AI is amazing. But AI is also awful. I see so much sloppy, low quality, absolutely midwit work done with AI. The worst part is that many AI users have no taste whatsoever, and are actually proud of their completely error-prone midwit slop. And they don’t understand that the numbers AI generates are often wrong, because they themselves aren’t good with numbers. Even worse is when they deny they used AI to generate the slop. AI has essentially created a completely new class of midwit, separate and distinct and only partially overlapping with the US university educated midwit. It’s similar in that both are just mindlessly regurgitating what they learned. But I can’t believe that I actually prefer, at least some of the time, the style of the old midwit. Bring NPC back?! AI is the new NPC. To be clear, I see all the positive sides of AI. It’s amazing for data analysis, for visualizing via Midjourney, for organizing search results, for summarizing large documents, and for cleaning up audio transcripts. It’s often a better doctor than a doctor, or a better lawyer than a lawyer. But only in skilled hands. High IQ prompting and verifying gets high IQ output. Low IQ prompting and verifying gets complete slop. I think of AI as amplified intelligence, as opposed to artificial intelligence. Moreover, the text that AI generates is placeholder text. It’s the new lorem ipsum, it’s lorem aipsum. It is at best a suggestion and meant to be completely rewritten in your voice. By contrast, AI images and video are higher quality, in part because they are more obviously AI. Meanwhile, the Chinese are focusing much more on AI in the physical world, for drones, robotics, self-driving, and manufacturing. Elon is doing this but relatively few others are. Those physical use cases, at least right now, are pulling out of their trough-of-sorrow stage and getting to full production. Self-driving cars are real and they are here. Anyway, the problems with digital AI may be solvable. We’ll have all kinds of tools to detect AI and prove that something is not AI. And it’s very possible that models could evolve to generate higher quality results, or at least highly personalized results, so that it’s not the same generic slop from so many. But just saying: I love AI, but I also kind of hate when it’s used badly. And I can see why others will just see the downsides and not the upsides. It’s extremely similar to crypto in this way. And it’s very simple for people to just say “no AI, no crypto” without parsing through the downsides to see the upsides. I do think we will get through the other side and turn AI into a power tool with a higher hit rate. Just like we don’t think that much about people getting electrocuted by power lines. The bugs will get worked out. But it’ll be both cultural and technological adaptation.