The unpredictable revolution of AI

A couple of days ago, I read two reports on what AI means in the near future. One was a news report on how the Indian IT services industry now has an AI component in almost every new deal. The article projected a 25-30 per cent growth in AI-related spending by clients of large tech outsourcing companies. There were some more details that supported this growth story, and all in all, the general tone was in support of the current narrative on AI use in business, that AI is well on its way to becoming an indispensable tool for companies looking to stay competitive and efficient.

Coincidentally, the same day, I read a report in US Bank, Goldman Sachs' Top of Mind journal titled 'Gen AI: Too Much Spend, Too Little Benefit?' While there is a question mark at the end of the title, the article is an excellent overview of why the AI hype might, in fact, be mostly hype, at least for a long time to come. The article begins with the statement, "Tech giants and beyond are set to spend over $1 trillion on AI capex in coming years, with so far little to show for it. So, will this large spend ever pay off?" and then the opening paragraph.

Much of the analysis is based on interviews with people whom one would expect to have a relevant opinion about AI and its impact. One of them is MIT professor Daron Acemoglu, who expresses deep scepticism about AI's impact. He predicts that only 25 per cent of AI-susceptible tasks will be economically viable to automate in the next decade, which would be less than 5 per cent of all tasks. Note that he's not talking about technical feasibility but economic viability. Acemoglu doubts that AI will advance as rapidly or impressively as many expect, challenging the historical pattern of technologies becoming more efficient and less expensive over time. He also questions whether AI adoption will generate new tasks and products, arguing this isn't guaranteed. Consequently, he forecasts AI will only boost US productivity by 0.5 per cent and GDP growth by 0.9 per cent over the next 10 years.

Somewhat along similar lines is the argument of Jim Covello, who is the Head of Global Equity Research at Goldman. He argues that to justify the estimated $1 trillion development and operational costs, AI must solve difficult and complex problems, which he believes it's not designed to do. Covello contrasts AI with truly transformative inventions like the internet, which disrupted high-cost solutions with low-cost alternatives even in its early stages. He's doubtful that AI's costs will decrease enough to make widespread task automation affordable, given its high initial costs and the complexity of crucial components like GPU chips, which may limit competition. As someone with an investment perspective, he is also sceptical about AI's ability to increase company valuations. He suggests that any efficiency gains would likely be offset by competition, and the path to revenue growth based on AI is not clear. He also questions whether AI models trained on historical data can ever truly replicate humans' most valuable skills, but that's a separate story.

So far, in my own and Value Research's dabbling in AI, I feel that it's a great tool to have in many businesses and professions. However, if you ask me whether the amount of money that has gone into and apparently will go into AI could be justifiable given what it seems capable of, then the answer is very much up in the air. Not only that, a sceptical point of view is entirely defendable.

Of course, this goes against the mood of the moment. In our various roles in life—as investors, professionals, business persons, or (most of all) parents wondering how AI will impact our children's careers—we can't help but feel that the rise of AI is a very big deal.

However, it's important to be aware of all sides of the story and not just agree with the consensus. New technologies take decades to have a widespread impact, and even then, they are never what was projected at the beginning.

AI is definitely a revolution, but revolutions never run along expected lines.

Also read: 10 mutual funds with the maximum exposure to AI stocks