A flagrant violation requires clearly established jurisprudence. What is considered fair use has always been nebulous, and to the extent that there are standards on it, it's mainly through established precedence, of which there is none on AI training.
One can argue that the closest thing to an AI's training on a book is a person reading a book and remembering the details, which i'm sure you reject, but it's not prima facie unreasonable.
you are exactly proving the point, the protests seem engineered to make people think they are doing something tantamount to ousting communist dictators when they really don't seem to be doing much at all except maybe annoying the moderates already on their side
I don't know if that's true. Protests - or in fact, any large scale ability to organize people - have the potential for action baked in. Remember that the fall of the USSR started with some relatively mild protests.
It wasn't long before there were Zomo's in the streets of Gdansk, Poznan, Warsaw and many other places to try to curb those protests and that in turn is what led to the fall of the Polish regime.
Yeah and nothing convinces the undecided independents, who now are mostly paranoid schizophrenics, to vote for you like appearing to shut down free assembly
No one cares about these protests who isn't already voting blue. If anything it alienates the majority of people who find endless futile protests annoying. Shutting down the account intentionally is so extraordinarily stupid politically that it seems pretty plausible that it was algorithmically flagged.
Trump has done many genuinely scary things but shutting down a social media account then turning it back on 90 minutes later is not exactly sending tanks into Tiananmen Square.
LLMs don't seem to be really that good at managing complex and conflicting needs, especially if it grows in increments, a ticket here and there. (To be fair people often suck at it too.)
I have a theory that mess in an LLM-generated codebase sort of grows exponentially. Each hack or sloppy abstraction chips away at the coherence of the project as a whole and makes the next feature harder to do. But perhaps more importantly, it demonstrates to the LLM, which is a probabilistic machine despite its impressive capabilities, that quality of the code is low, and it generates more low-quality hacks.
And eventually you reach knots upon knots that you have to untangle yourself because the LLM becomes too confused by the code it wrote itself. As models get better, the problem seems to be getting better. I'm needing to do this more and more rarely, and the tradeoff is still clearly worth it.
But I've seen Fable 5.1 absolutely lobotomized by someone who fills each chat context with half a million tokens of insipid garbage, and I don't see why this couldn't happen when coding as well. I'm not sure a solution to this general problem that LLMs can simulate idiots really well will be fixed any time soon.
reply