| ..and the last solo writers, and the last solo accountants, and the last solo medical insurance evaluators, math tutors, etc. I am doing a medical claim evaluation system (which I have been clear that they need the human to do their own review after, but they do have "copy" buttons next to each field) because I needed the money and that was the best I could find on short notice. As it is currently, the humans seem to be easily overwhelmed by the constant flood of requests which all seem to require them to apply as little human judgement as possible. They do have actual doctors there, but it's like a legal exercise more than it is applying any actual human judgement in 95% of cases. The humans in this case are actually finding and applying one or more of thousands and thousands of rules and predetermined decisions based on prior studies. There is an enormous effort in these guidelines to avoid as much actual judgement on the spot as possible by including criteria for what treatments have to have already occurred to unlock the next thing, etc. So it's clear to me that this entire industry would have already been 98% automated if they had LLMs before. The next project I am going after is an instructional design project that generates practice
standardized test questions for different grade levels. In a previous project I made an actual voice-to-voice AI tutor. I think we are rapidly heading to a point in the next 0-3 years where if you want an affordable service, you will only be able to get an AI. Interacting with actual humans will be for wealthy people. I mean the educational project is in a very elite neighborhood, and yet from looking into them, I see that parents complain about how expensive their tutoring services are, which is surely the next project after the practice test generation. Also, the last interview I had to try to get a freelance contract online, the potential client had pre-generated all of the requirements, architecture, and indeed entire project code using ChatGPT. The only thing preventing them from doing the entire thing was the complexity of deploying it and the fact that the future of their business depended on it being a robust solution. But I am literally competing with the AI in terms of the architecture because the degree to which I deviate from what it suggested to this potential client (which is actually fine, it's just the using the most popular approaches of the moment) I feel the client may question all such deviations. And I tried to give a realistic but extremely tight timeline that would allow me time to actually write a lot of the code myself if I wanted. But this may have been a mistake because they will have dozens of other candidates pitching an estimate that assumes the use of AI with minimal human intervention. |