Comment by sixhobbits
1 day ago
(author here) I think there's a difference between "I'm no longer impressed" (good) and "I was never impressed and never would have been impressed" (bad, but common).
Yes it's easy now so its by definition no longer impressive, but that in itself is impressive if you can correctly remember or imagine what your reaction _would_ have been 6 months ago.
Never impressed, no longer impressed, feeling depressed ... Another option, newly impressed by the next iteration.
Up to a point these have been probability machines. There's probably a lot of code that does certain likely things. An almost astonishing amount doing the same things, in fact. As such, perhaps we shouldn't be surprised or impressed by the stochastic parrot aspect any more than we're impressed by 80% of such sites being copy pasta from Stack Overflow a few years ago.
However, what we perhaps didn't expect is that on the margins of the mass probability space, there are any number of less common things, yet still enough of those in aggregate that these tools can guess well how to do those things too, even things that we might not be able to search for. Same reason Perplexity has a business model when Google or DDG exist.
And now, recently, many didn't expect one might be able to simulate a tiny "society of mind" made of "agents" out of these parrots, a tiny society that's proving actually useful.
Parrots themselves still impress me, but a society of them making plans at our beck and call? That can keep us all peeking, pecking, and poking for a while yet.
// given enough time and typewriters, who wins: a million monkeys, a society of parrots, or six hobbits?