There’s a lot of AI doom and gloom around at the moment. But the cycle of disappointment and grief is creating some narratives which are bizarrely contradictory. We’re not investing enough in AI! The AI isn’t working! AI will destroy humanity!!!
So a question: what is genuinely stopping AI being adopted, working properly, or being a force for good in the world?
The fact that AI is not a thing.
Since the label was first coined in the mid-50s, it’s always been applied to whichever technology is currently promising to produce intelligent-seeming outcomes. Once upon a time, document scanning was AI.
In 2020, what people refer to as AI is a collection of techniques around data mining, statistical analysis and machine learning, set loose on large datasets, fuelled by increasing computational power.
Alas: endemic sampling bias, naive realism and statistical fallacies conspire to create disappointing outcomes, and this is shaking our confidence in the AI unicorn.
That’s fine – because AI isn’t, and never was, a thing.
What we need to pay attention to is human outcomes, not AI. Let’s invest in solving human problems, using whatever techniques of automation that are proven to work in each case. And it would help if we just stopped using ‘AI’ as shorthand for ‘new technology’.