On AI, and the reason why my 2022 predictions failed completely

Ever since GPT4 came around, I've been very keen on AI. Not so much on AI itself (though I do use a lot for research and daily use) but on the social, psychological influence AI has brought us. 

Around four years ago I wrote an article "meditating" on GPT4, which was like a miracle back then. I remember Ted Chiang's article deeming LLM was just a "blurry image of the Internet," which is still a pretty solid conclusion now, and his very to-the-point forecast that LLMs would only bring out floods of "content mills" that gave us only the "abstracts" of everything, of which everyone soon got very tired. And I remember myself reflecting on how LLMs contained infinite information and almost zero knowledge of the world, making it still the information machine, having nothing to do with being intelligent, as what we would usually call a smart person so, be it book smart or street smart. All the more distanced for the street smart type. And yet such information machine would indeed change something, something very subtle, and very different from what I predicted back then. At the time I thought there would be a cognitive revolution, not from the AI world but from the human world, that such powerful tool as GPT4 would greatly empower the geniuses, the "Superman" (or "Overman") in Nietzsche's sense, making them truely powerful in almost all aspects of live. The Last Man or the Mass Man would still enjoy their miserly happiness, but they will inevitably step down from the throne, making way for the new species, the LLM-empowered Superman to rule the world. That might be a bit exaggerated, but it was genuine enough.

Four years have gone by, and none of the above predictions happened. Why? And what happened instead?

I can only provide some personal observations over the years. First of all is what happened to me. I find myself constantly relying on LLMs to provide practical advices on almost everything. Got a headache? Ask ChatGPT what type it was and what to do with it. Looking for a present for my girlfriend? Let Gemini list 10 gift choices based on my description of her. Had a fight with my family? Again, ask Gemini for help. Now, all of the above usages are clearly very down to earth, daily, mundane, and therefore never before occured to me that these could be articulated as questions. Also, due to my former experience with ChatGPT (that was when Google's Gemini was not nearly as good) for my pseudo-academic usages, I tend to ask Gemini more practical questions and leave the academic ones to ChatGPT. Seems I naturally seperate two pragmatic fields of my life and apply accordingly to the two LLMs. 

I also urged people around me to start using LLMs to deal with these small questions, like my parents. To be fair, people like my parents, that is, Chinese people in their 60s with crappy elementary education and almost zero understanding of the whole Internet and smart phone world, still have a lot of questions for daily usages (maybe more than the younger generation) like how to cook a certain dish, or whether a diabetes patient can eat a certain fruit, or maybe recommendations on TV dramas. The last question is actually my supposition for them, because they would never raise such questions to themselves. They wouldn't even question the display on the TV app home page, which is clearly driven by promotions, advertisements and so on.

So this is just a beginning of my analysis. The difference is much more on the usage than on the cognition or the worldview. My four-year-old prediction failed precisely because I focused too much on people's intelligence, rather than people's daily usages, just as nowadays fancy talks on AI tend to do. It is never about the intelligence, it is about usage, custom, practice, habit. And if something is to have an influence on the society, it's got to be on the usage. The mental damage Tiktok has on a society is precisely on the usage, even though the product itself lacks intelligence, or precisely because it lacks intelligence. It is designed to make full use of the biological weaknesses of the human species and to form unnatural usages therewith. Whether the AI tech becomes a boom or a scam depends on whether the usages a society forms from it will be good ones or bad ones, only this time the externality will be so vast, so pervasive, that we can hardly see it on an industrial scale, but only on a national scale.