Before a technology hits a threshold of "becoming useful", it may have a long history of progress behind it. But that progress is only visible and felt to researchers. In practical terms, there is no progress being made as long as the thing is going from not-useful to still not-useful.
So then it goes from not-useful to useful-but-bad and it's instantaneous progress. Then as more applications cross the threshold, and as they go from useful-but-bad to useful-but-OK, progress all feels very fast. Even if it's the same speed as before.
So we overestimate short term progress because we overestimate how fast things are moving when they cross these thresholds. But then as fewer applications cross the threshold, and as things go from OK-to-decent instead of bad-to-OK, that progress feels a bit slowed. And again, it might not be any different in reality, but that's how it feels. So then we underestimate long-term progress because we've extrapolated a slowdown that might not really exist.
I think it's also why we see a divide where there's lots of people here who are way overhyped on this stuff, and also lots of people here who think it's all totally useless.
2 years later my sister uses it for almost everything and despite her duties increasing she says she gets a lot more done rarely has to bring work home. And in the past they had an English major specially to go over all correspondences to make sure there were no grammatical or language mistakes that person was assigned a different role as she was no longer needed. I think as newer generations used to using LLM for things start getting into the work force and higher roles the real effect of LLM will be felt more broadly as currently apart from early adopters the number of people that use LLM for all the things that they can be used for is still not that high.
[1] https://bughunters.google.com/about/rules/chrome-friends/574...
[2] https://www.mozilla.org/en-US/security/client-bug-bounty/