In San Francisco “it is taken as read” that artificial intelligence (AI) “will transform” the world, says The Economist . But for all the hype, the technology has so far had “almost no” discernible impact on business practices or productivity. There are signs of a “backlash” to the AI boom , says Neil Shearing of Capital Economics .
Recent investment analysts’ notes have highlighted “the technology’s shortcomings”. Such a period of disillusionment was inevitable. New technologies always see a lag between introduction and wider economic effects. IT developed in the 1980s and early 1990s didn’t raise US productivity until the late 1990s.
“The boost to productivity from AI will be substantial”, but it could be a few years yet before that feeds through to corporate earnings and GDP . Just as during the “railway mania” of the 1840s, “investors are attempting to capture the benefits of new technologies ahead of them fully materialising in the real economy”.
Is AI overrated?
At the peak of 19th-century railway mania, investors were pouring the equivalent of “around 7% of Britain’s national income” into rail projects, says Edward Chancellor on Breakingviews .
As today, there was much grand talk of technology advancing “human civilisation”. Overcapacity and resulting losses were “large and largely foreseeable”. There were, for example, “three separate lines connecting Liverpool with Leeds”. The boom ended with a financial crisis in 1847. Yet the parallel between AI and rail is not perfect. By the 1840s steam locomotives were a fairly mature technology.
By contrast, the “hype over self-teaching computers” appears much more exaggerated. ”Daron Acemoglu of MIT argues that AI buzz 'obscures the reality of what is really a quite limited technology',” says Andrew Orlowski in The Telegraph . By his estimate, “just 4.6% of tasks can be reliably automated”. That makes AI more comparable to an Excel spreadsheet than the harbinger of a “fourth industrial revolution”.
AI uses pattern recognition to guess the right response to a query, but an “intrinsic feature” of such machines is that they confidently “make stuff up” when they can’t find a reliable answer (a problem known as “hallucination”). This renders AI “useless for many potential use cases”. Previous technology bubbles left us with Victorian railways and internet fibre infrastructure.
But today’s tech firms are “spending $1 trillion on data centres” that will soon become outdated as new chips hit the market. “This is capital incineration on a vast scale.” Financial “bubbles can take a long time to burst”, so there is reason to stay invested for now in the beneficiaries of the AI spending splurge : the “chip manufacturers, utilities and other companies exposed to the coming buildout of the power grid”, says Jim Covello of Goldman Sachs .
AI is hardly the first Silicon Valley “tech hype cycle. Virtual reality, the metaverse and blockchain” also attracted huge investments, but still have few real-world applications. Covello thinks that “if important-use cases” for AI “don’t start to become more apparent in the next 12-18 months”, then investors’ enthusiasm “may begin to fade”.
This article was first published in MoneyWeek's magazine. Enjoy exclusive early access to news, opinion and analysis from our team of financial experts with a MoneyWeek subscription .
Most Popular
Comments / 0