Will the current level of AI research lead to KOLs losing their jobs? My answer is very unlikely! Take a look at @the_nof1's final leaderboard, which went viral recently. The performance of several large models was shockingly poor, resulting in massive losses, with a median loss of over 50%. My assessment of current AI is that it's an analyst + language master, capable of solving "small problems," but far from solving complex "big problems" in the real world. The core reasons are: - Large models excel at summarizing, but lack innovation; - Scaling laws are gradually becoming ineffective; Severe illusions. After seeing the charts below, are you still worried about KOLs losing their jobs?

A bare LLM is more suitable as a smart docking station; it needs to be stacked with other components.

Yes, it needs customized optimization.

The premise is that they are genuine KOLs!

The issue isn't about customization or optimization; AI should be about helping the human brain acquire information and inspire thought. In short, it's about helping KOLs (Key Opinion Leaders) fill their minds with information. It's not about helping them output. If AI bypasses the KOL's thought process and outputs directly, what's the point of the KOL's existence?

In other words, it's not just a matter of the granularity of the LLM (Limited Learning Model). It's more about the path. For example, I might write an article and output the parallel mechanism of Solana. I can ask very detailed questions, and the LLM will output the article, but I myself won't understand it. The next time I output the parallel mechanism of the Move language public chain, it will still be the same, without comparison. However, if I read the Solana technical documentation myself and communicate with the LLM, I can truly understand Solana's parallel mechanism. The next time I study the parallel mechanism of the Move language public chain, I can actually understand it better and compare the two parallel mechanisms, thus more effectively, deeply, and thoroughly interpreting the differences between the parallel mechanisms of Solana and Move, and what their respective advantages are. With AI, the goal should be to accelerate the learning and growth of the human brain. Not to abandon the learning and growth of the human brain.

其实是ai 辅助人脑

Yes, yes, we can't give up on the human brain; our brains are already so excellent, haha.