ChatGPT on its own isn't that dangerous. But it starts making things weird once you combine the GPT API with a framework like langchain, which helps you make "decision pipelines".
This is because ChatGPT is a multi-use LLM, you can use it for text generation, classification, sentiment analysis, and even some coding. If you pair that up with a decision making pipeline, you can set up something like the following:
ChatGPT API generates an article title, passes the article title to a ChatGPT API relevance score calculator - - this happens in a loop until an article title with a high relevance title has been found, then the title is passed to another prompt which has the system/agent prompt for creating subtopics and it's own relevance score calculator . . etc so the entire pipeline can be automated with close to zero knowledge of coding. It used to be that you needed a background in ML to do this stuff, find cosine distances and all that, now all it takes is a system prompt.
This is awesome (for people like myself, a rookie coder) but it also has a side effect of rendering vast chunks of the economy irrelevant. Because I doubt I'll EVER have to outsource any of my work ever again, and there are probably many business owners like me who are realizing this as they use ChatGPT. So yeah, it's good for us, but a huge net negative for many people. The "Darwinian" part of me thiks "meh, so what, survival of the fittest" and all that, but do I really want 60% of society to be unemployed? I don't, even for the most selfish reasons, that means there'll be an expansion of the tax base and I'll have to pay more taxes, let alone the complete dystopia of a whole society of people who feel they have no productive role in it.
