I had the privilege of attending the executive programme on AI at the Massachusetts Institute of Technology (MIT). It was an extraordinary two weeks spent alongside 60 business leaders from more than 20 countries, representing a wide range of industries, being taught by some of the world’s brightest minds on the subject.
Amara’s Law states that we tend to overestimate the impact of new technology in the short term and underestimate it in the long term. Given the present level of hype about AI, the idea that we might still be underestimating the long-term implications is hard to take in. But two weeks at MIT made it clear to me that we probably are.
AI has already matched or exceeded human performance across a wide range of expert tasks, in some cases by a big margin. That gap will widen. Hundreds of billions of dollars are being invested, and the scaling laws that have driven recent progress still hold. Whatever you think AI can do now, it will do more in future.
This leaves an obvious question: what is left for us? The instinct is to answer by listing the things AI cannot do. But that is the wrong approach. AI’s capabilities will keep changing, and any answer based on what it cannot do now will be out of date soon enough. The better question is this: what should remain human, and why?
