Investors warned of AI bubble
Spending on artificial intelligence (AI) in the wake of hype surrounding large language models is set to hit US$1 trillion (approx €926 billion) in the coming years, with cash splashed on data centres, computer chips, and even the bolstering of power grids.
That impressive spend may yet turn out to be a waste, according to a major investment bank.
Goldman Sachs, in a report published this week and intended to inform would-be investors, said that despite the spending spree, generative AI would have a limited economic upside in the next decade, arguing that “the technology isn’t designed to solve the complex problems that would justify the costs”, and that there was “little to show for the spending so far”.
The report, entitled Gen AI: Too Much Spend, Too Little Benefit? – co-authored by Daron Acemoglu, an economist at the Massachusetts Institute of Technology, and Goldman Sachs researcher Jim Covello – stands in contrast to not only spending on AI but breathless research reports from sundry analysts, management consultants and banks. Banks including, for the record, Goldman Sachs itself, which in 2023 said generative AI could see GDP rise by 7%.
No ‘killer app’ has yet emerged for generative AI, the report says, going on to say that even if one does, its impact on investment returns would be unclear.
Notably, the report is scathing about the cost of the hardware and infrastructure required to run the technology. However, it is precisely this that is causing shares in companies such as chip designer Nvidia to remain at elevated levels.
AI is a bubble, Covello said, but like many bubbles of the past it could bob along on a sea of cash for some time because investors in “picks and shovels”, referring to more money being made by sellers of equipment than miners during the 1848 California gold rush, will continue to pump money into infrastructure despite the fact that claims made for AI as whole are “unlikely to hold up”.
However, Covello’s criticisms are broader than chip company share price inflation.
Unlike previous waves of technology, such as the Internet, which had an immediate and obvious impact on the economy, AI does not actually allow anyone to do things that were not before possible.
“The crucial question is: what $1tn problem will AI solve? Replacing low wage jobs with tremendously costly technology is basically the polar opposite of the prior technology transitions I’ve witnessed in my 30 years of closely following the tech industry,” said Covello in the Goldman Sachs’ newsletter.
The criticism is timely. While every tech company on earth, not to mention every consultant the nature of whose actual job is a mystery even to them, is champing at the bit to get some of the sweet cash being poured into AI, concern is growing that even in cases where the technology can be used it simply costs too much to bother. AI’s energy costs and attendant emissions, are nothing short of ferocious.
Whatever about what AI may or may not deliver ‘real-soon-now’, Goldman Sachs’ report should be welcomed as a riposte to the fantasies of the coherent but uninformed. Just this week, former British prime minister Tony Blair gave a speech claiming that “the only long-term solution to grow productivity and better public services is the full embrace of this technology revolution”.
Ttalk is cheap. The likes of Blair can stand in front of a nodding audience, speaking of “a revolution on top of the revolution,” but firing off superlatives does not equal cold-eyed analysis. In fact, it’s tempting to conclude from the speech that perhaps it really is the case that some people’s jobs can be replaced by machines that spew out word salads imitating thought. Our misfortune is that they won’t be.
Subscribers 0
Fans 0
Followers 0
Followers