Openi Chief Executive Officer (CEO) Sam Altman On Wednesday, India is commonly called an incredible market for Artificial Intelligence (AI). He said that the Indian users of OpenaiI have tripled in the previous year. Altman, who is in India, spoke to HT’s editor-in-chief R Sukumar On the future plans of chat for India, Artificial General Intelligence and Diprach. Edited Excerpt:
I am sure you are looking at the announcements that India has done on its AI program. You were here some time ago and you made these comments – how India was not trying to do its own frontier model – it became controversial. Has your scene changed? And do you think the Indian AI scheme is on the right path?
It was in a different context. It was a different time when the frontier was super expensive to model. And you know, now, I think the world is a very different paradigm. I think you can do them at low cost and perhaps do incredible work. India is generally an incredible market for AI, also for us. This is our second largest market after America. The users here have tripled the last year. The innovation that is happening, what people are making (in India), it is really incredible. We are excited to do a lot here, and I think this (Indian AI program) is a great scheme. And India will build a great model.
What are your plans in India? Because there is, while everyone sees the end of AI, this is a very big end. For example, now what you are doing in America, in partnership with SoftBank, it is creating a huge infrastructure. Are you planning to bring that infrastructure in India?
We have nothing to announce today, but we are difficult to work, and we hope that something exciting to share soon.
In the late 2022, when you announced the chat, and over the weekend, you announced the Diprack. The pace of change seems quite shocking. Moore has a rule in microprocessors. Is there any law on the pace of change here?
Deepresearch is the thing that is most felt, such as chat, how are people reacting. I was watching and reading online last night – I was very busy for the last few days, so I didn’t find the reviews to read – and people look as if they have a magical experience, as if they have them Thi when the chat launched for the first time. So I think this step takes this step in agents from chatbots, it is the effect that we dreamed at night, and it is great to see that people have another moment. Moore’s law is, you know, every 18 months 2x (the processing power of chips is doubled every 18 months), and it changed the world. But if you look at the cost curve for AI, we are able to reduce the cost of the intelligence level given about 10x (ten times) every 12 months, which is incredibly more than the law of Moore Is powerful. If you compound both for more than a decade, it is completely different. However, although it is true that the best of the frontier model costs the cost of this steep, upward, at the exponential (curve), the rate of reduction in the cost of the unit of intelligence is just incredible. And I think the world has still not made it quite internal.
When the news of the Chinese model, there was news of deep search, what was your first response? At least the title was that they managed to train their models at a very low cost, although it later came out that it was not really so.
I had a lot of doubt from the cost number. It was so, some zero are missing. But, yes, this is a good model, and we will need to make a better model, which we will do.
AI seems to be a very infrastructure intensive and capital intensive. it is a matter. Does this mean that there are very few players who can actually work on that scale?
It is changing as we had spoken earlier. For me, the most exciting development of the previous year is that we came to know how to make very powerful small models. Therefore, the frontier will be expensive on a large scale and a large amount of infrastructure will be required, and that is why we are doing this Stargate project. But, you know, we will also get GPT 4 level models on the phone at some point. So I think you can see it in any direction.
Where you are, there is one of the challenges of being there, and who you are, your company was the first company that captured a lot of public imagination for artificial intelligence. When you are the first company, it is your responsibility, not only for your company, but also for the industry and how the whole industry intervenes with society. And there, there are many issues that are cropped…
We have a role as I think, if you are on the border … there is a role as a teacher, and the role is like a search to tell the society what you think and what you think Looks like what the effect is going to happen; This will not always be right, but it is not dependent on us or any other company, to say, okay, given this change, what the society should do here.
This is for us to say this, the changes we are seeing here are some ideas here, here are our recommendations. But the society has to decide how we think how we are going to reduce economic impact, how we are going to distribute wide benefits, we are going to solve the challenges coming with it. So we are a voice, an important voice, in it. And I do not even mean to say that we do not have the responsibility for the technology we created. Of course we do, but this is a conversation between all stakeholders.
If you look at the Indian IT industry, they have really performed well in taking goods that other people have built and built very smart models over it, and provide services above it instead of making models. Do you think what they should do with AI? Or do you think, should they do more?
I think India should go for a full stack approach …
… which will require a lot of capital.
Well, this is not a cheap project, but I think it is worth it.
You have more than 300 million users …
More…
… Okay, and what have you learned what they are using for chat GPT?
Can I show you something? Because this is really a meaningful thing. I was just looking at X (changing the computer to show the screen). So this man, we are not really friends, but I know him a little. Deep Research launched a few days ago, and her daughter has a very rare form of cancer, and she has stopped her job, I think, or maybe changed their job, and super hard work Is; They have put a large private research team (to understand his illness) together. They have raised all this money, and intensive research is giving them better answers than the private research team. And it is really meaningful for us to see that kind of items.
Do you hope that President Trump will take more steps to protect the US leadership in AI? Do you see that happening? Or, to do the question differently, is there a national game to be played in AI?
Of course there is there. But our mission, which we take the super seriously, is for AGI (Artificial general intelligence) To benefit everyone of humanity. I think it is one of these rare things that transfer national borders. AI wheel and fire, industrial revolution, agricultural revolution, and it is not a matter of a country. This belongs to everyone. I think AI is one of these things. It is like the next step in it. And they are not related to nations.
You first talked about Artificial General Intelligence a few years ago. Have we gone close to him?
Yes, when I wonder what the models are able to do now, what they can do a few years ago. I think we are undeniably close …
Are we more courageous with our failures now?
Where we have gone a few years ago … I think two years ago we have made so much progress in model security and strength. You know, look at the Elucidation rate of a current model, or the ability to follow a set of policies, we are in better size than two years ago. This does not mean that we do not have to solve the superintendent (an theoretical creation of AI or more than far away). Of course we do, but we are on a good trajectory.
Have you seen the lensate paper on the study of Swedish breast cancer that came yesterday? He used an AI model called Transpara, which I do not know if you are familiar with it, and found that accurate diagnosis increased by 29%, with no wrong positive …
It is fantastic. I was thinking the second day, you know, how much better it is to drive AI? How much better should be as a diagnter than a human doctor before diagnosing AI? It has become clearly better; Self-driving cars must be more secure than human drivers to accept them. But, how many of those studies and us before we want an AI doctor?
Although I think when it comes to diagnosis, there will be very little for cars…
I think for cars, perhaps thematic, you want it to be 100 times safe. For a diagnosis, it must be Very little.
