Openai’s CEO (CEO) Sam Altman was generally called India in an incredible market for artificial intelligence (AI) on Wednesday. He said that Openai Indian users have tripled last year. Altman in India talked to R Sukumar, the editor -in -chief of HT, about Chatgpt’s future plans, artificial general information, and Deepresearch. Edited excerpts:

I am convinced that India has seen the announcements on the AI program. You were here someday and made these comments -about how India should not try to do his own frontier model -there was room for debate. Has your view changed? And do you think the Indian AI plan is on the right track?
It was in another context. It was a time when the frontier model was very expensive. And, as you know, I think the world is a very different paradigm. I think you can do them at lower costs and probably do incredible work. India is an incredible market for us, in general. This is the second largest market after the United States. The user here has tripled last year. It’s really incredible that the innovation and people are built (in India). I’m looking forward to doing more things here. I think (Indian AI program) is a great plan. And India builds a wonderful model.
What is your plan in India? When everyone looks at the AI front end, there is this huge back end. For example, what you are doing in the United States in cooperation with SoftBank is to create this huge infrastructure. Are you planning to bring some of the infrastructure to India?
I haven’t announced anything today, but we are working hard. I hope that there is something exciting to share immediately.
In the latter half of 2022, when Chatgpt was announced, Deepresearch was announced on the weekend. The pace of change seems to be very amazing. Micro processor has Moore law. Is there a law on the pace of change here?
Deepresearch is the most felt like Chatgpt in how people are reacting. I was reading online last night and reading -I have been very busy for the last few days, so I couldn’t read the reviews -and people had magic experience as they had. It looks like it’s done when Chatgpt starts first. Therefore, I think this movement from chatbot to agent has the influence we had dreamed at night. And it’s very cool to see people have such a moment. Moore’s law twice the world (twice the processing capacity of chips every 18 months), which has changed the world. However, looking at the cost curve of AI can reduce the cost of a specific level of about 10 times every 12 months (10 times). This is incredible than Moore’s law. It’s completely different if both are complicated for over 10 years. Therefore, it is true that the highest cost of the frontier model lies in this steep, up -faced index function (curve), but the cost reduction rate of intelligence units is incredible. And I don’t think the world has completely completely inner inside.
What was your first response when the news of Deep Shik, a Chinese model, came out? At least the headline was that they were able to train models at a much lower cost, but it turned out to be in fact.
I was very skeptical of the cost. There are some zero that is lacking. But yeah, it’s a good model and needs to create a better model.
AI seems to be very infrastructure -consolidated and integrated in capital. In that case. Does that mean that there are very few players that can actually operate on that scale?
As I mentioned earlier, it is changing. For me, the most exciting development of last year is that I have found a way to create a very powerful small model. Therefore, the frontier is still very expensive and requires a huge amount of infrastructure. Therefore, we are conducting this Stargate project. However, as you know, you will also get a GPT 4 -level model running on the phone at some point. So I think you can see both directions.
One of the places you are and one of the challenges of who you are is the first company that your company almost captured public imagination. When you are the first company, you are responsible for how the industry and the industry as a whole interface not only your company, but also society. And there are several problems …
We have a role as you think you are in Frontier … we have the role of an educator, and what you think is coming, and you It’s like a lookout to convey society what the effects will be. It is not always right, but it does not depend on us or other companies.
It is up to us, there is a change we are here, there are some ideas here, here is our recommendation. However, society needs to determine how economic impacts are eased, how to distribute profits, and how to deal with the associated issues. So we are voices, important voices. Also, we don’t intend to say that the technology we create is not responsible. Of course, we do it, but it must be a conversation between all stakeholders.
Looking at the Indian IT industry, they do not build a very smart model on it, and provide services on it, but provide services on top of it. It came really well. Do you think they should do it with AI? Or do you think you should do more?
I think India should look for a complete stack approach …
… This requires a lot of capital.
Well, it’s not a cheap project, but it’s worth it.
You have more than 300 million users …
more…
… Well, and what did you learn about the use of chat GPT?
Can I show you something? Because it is really meaningful. I was just looking at X (turn the computer to display the screen). So this guy, we are not really friends, but I know him a little. The deep -search was released a few days ago, and his daughter has a very rare cancer. It’s like stopping work. He summarizes a big private research team (to understand her illness). He is collecting all of these money, giving him a better answer than his hired private research team. And I see that really meaningful for us.
Do you expect President Trump to take more measures to protect American leadership in AI? Do you see it happening? Or are there any national games to play with AI to make questions another way?
Of course there is. But our mission is that we are very serious, but the AGI (artificial general information) will benefit all humanity. I think this is one of these rare things across borders. AI is like wheels and fire, industrial revolution, and agricultural revolution, not in the countryside. It is everyone. I think AI is one of these. It’s like the next step in it. And they do not belong to the country.
You first talked about artificial general information a few years ago. Did we approach it?
Yes, when you think about what you can do now, compared to what they can do a few years ago. I think we are definitely approaching …
Are we more adventurous in our failure now?
A place where we moved a few years ago … I am thinking about how much progress has made model safety and robustness compared to two years ago. As you know, look at the current model elucidation rate or the ability to comply with a series of policies. We are much better than two years ago. That doesn’t mean we don’t need to solve it for thinking like Super Intelligence (far exceeding human intelligence or theoretical components of intelligence). Of course, we do so, but we are on a wonderful track there.
Did you see a ranced paper on Sweden’s breast cancer research yesterday? They used the AI model called Transpara, but I don’t know if you’re familiar with it. They were accurate diagnosis by 29 % and found without accidental detect …
That’s great. How well did you think that AI needs to be allowed to be driven the other day? How much is AI as a diagnosis than a human doctor before the diagnosis is allowed? Obviously you need to get better. Autonomous driving cars must be much safer than human drivers for the world to accept them. But do you need many of these research before an AI doctor wants?
Regarding the diagnosis, I think the bar is much lower than a car …
In the case of a car, probably subjectively, you want to make it 100 times safer. It should be much lower for diagnosis.