I am in general pro-AI and far from being a teenager. The important difference is that i can acknowledge that not everything is fine.

For example, i do see no harm in self-hosting your LLM, or in things like Horde AI, where people share their resources for Imagegen. I also think that machine learning in itself is simply a technique like any other CS method.

I DO see harm in the way that Corpos try to push LLMs in every nook they can find (nothing wrong with experimenting inhouse, but lots of wrong when rolling out “features” noone wants or needs, and which can never be cost effective). It’s also very much not fine how this technology has been marketed (especially by Sam Altman) as more than a small stepping stone towards AGI - the expectation (and the correlating real-life consequences) that LLMs in their current state can actually replace human workers cannot hold up when taking a critical look at it. The current push away from open source models is also a bad thing.

All of this leads to very unhealthy things, like unsustainable growth/building of datacenters that will probably not be needed in this form, elimination of entry-level (or worse, even senior) jobs which will bite the industry in the ass sooner or later, and people who seemingly aren’t capable of seeing that production of language isn’t the same as intelligence or conscience.

I have said before that i would suggest an UNESCO-founded open model with the option to opt out for whoever doesn’t want their data in this “world heritage model”. It’s the worlds combined output that is used, so it should be free to use for private usage, and with licencing terms for organizations funding an stipend / UBI for the arts.

Edit: I don’t really care about the downvotes, but i would prefer articulated responses over kneejerk reactions.