I am glad I have been useful to your considerations, providing links you didn't run into elsewhere. That is all I seek to achieve with my posts and comments, to support the understanding people achieve of the important issues we face. You are kind to consider my words and include me in the conversation.

There is a point I have tried to make about AI that seems to be neglected. These meltdowns and 'hallucinations' aren't potential to LLM's. Hard drives don't just hallucinate new directories when they go bad. If your hard drive goes bad and starts creating new folders and files on your computer, you know that someone has programmed it to do that. There's no failure mode that enables it to create new volumes of information.

Just so with LLM's that are devices supplied with algorithms with which to weight text from samples they have been trained on. The meltdowns and hallucinations are being misrepresented as some kind of failure mode, but that's not what they are. These malevolent attacks and creation of false information has been programmed into them, or is being added on the fly during their operations.

I think this is extremely important to understand, because what it reveals is the malice of the programmers, the companies that are releasing these products. Hard drives aren't evil. AI isn't evil. If these products are doing evil, they have those features built in at the factory, so to speak, and we can observe the producers are evil.

What it reveals is that these products are weapons used to attack us, and those companies are our enemies.

That is why I have sought links to AI that can be run by users, rather than accessed through their corporate platforms, because we won't turn AI into weapons that do us harm, and can enable AI to provide the full benefits to us it's capable of.

Thanks!