Interview with Andrew Ng: India’s big opportunity in AI is at the application level, as well as in LLMs: Andrew Ng

He Generative AI The moment has come Startups and companies today have more equal conditions than ever, with India on the verge of opportunities and should focus on developing the application layer on fundamental models, Cursor Co-Founder and President based in California Andres Ng he told Surabhi Agarwal and Annapurna Roy in a virtual interview.

Ng, 48, one of the most influential voices in AI globally, who previously headed Google Brain and was chief scientist at Chinese tech firm Baidu, is also founder of AI learning platform DeepLearning.ai, a general partner at AI Fund and an adjunct professor of computer science at Stanford University.

Much of the media coverage of GenAI has focused on the large language model (LLM) or the base model layer, but it is a capital-intensive exercise and the real value will now come from building applications, he said, adding that his fund is focusing on such startups and they are making an average investment of 1000 yuan per month. He also talked about AI democratising coding for everyone and that AI will not eliminate jobs, but people who do not use AI can be replaced by those who do. Edited excerpts:

What will drive the next phase of Coursera, especially considering the AI ​​reset?

The entire world is at a tipping point. With the rise of generative AI (GenAI), the nature of work is being transformed. This creates a lot of pressure for organizations to upskill their employees. In many jobs, AI will be able to play a major role in 20-30% of tasks. I don’t think AI will replace people, but people who use it will replace those who don’t.

What kind of approach should a country like India take towards AI?

Much of the media coverage of GenAI has focused on the large language model (LLM) or base model layer, but this turns out to be a capital-intensive exercise. A state-of-the-art base model can cost $100 million, maybe more. But the application layerFor some reason, GenAI has had much less media coverage, but almost by definition it is going to be even more valuable in terms of revenue generated, for example, than the base model layer. India has tons of highly skilled people. It would be worthwhile for some to work on training base models, perhaps adapting them to local languages. But I think a much larger fraction of the value that would be created and captured in India would be in building applications on top of these wonderful base models. India’s economy has a huge service sector or a huge industrial sector, a huge agricultural sector, and also a lot of FDI (foreign direct investment). India can find out how GenAI transforms IT services, telecom, financial services, hospitality, manufacturing, mining and construction, textiles, chemicals, or other major industries.

The biggest opportunity for India is to look at the sectors where the Indian economy is already strong and find applications of AI in those sectors that will enable it to maintain its advantages.

People are asking what the ROI is for GenAI investments. Which side of the debate are you on?

As far as the application layer, it’s very clear. I think it’s totally worth it, partly because it’s very capital efficient – it doesn’t cost that much to build valuable applications. And I’m seeing revenues go up. So on the application layer, I’m not worried. Now, on the base model layer, given the capital expenditure on GPUs, I think from a long-term economic standpoint, it’s going to be worth it. But I think from a time standpoint, there’s a lot of pressure to prove the value of that investment in the short term. And I think that’s something we need to keep in mind.

There is an AI race underway. Do you think others will be able to challenge US dominance?

The field is more open than it used to be, but frankly, the highest concentration of talent in GenAI right now is in Silicon Valley. In my entire career, I’ve never seen such a high concentration in one place. I think that’s because a lot of the innovations came out of two teams: OpenAI and my old team, Google Brain.

While there are pockets of talent in Canada, France, the UK and China, it’s clear that China is growing rapidly. No country seems to have an insurmountable advantage, especially when looking at application opportunities. But Silicon Valley does have more technical know-how than any other city in the world right now.

What would you say is the right approach to regulating AI?

The hope that comes from regulation is that it will increase investment in science and research, support industry growth, and support open source development. I was very surprised last year by the intensity of the push to try to crack down on open source. If someone tries to crack down on open source or make it harder and succeeds, almost every country, including India specifically, will lose out because it will deprive much of the world of access to cutting-edge AI.

If governments identify specific use cases that are harmful, such as deepfake pornography, and then regulate their use, that would be far better than blanket attempts to regulate technology against hypothetical harms. They would only stifle innovation.

Regulations that support open source will allow many startups and large enterprises to access it, rather than a small number of large US tech companies becoming the choke point for accessing GenAI.

What is the focus of the AI ​​Fund and where do you see the most opportunities? How does India look from a startup investment perspective?

We focus mainly on the application layer, a little bit on the tools layer. I see a lot of opportunity. We end up creating, on average, one startup per month. We often work with large corporations to try to combine the best ideas in AI with unique insights in the industrial sector to address an important industry problem. One example is that we created a startup to use AI to make ships more fuel efficient, saving 600 or 700 ships around half a million US dollars a year in fuel costs.

We seem to be lucky enough to come across a lot of interesting opportunities to develop valuable AI projects. One of them ended up launching in India: Jivi Health. We were lucky enough to find Ankur Jain, who is a fantastic CEO, whose team in India created some pretty advanced technology that was, maybe still is, at the top of the Hugging Face rankings for LLMs in the healthcare area.

But Indian AI startups are not getting as much funding compared to the rest of the world.

High interest rates make life harder for many entrepreneurs looking for funding. At AI Fund, we budget $55,000 to build a working prototype. After that, if it looks promising, the check is usually $1 million. That level of investment is usually enough for the startup to build a real product. People think that developing AI is very expensive and that is absolutely true if you want to build base models, but because of all this investment in base models, the capital required to build an application on top of that is now lower than ever. But you need to identify the valuable idea and then execute it well.

What is the future of programming, with AI making it accessible to everyone?

Now more than ever, many people need to learn a little programming. The barrier to learning to program is lower than ever because they have a chatbot as a companion.

If everyone learns to program, you can customize it and make your computer do many more tasks than a website can. So this is a major boost to personal productivity. Just as it is not controversial today that we want our children to learn to read and write, I hope that in the future we will want our children to learn a little programming.