Throughout 2025 I kept seeing CEOs of large companies (including AI labs) scaring people by mentioning the possibility that white-collar jobs would be automated by AI. Jobs, representing a source of livelihood and meaning for a lot of people, were directly threatened. As a result, a lot of discussion happened online and offline. While people were feeling all sorts of things throughout these discussions, I felt quite disengaged. I thought that they were worried for nothing. The capabilities of large language models at that time were impressive, but with greatly noticeable technical flaws. Despite my lukewarm reaction, I was lucky to witness the emergence of three interesting archetypes among the people participating in the discussions: the ones swayed by hype, the natural adopters and the fearing ones. The first subset consisted of users who were excited not just about the technology, but about the acceleration of human technological progress in general. They did not agree about the best way to cope with the huge displacement of jobs that would be created by advanced automation, but they were hopeful that, eventually, we would find ways that benefit everyone. The second subset took to the up-and-coming technology like fish to water (for a good example, see LLMs for the Working Programmer, published in August 2024). These people seemed to be better attuned to the nature of these machines. Instead of projecting human traits onto them, they strove to understand them as they were—including hallucinations, stagnation due to mode collapse and other things. They stood out to me quite significantly. They did not seem to be worried unlike the third group, whose online output seemed to be tainted in large part by fear and envy.
No pages have linked to this URL yet.