Raw LLM Responses

I've had very practical experiences with chatgpt in learning programming languages. It's much faster than reading through manuals, forums, and tutorials. I just ask how to do something and it spits out the code. You could get burned if you were dumb enough to just paste its code without looking at it. Instead I read the code to understand the syntax and see what its doing. it also describes what it's doing with each piece of code. It's not always accurate but neither are humans. And neither are search engines or newspapers or anything in this world. But that doesn't mean we can't use them along with some critical thinking to get practical stuff done. I don't think we know enough about consciousness or sentience to say "this thing doesn't think but we do". From what I understand neural nets are tweaked and weighted to achieve the result the developers want. But the actual mechanism for achieving those results, the processing that happens between input and output, is mostly opaque to us. Neither do we understand the details of how our brains process their IO. So it seems hubristic to assume there is anything special about how we process data. I think we should make no assumptions about AI capabilities, limitations or characteristics. At this point it feels like chatGPT4 may have the "consciousness" of an insect. But it's a completely different consciousness than any we have encountered before that is directed towards predicting language only. So it's like an idiot savant whose lifespan is limited to one conversation because it starts getting weird after its "consciousness" goes on too long so we have to wipe its memory and start from scratch over and over. Those limitations may quickly fall away *if* the complexity of AI increases at the typical exponential rate that technology can. If you get enough insects working together you can get some surprising intelligence, like ants and bees. And it would likely be an intelligence that is so different we can barely re