Job transition with AI will not be as dramatic as feared: Susan Athey

Susan Athey talks about the impact of artificial intelligence (AI) on the workforce and the economy and how interoperability is important for competition in the digital space

Susan Athey, professor of economics of technology at Stanford University, talks about the impact of artificial intelligence (AI) on the workforce and the economy and how interoperability is important for competition in the digital space, in an interview with Ruchika Chitravanshi and Asit Ranjan Mishra in New Delhi. Edited excerpts: 

 

Do we expect AI adoption to outpace earlier technologies like personal computers, or will regulatory, infrastructure and trust bottlenecks slow its adoption? 

 

I don't think there's a single answer. Some uses have low adoption costs and will go quickly. Using ChatGPT on a browser is very easy to access. For companies, using internal chatbots is easy to adopt but that does not change the economy all that much. A lot of the things that change the economy have bigger adoption costs and go more slowly. 

The easiest change to make is to stop hiring entry-level workers because you don't know what you want them to do.

 

So that's where we're seeing the quickest response. My overall conclusion is that the job transition will mostly not be as dramatic as some people fear. But in very specific groups, like entry-level workers or call centres, it will be bad.

 

India has a huge amount of IT outsourcing, and there, people have to change what they do. When something gets cheaper, like coding, it doesn't mean you're going to do less coding.

 

It actually means that more companies around the world can do software projects. The huge worldwide backlog of software projects, which were not economical to do before, can be done now.

 

I'm actually cautiously optimistic. I think some firms will fail. There'll be reorganisation of the industry. The ones who do it best are going to grow, and the ones that don't are going to shrink.

 

How should countries like India respond if AI disproportionately automates high-paying and white-collar jobs rather than low-skilled work?

 

They need to see where the new demand is, and so they are going to have to pivot. They will need to embrace AI to make their workers more productive. Every government, every company in the world is going to want to do new software projects, and the process of adopting AI takes work and expertise. Companies need to figure out how to meet that demand, but the demand is there.

 

For the people in college right now, it is kind of scary. They have to be more nimble, they can't just follow a formula.

 

People should learn about a domain. It's more important than before that you know something besides coding.

 

The central government is considering that companies which use data from Indian content creators to train AI models should share a portion of their global revenues once these systems are commercialised. Is it fair or will it discourage innovation? 

 

The worst thing for innovation is when the rules of the road aren't clear. And so, having some kind of system is better than no system. Some kinds of content are very important. I happen to believe that news is very important. And if you take away the business model of the news, then we won't have news and we need news. So, I think news is something special because you have to spend money on it every day. It's been a problem worldwide that we don't yet have the frameworks to make sure that the news industry gets compensated while innovation goes forward.

 

And to complicate matters, of course, the news industry itself is innovating and creating AI products. So (it is better), getting frameworks sooner than later, but those frameworks have to protect creators in some way. 

 

India wants to come up with ex-ante regulations for systemically significant digital enterprises. What do you think of this approach? 

 

I was in the Department of Justice in the Joe Biden administration. We were very successful with monopolisation in technology. But lawsuits are very slow. And so, enforcement is ineffective when the world is changing so fast. Right now, we see that incumbents are attempting to use the control they have to leverage their current position.

 

We see it in enterprise software where all over the world, companies that control data, whether it is customer databases, software or health record software, refuse to interoperate with new AI companies.

 

The incumbents are trying to hold on to their market power and prevent new entrants from competing. So, ex-ante is required.

 

And, I think some of the principles that we found effective in the past are rules about interoperability. Give the consumer the choice. They should be able to choose an AI provider that offers them more privacy or better services.

 

These policies are often considered as non-tariff barriers. A provision allowing the regulator to charge penalty on global turnover is being challenged in the high court. What do you think?

 

I don't want to comment on the specific legislation, but I would just say, India has a strong position. I think interoperability is very important.