I think we’re in a time where generative AI is at peak accessibility for a long time tbh. The underlying economics do not make sense. This is also a somewhat related topic that ties into my frustrations on it, but like… it can’t actually get much better or widespread than it is right now without fundamental changes to how it works and they’re still using algorithms developed in the 70s so I don’t super expect that innovation to be right around the corner at this point. It costs enormous, enormous amounts of time and money to train the models. Billions and billions of dollars. It takes an enormous amount of computing power to run them and water to cool them down. AI singlehandedly reversed trends in fossil fuel power plants. The data set uses basically everything they could get their hands on, even if they couldn’t legally use it, and people are talking about “just wait until it gets better”.
They’re having trouble making money right now with AI, and there isn’t a whole lot of novel data left to scrape. They need an order of magnitude of more data to make significant improvements which would mean more time labeling the data and curating it, and then time to train it, and all of this will cost billions and billions of more dollars (and again a lot of the people doing the labeling and such are being paid slave wages). Like we can use ChatGPT and similar things for free because someone is bankrolling this expecting future profits. The barrier to entry must go up because I know OpenAI isn’t turning a profit and I’d be surprised if any of these companies are. Idk, just seems like it’s all a bubble and it’s not feasible to expand or improve upon it because of we’re already having to build power plants and drain water reservoirs to keep up with the “not economically viable” stage. The cost must go up and the rate of improvement will begin to plateau.