I think it is reasonable to say we are all experiencing quite a lot of hype about AI related matters at the moment. But what is the specific subject of that hype? Really, it is about a subset of AI called deep layer neural networks and within that another subset called transformer-based large language models. If we ask ourselves about AI overall – the high level general concept of machine intelligence – the hype cycle is not a short-run, recent thing.
Remember that the Y axis of the hype cycle is ‘expectations’. That is – the future expected value of the innovation, in the collective minds of its potential adopting market. There is no simple, single way to measure those expectations. Any measure will be indirect and the kind of indicator that best helps evidence it will vary by the context of the specific innovation. One indicator I often find useful to discover the position of bigger, longer term concepts is Google’s Ngram Viewer. It plots the frequency of occurrence of a term in all books. Sadly they only update it every few years and the last update takes us to 2019. But even so, the curve for AI (a term coined at a specific conference in 1956) is an elegant illustration of just how deep and long the biggest hype cycles can be.
