The sweet spot of the current LLMs (not whatever the next gen might or might not improve on) for me is similar to suddenly having an army of idiots at my finger

The sweet spot of the current LLMs (not whatever the next gen might or might not improve on) for me is similar to suddenly having an army of idiots at my fingertips. There are a lot of smallish tasks/problems people/systems needs to deal with, some of them even waste notable real engineering capacity, and a highschooler could do manually quite easily by hand. Example: find out if a text contains an email address, including all kinds of shenanigans people do to mask it (may not be allowd, ... whatever). From a purely coding standpoint, this is a cats-and-mouse game of improving regex solutions in many cases to also find the more sophisticated patterns, but there will always be uncatched/new ways or simply errors that produce false positives. But a highschooler can be given a text and instantly spot the email address (or confirm none is in there). In order to "solve" these types of small problems, LLMs are pretty much fantastic. It needs to only be reliable enough to produce a structured answer within a few attempts and cheap enough to not be a concern for finance/operations. Thats why for me it makes absolutely sense that the #1 priority for OpenAI since GPT4 has been building smaller/faster/cheaper models. Automators need exactly that, not genius-level AGI. Also for me I think we're not even scratching the surface still about many tasks can be automated away within the current constraints/flaws of LLMs (hallucination, accuracy, ...). Everyone tries to hype up some super generic powerful future (that usually falls flat after a while), whereas the true value of LLMs is in the many small things where hardcoding solutions is expensive but an intern could do it right away.