“The 7 deadly sins of predicting the future of AI” is a really long read. I haven’t finished it yet (and probably won’t this week). But it does illustrate, even with skimming, not only some of the problems we have with predicting the future of AI, but the problems we have with predicting the future itself. There are all sorts of biases that creep in. I think the first one – “we overestimate the short-term and underestimate the long-term… but how ‘long’ is the ‘long-term’?” is the most common mistake we make in forecasting.