In Gartner’s Hype Cycle of Emerging Techs, Machine Learning - the most common AI tech in the digital sector - has crossed the Peak of Inflated Expectations, speeding downwards to the Trough of Disillusionment. Looking at how the trends develop, the prediction appears to hold true, with the first cracks appearing where the glass ball of hype meets hard reality.
Heading down the Trough
The first point of impact are overstretched expectations about what the technology can actually do. Robots find IKEA chairs hard to build, and Tesla’s boss tweets that “humans are underrated”, as excessive automation hits production targets. Even if you are familiar with Moravec’s paradox (which explains how AI can beat humans in chess, but has trouble crossing a room without bumping at things), your expectations are soon to be adjusted.
Secondly, ML algorithms feed on data - high volumes of data laid out in analytical cubes. Our friends at Enjins, a Data Science start-up in Amsterdam, emphasise that before they can create predictive ML models, the hard work of data collection and consolidation needs to be done. Dealing with legacy systems and organisation silos is no mean feat, and organisations need to be prepared for the groundwork required.
Finally, as ML moves to the mainstream, the accuracy of its predictive powers will come under the microscope. At a recent Episerver Partner event, we were presented toolkit that delivers Individualised Content. Among the excitement, it was easy to miss a point that unlike product recommendations, irrelevant content is a major own-goal. Like all techs, ML needs to be tested and we need to be clear that testing ML is as new as the tech itself.
Towards the Plateau of Productivity
Web developers have seen this before. When responsive design was all the rage, clients thought it would solve all their problems, and Bootstrap would make it effortless to implement and test. As reality settled in, we hit the Trough hard (remember “Web Design is Dead”?), but we came through, and can now enjoy making Design Systems in the Plateau of Productivity.
Like the Hype Cycle model, I am a techno-optimist: AI will eventually make it to the Plateau, and deliver the promised value to organisations. But as we breathlessly present its potential, we must also be realistic about its limitations, as well as the human effort required to implement and test it. The view from the Plateau is beautiful, but to get there we need a little less hype and a lot more pragmatism.