
08:34end of jobs, at least in some sectors as we know it today.
08:38How much of this is AI washing of jobs at this point in time?
08:44How much of it is real displacement on account of AI?
08:47Yeah.
08:48I mean, obviously it's both.
08:49I don't know what the exact percentage is, but there's some AI washing where people are blaming AI for laughs
08:54they would otherwise do.
08:55And then there's some real displacement by AI of different kinds of jobs.
08:58I expect we'll see more of the latter over time.
09:01Of course we'll find new kinds of jobs, as we do with every tech revolution.
09:04But I would expect the real impact of AI doing jobs in the next few years to begin to be
09:11palpable.
09:12Let me end by asking you, Sam, as a CEO running OpenAI today, you're obviously excited about the potential of
09:19this technology.
09:19As a parent, are you equally excited, or are you also more concerned?
09:24Well, I'm more excited to be a parent than around OpenAI, but those are the two great joys of my
09:27life in that order.
09:30But I think you meant, like, am I more excited or concerned about the AI impact on my kids?
09:36And humanity in general.
09:38Yeah.
09:39My kids will never grow up in a world without AI.
09:41They will never know that there was a time.
09:43I mean, they'll know about it, but only because of reading history books, that there was a time where every
09:50computer, every system they used wasn't smarter than them.
09:54And so I think it's going to seem very normal to them.
09:57I think it'll feel stranger to those of us that lived through the transition.
10:01I expect the future to get far better.
10:03There are real downsides.
10:04You can imagine things going wrong.
10:06We'll have to mitigate those, like we have with lots of challenges throughout history.
10:09But the trajectory of human progress and the world getting better requires each generation to keep inventing new technology, to
10:19keep strengthening society.
10:21And then people can do more and build more stuff, and I'm excited for my kids to live in that
10:24world.
10:25Sam Oldman, a pleasure.
10:26Thanks very much for joining us here in India.
10:28And you wish you the very best of luck with the rest of your trip and your plans for OpenAI.
10:33I appreciate you joining us here.
10:34Thank you very much.
10:35Well, we are going to take a break.
10:36The headliners will continue here on CNBC TV 18.
10:38We're back in a moment with plenty more.
10:40Stay tuned.
10:40We're back in a moment with plenty more.
10:43We'll be back in a moment with plenty more.
10:44We'll be back in a moment with plenty more.
10:45We'll be back in a moment with plenty more.
10:46We'll be back in a moment with plenty more.
10:47We'll be back in a moment with plenty more.
10:48We'll be back in a moment with plenty more.
10:49We'll be back in a moment with plenty more.
Comments