Woozle/LLM

my thoughts on LLM policy

Generally, I feel that the current LLM/"AI" technology (which I'll just call "Gen3 AI" or "G3AI" here) is pretty amazing and has enormous potential, but is also being deployed in horrible and abusive ways.

My preference would be to have a more fine-grained policy that specifically targets those abuses (allowing only G3AI whose ethicality can be demonstrated) but between the intense negative sentiment which exists around them and the present lack of adequate tools for reaching evidence-based consensus around complex issues, it feels necessary to use the blunt instrument of a blanket ban rather than taking more limited action.

Ethical Issues

The major problems with G3AI technology appear to be:

• theft of creative work: The data-set used for training of models often includes copyrighted work that is being used without permission or any attempt at compensation.
• massive resource consumption: despite the lack of a single instance of G3AI generating profit, there has been a huge rush to develop mindbogglingly enormous datacenters dedicated to Gen3 processing. Worse, these centers have been largely unregulated and are creating massive resource-drains (electricity and cooling-water) wherever they are built, not to mention pumping gigawatts of heat into the atmosphere of an already-overheating planet. To satisfy their unhinged need for electrical power, they have begun increasing the supply of unsustainable energy-sources (e.g. recommissioning nuclear power and fossil-fuel-burning plants, or even ordering new ones built), with absolutely zero apparent concern for the damage this will cause.
• use as a surveillance tool: G3AI provides an unprecedented ability to spy on the activities of enormous numbers of people, given its ability to recognize meaning in text, audio, and video feeds. It can sit quietly in the background, looking for whatever activity its controllers have configured it to watch for, and alert them only when that activity is detected. Unlike human observers, it can operate around the clock and does not experience fatigue.
• nonconsensual and inappropriate intrusion: The G3AI industry is actively pushing the technology into every corner of our existence, including many where it is inappropriate or even dangerous. People are being encouraged to trust its output (and many apparently do), even though it is well-known for making egregious errors worse than those likely to be made by humans. (This includes a lot of arguably-consensual misuse, which could probably be another entry in this list.)

Negative Sentiment

The "AI" industry has so far been absolutely oblivious to these ethical issues – even to the point of imagining a completely different set of ethical issues (unsupported by any evidence) to justify its continued abuses. This obliviousness and heartlessness by itself is enough to make people want to avoid the technology altogether.

On top of that, G3AI's largest users all seem to have in common the goal of unemploying the rest of humanity – and keeping the revenue for themselves. They talk enthusiastically about how "labor saving" the tech is, but then don't give a moment's consideration for what happens to the producers of that labor when they're no longer being paid for it. They appear to be fascistically devoted to a "might makes right" philosophy: if you can dominate and control, then you deserve to do so and you totally should. This is somehow "the greater good" (but that's a whole other discussion too).

In summary: people might be able to work around the ethical issues or tolerate honest errors of judgement around G3AI if its primary proponents weren't being such utter entitled assholes and powermongers about it. Instead, we are all left with very sour feelings about it and many people don't want to touch it with a 40-meter pole. G3AI has been abusive to them, and nobody wants to be around their abuser.