>Yes, but who’s to say the barrier of cost will be taken down? It’s not as easily reproducible as you say, training it takes tons of data and computing power etc.

It's easy enough, we have for example about a dozen open source language models that we can run, and the quality of them improves even as their size decreases (Flan-T5 is a good example). We can also see that with image models - cheaper, faster, and more variety. We already have quite a few different models, not including fine tuned off shoots.

And even for new models today, the costs are measured in the hundreds of thousands to millions. Those costs, while not cheap enough for me to build, are still incredibly cheap. And it will get cheaper to build the same models as techniques improve and hardware improves, there's no reason that trend wouldn't continue.

> For all we know if things like posthumanism etc. become real they might as well just charge an unreasonable price and only the select few will get it, leading to a gattaca-like scenario in real life. These sorts of dystopias are a genuine possibility with the advancements we’re seeing.

But that's a fear based conclusion, not something you are really coming to from an informed place. This isn't how technology has worked so far, and technology has made us all more "powerful", the internet, smartphones, and now these models. Why assume that at some arbitrary point this will suddenly no longer be possible? Why assume that the world is filled with mustache twirling rich villains?