Even by the standards of technology industry hype, the excitement around artificial intelligence (AI) marks a new gold standard for hyperbole.
Giving AI the highest possible accolade, Google’s CEO Sundar Pichai has stated that AI technology is ‘more profound‘ than the invention of electricity or the discovery of fire.
Meanwhile, research giant McKinsey Global Institute estimates that artificial intelligence could increase corporate profits by up to $4 trillion a year worldwide in the coming years, and organizations such as Microsoft and Google are racing to integrate generative AI into the apps and operating systems we use every day.
But what if AI isn’t the world-changing, profit-boosting force that everyone believes?
Several high-profile skeptics suggest that AI may not deliver the rapid returns many expect – and that the current AI mania might be a bubble.
Could they be right?
Could AI Be a Bubble?
Bank of America investment strategist Michael Hartnett suggested earlier this year that the current excitement around AI was a ‘baby bubble‘.
The Nasdaq 100 has surged this year, driven by rallies in AI-related stocks such as Nvidia, whose chips power AI applications: Nvidia’s earnings have hit $13.51 billion, up 101% on last year.
Hartnett has compared the excitement around AI-related stocks to the dot-com crash in 2000, which saw investments in the Nasdaq Composite rise 800% before falling 740% by 2002.
Veteran economist David Rosenberg agrees, describing the current interest in AI as a ‘mania of sorts’, writing in a column,
“This type of corporate behaviour is not too different from what took place in the dotcom bubble, with company after company satisfying investors’ appetite for news on how it plans to incorporate the internet into its business — or boosting stocks just because they added ‘.com’ to the name.’
Rosenberg qualified this, however, by saying that he was a believer in the benefits of AI over the longer term.
The Problem of AI Hallucinations – Why Do They Matter?
Other skeptics are critical of the technology itself, with AI analyst Gary Marcus saying that hallucinations (where AI systems such as ChatGPT make up or mangle facts) are not a simple problem to get rid of.
Marcus believes this intractable problem could stop AI from delivering the financial returns its advocates expect.
Marcus told the Financial Times, ‘There is a fantasy that if you add more data, it will work. But you cannot succeed in crushing the problem with data.’
Marcus pointed to previously highly hyped AI technologies, such as Facebook’s AI assistant M or IBM’s Watson, which promised world-changing results but were under-delivering.
Marcus wrote on his Substack, ‘Hallucinations are in their silicon blood, a byproduct of the way they compress their inputs, losing track of factual relations in the process. I first pointed out this risk in 2001, in the fifth chapter of my book The Algebraic Mind, and the problem has persisted ever since. To blithely assume that the problem will soon disappear is to ignore 20 years of history.”
Marcus says that he believes the problem of AI hallucinations will eventually be ironed out, but no one knows whether the technological breakthroughs to solve the problem will arrive in months, years, or decades.
Will AI Boost Productivity?
Claims that AI will boost worker productivity have been central to this year’s AI gold rush – but is it sure that technologies such as large language models (LLMs) will boost productivity?
McKinsey has predicted that up to 50% of tasks could be automated as early as 2030. Still, skeptics such as technology consultant Jeffrey Funk say these gains will likely arrive more slowly than people imagine.
Funk writes on Linkedin, ‘How is AI helping companies improve the productivity and quality of output factories, construction sites, farms, mines, software development projects, or doctors treating hospital patients?
At present, AI tends to focus on tasks, Funk believes, rather than ‘systems thinking’.
The Curse of Recursion
The internet is already filling up with AI-generated content – and some have warned that this will make it harder to train AI systems in the future.
A paper published this summer, ‘The Curse of Recursion: Training on Generated Data Makes Models Forget‘ (PDF), showed that large language models trained on AI-created data degenerate – and that training AI models on the whole of the internet will be increasingly unrewarding.
Ross Anderson, professor of security engineering at Cambridge University and the University of Edinburgh, wrote, ‘Just as we’ve strewn the oceans with plastic trash and filled the atmosphere with carbon dioxide, so we’re about to fill the internet with blah. This will make it harder to train newer models by scraping the web.’
Is the invention of AI, in all its forms, as transformative as fire — or at least up there in terms of changing the path of humanity? Time will tell on this one. But one thing is sure — changes to how we live and work are happening rapidly.
Is the AI boom overblown? It’s certainly the case that investors have gotten very hot under the collar in recent months, and there are known drawbacks to the technology — so it’s perhaps wise to remain cool and collected around AI. Not every AI startup is going to become a billion-dollar giant.
But with investment pouring into AI and business leaders enthusiastically embracing its potential, AI seems poised — despite its drawbacks — to change our lives in a way not seen since the dawn of the internet era.
I, for one, am here for it.