Is anyone actually using generative AI?

© Steven Boykey Sidley

(ideogram.ai)

Here is a surprise. It turns out that not very many people are using generative AI. I should add the critical qualifier ‘yet’ but, given the oxygen that this new technology has sucked up over the past 18 months, I would have expected it to have embedded a little more deeply into our lives by now. But it hasn’t.

Before presenting the stats, we need to get a bit of taxonomy out of the way. We all use AI all the time. It has been part of our daily lives for a decade or two – features such as spell checkers and spam filters and facial recognition and the like. Generative AI (as typified by ChatGPT and its ilk) is a new strand of AI. These systems have been trained on a large corpus of human knowledge (text, sound and imagery) and use some fancy statistical wizardry to make various predictions and assumptions about the way we use language and images and such. Then we tell them to do things as though they were all-knowing, multi-skilled assistants.

The first time you try it, it feels like magic. And the second. And the third. And pretty much every time thereafter. There have been a couple of glitches, widely reported, but those become rarer as the software gets hammered into shape by some of the smartest people on earth.

I am one of those who have found GenAI really useful in the research that is an important part of my daily life (I use a new search program called perplexity.ai, which is superior to good ‘ol Google in myriad ways). I have a few friends and colleagues who use GenAI extensively (for summaries, abstracts, story ideas, plots, legal opinions, cover designs, graphics, titles, list ideas, curricula development, recipes, tourism options, writing code). And yet, if I do an unscientific poll, I am pretty sure that most of the people I know will report they have tried it, said wow for a minute, but have not really bothered to dabble further, sliding happily back to their pre-ChatGPT lives instead.

This is borne out by some recent stats. (I am quoting from a blog by the excellent tech commentator Benedict Evans – anyone looking for sources can find them here).

Here is the first shocker. Somewhere around 50% of people in the US, UK, Denmark, France, Japan and Argentina have tried ChatGPT (or similar). But less than 5% use it daily, less than 10% use it weekly and those numbers have not grown much over the last 9 months. The fact that 50% of people have tried it is, of course, pretty astonishing a mere 18 months after this thing arrived on the scene. The fact that most have abandoned it tells a bigger story. It seems to be a solution without a clearly defined problem.

Wait (I hear some of you object), killer use cases are coming, they are in development, in beta, in proof-concept, in pilot. Yes, they are, and I have seen a few (and am involved in one). But there is a difference – the ‘killer use cases’ are products, whereas ChatGPT, Mistral, Gemini, Claude and the rest are more like general purpose know-it-alls. They can do just about anything and therein lies the problem - what MBAs refer to as a lack of product/market fit.

Evans puts it succinctly – “...it’s a technology that can enable a tool or a feature, and it needs to be unbundled or rebundled into new framings, UX and tools to become useful. That takes… time.”

I remember this exact problem when I was a newly hatched programmer and I had just bought my first computer (before they were freely available off the shelf). I had the machine, I had a software development kit, I had the development skills and I had the time. I had no idea what to do with my machine. It sat gathering dust until I went out and bought products which addressed my specific needs – a word processor, an email system, a video game.

Which leads us to the next conundrum – the role of GenAI in business. Various big players like Accenture have reported huge increases in AI revenue over the past 18 months by providing services to major enterprises. But, when Mr. Evans looked under the hood of this reported increase, he found out that almost all of the services provided were basically starter kits. ‘We’re going to try and explain what this AI thing is, and how you might be able to use it’. Theey were not development contracts. They were mostly, erm, high level explainers, suggestions. There was not a lot  of production going on that might actually reduce costs or increase productivity.

Evans refers to an Accenture study that looks at GenAI usage within enterprises. It is finding some early traction amongst the low hanging fruit like software, customer support and marketing but, as for the rest – finance, supply chains, HR, manufacturing, sales and legal – not so much.

Evans comments wryly -  ‘(this is) what happens when the utopian dreams of AI maximalism meet the messy reality of consumer behaviour and enterprise IT budgets - it takes longer than you think, and it’s complicated”.

Early days? Perhaps. But I suggest that there will soon be a massive popping of bubbles underway. Expectations have been raised to silly levels over the last year or so. Venture capital faucets have become firehoses and scores of new products are being released every day, most of which will quietly wither and die on the vine.

The world – us as individuals, and the companies that drive our economies, is less nimble than we like to believe. We don’t really like change as much as we like to think we do. Change is exhausting, after all, even scary. We change when we have an incentive to change, such as the potential of a better experience or the threat of job loss, competitive irrelevance or reduced profits.

AI-fuelled change is coming, without a doubt, but not tomorrow. Right now, we are not all ready for it.

Steven Boykey Sidley is a professor of practice at JBS, University of Johannesburg. His new book It’s Mine: How the Crypto Industry is Redefining Ownership is published by Maverick451 in SA and Legend Times Group in UK/EU, available now.