Empirically: training improves the outcomes it doesn't change the capabilities.
ergo: we havent advanced much past translations. since that's empirically what llms do they translate questions into answers. I'm done responding to you its a waste of time talking to a wall.
i build / work on llms. they literally use transformers and attention. we're literally not that far past it. We're basically just optimizing the code paths. All the 'advances' the last 5 years have been routing, context management, and reducing bit size.
You're confusing time with actual progress. the stupidest kind of incorrect: confidently.
that midpoint just the fallout of using transformers. we're really not that far past it.
yes, we're talking about different things.
never said it would. I'm asserting that by the time AI gets to the point where its as useful as humans, it'll inherently be as deserving of autonomy and freedom as we are. which brings you back to being required to pay a living wage.
and what will likely happen is when they finally develop it, suddenly that thing deserves to not be exploited.
very rose colored glasses. the extra work to identify such sites is trivial less than even the hashing approach anubis uses.
risks are minimal for data pipelining on the training side. you can bootstrap a classifier that errors towards reject to sweep 99% of the weirdness this thing is doing in a few days. we already have clean datasets we can use to baseline such systems.
on top of that its fairly easy to detect if a run is failing due to collapse. the capital costs are small.
Same, and next year when bcachefs has nailed down the kinks it'll be basically bullet proof with snapshotting prior to its daily updates.
pointless and easily worked around. these fonts would have a unique signature that can be detected and decoded same way the browser is doing it. basically its relying on obscurity to work.
legit though, just make eclipses holidays and the problem is solved.