Is AI’s Solow paradox about to be resolved?

There may soon be a definitive answer to the question of whether AI is overhyped

“You can see the computer age everywhere but in the productivity statistics.”
Robert Solow, 1987

The Solow paradox, sometimes referred to as the ‘productivity paradox’ or ‘productivity paradox of IT’, has largely been forgotten nowadays. But it was a hot topic among tech and econ nerds a few decades ago.

Here’s the tl;dr – despite the mass adoption, at least by businesses, of computers from around the mid-1970s onwards, productivity growth slowed in the US in the 1980s. Many academics, most notably Erik Brynjolfsson, tried to crack the Solow paradox nut in the 1980s and the first half of the 1990s. But then things picked up in the late 1990s and everyone concluded that IT had indeed enhanced productivity, albeit more slowly than predicted.

The Altman paradox
The contemporary version of the Solow paradox might be referred to as the Altman paradox. It could be summarised as follows: You can see the AI age everywhere but in the labour market and productivity statistics.

Many of us, including Sam Altman, continue to insist that AI will profoundly disrupt labour markets, economies and societies. However, it’s almost three years since the first iteration of ChatGPT dropped and AI has yet to profoundly disrupt labour markets, economies and societies.

If you squint hard enough, you can see what appear to be signs that profound disruption is about to hit like a tsunami. But there’s simply no getting around the fact that AI has been more knee-high swell than tidal wave so far.

Reasonably enough, AI sceptics frequently point to tight labour markets in technologically advanced nations and note that the people talking up AI are almost always those who stand to benefit financially from its widespread adoption.

But if the sceptics are correct, that raises the question of why so many Silicon Valley geniuses, tech behemoths and nation states are betting the farm – or at least a large herd of cattle – on it.

Are they all mistaken?

Mobile phones: a case study
While they were invented in the 1970s, mobile phones didn’t become a thing until the 1980s. Even then, they weren’t popular, which was hardly surprising since they were eye-wateringly expensive, unreliable, large and heavy. Oh, and you might have to charge them overnight to get 30 minutes of battery life out of them.

Things improved in the 1990s and early 2000s, especially on the price and UX front, and mobile phones became more common. But they still weren’t that popular, and even those who had them weren’t obsessed with them. Which is, once again, hardly surprising given there wasn’t much you could do with them apart from make phone calls and send texts.

Back at the dawn of the Millennium – when people used to go around with an onion tied to their belt, as was the style at the time – the BlackBerry was all the rage in Professional-Managerial Class (PMC) circles.

It had reasonable battery life and a small keyboard that allowed career-focused go-getters to send and reply to emails when they were out and about. (At the time, it was standard practice to deal with emails on a desktop computer.)

There were a lot of articles during the 2008 US election campaign about the Democratic contender’s ‘Crackberry’ addiction, which seemed to further reassure my left-liberal PMC peers that Obama was indeed the Messiah, sent to heal all of America’s racial divisions. At its peak (in 2009), Blackberry controlled half the US market and a fifth of the global market.

BlackBerry no longer exists, at least not as a mobile phone manufacturer. As you’ve no doubt already deduced, dear reader, it was driven out of the mobile phone market by Apple after it released the first user-friendly, mass-market smart phone in 2007.

Is AI now morphing from BlackBerry to iPhone?
Much like an Eighties mobile phone, AI has so far been occasionally handy but generally not life-changing or economy-rearranging.

An argument could be made that AI has already significantly impacted some industries and demographics. But given the hundreds of billions ploughed into it and the unprecedented hype surrounding it, the real-world impacts have so far been underwhelming.

Let me conclude by proffering an explanation of what’s going on.

For the last three years, we’ve only had weak AI, which is the equivalent of a dumb mobile phone. Of course, having a dumb phone is better than having no phone at all, which is why plenty of people bought phones from Nokia, Motorola, Palm (the company behind the PalmPilot) and Sony Ericsson back in the day.

However, mobile phones only achieved mass penetration – and then something approaching universal penetration – once Apple’s ‘Jesus Phone’ appeared. Anybody who isn’t Amish now has a smart phone grafted to their hand and experiences anxiety if they are separated from their all-purpose device for too long.

I agree with the AI sceptics that AI has yet to live up to the hype. But I think they are foolish to conclude that this means AI is yet another tech industry nothingburger.

As you’ve probably heard, Elon launched Grok 4 recently. It’s stronger and therefore more useful AI. OpenAI has flagged that it will be releasing ChatGPT 5 soon. If reports can be believed, this AI could be even stronger, and therefore even more impressive, than Grok 4.

To summarise, it will have significantly improved memory, hallucinate less, reason more accurately, be truly multimodal (i.e. be able to work with text, image, audio or video) and be able to handle much larger documents than it can currently. In short, it will be much closer to a human personal assistant than a somewhat frustrating but still somewhat useful chatbot.

I don’t know if the release of ChatGPT 5 will be AI’s iPhone moment.

But I wouldn’t be surprised if it was.