I've been wanting to sit down and write about AI forecasters for over 3 months now, but I just haven't gotten around to it. This year they really earned at the level of leading forecasters, if not better - it is much easier for agents to analyze and filter thousands of news, evaluate their outcomes and probabilities than a living person. The most interesting question for me now is how society and politicians should implement predictive systems, and how to use them for the common good. From AI-2040: Honest AI predictors help a lot here. Just a few years ago, if AI said one presidential candidate was better than another, people would suspect bias, and the software company would be embarrassed and retrain the model to avoid such questions. Now, thanks to transparency and improved AI alignment techniques, there are much smarter models that people can see for themselves without bias. Over the years, these models have gained an excellent reputation for the accuracy of their predictions.
When such AIs speak about public policy issues, people listen to their opinions—especially if different models developed and trained by independent companies reach the same conclusion.