The most important way to understand why some folks are incredibly optimistic about the future of AI capabilities to solve problems is to realize that the human ability to think abstractly, cleanly, and logically, in a structured way that allows us to solve modern technical and theoretical problems, is a very new feature in evolution that most animals have hardly at all, and that humans, on a scale from zero to what is theoretically possible in computational science, exhibit this ability to a tiny, tiny, tiny extent. For example, the OpenAI model that solved Navier-Stokes literally used as many tokens as a human would use if they thought for 4,000 years at a typical work week. If you can appreciate the scale of model advances and how far away we are from any theoretical upper limit, then you can quickly realize that sometime in the early 2030s, the amount of total organized, logical, clean, and solution-oriented thought given to our major problems will be hundreds and thousands of times greater than we’ve ever put into them as an entire civilization. Every year of machines and people in concert thinking about and solving our biggest problems is going to be the functional equivalent of a very advanced human-only civilization thinking about these things for thousands of years. That is the reason right there why completely insane possibilities of massive abundance, discovery, genetic engineering, space flight, etc. are not just possible with advanced AI, but probably very likely.