I'm having second thoughts about the speed of AGI

I used to be in the camp of "AGI in the next 5-10 years" but after reading the book "Difficult Things" yesterday, I'm in the camp of "there is no way we are within 10 years of AGI."<br><br>I've always thought that it's fine for these models to be dumber and more flawed than us, as long as they are faster and more patient. But after reading "Difficult Things," I think it's really hard for AI to understand how to be faster or more patient without understanding the underlying science behind what they are doing.<br><br>For example, if I were to write an entire book on AGI, I would be able to do it faster than you, but I would do a terrible job explaining the underlying science behind how it works. Even if AI were twice as smart as us but had the ability to write an entire book about AGI in 5 minutes, they would probably say things like "we need to get rid of [insert problem in the field of AGI]... somehow."<br><br>I also don't think it's possible for an AI to make a mistake for a couple of years and then be like "well, I finally figured it out. I was wrong for 2 years.<br><br>Am I right about the pace of AGI? What are your thoughts?