//
sign in
Profile
by @danabra.mov
Profile
by @dansshadow.bsky.social
AviHandle
by @danabra.mov
AviHandle
by @dansshadow.bsky.social
ProfileHeader
by @dansshadow.bsky.social
ProfileHeader
by @danabra.mov
ProfileHeaderAlt
by @jakesimonds.com
ProfileMedia
by @danabra.mov
ProfilePlays
by @danabra.mov
ProfilePosts
by @danabra.mov
ProfilePosts
by @dansshadow.bsky.social
ProfileReplies
by @danabra.mov
Record
by @atsui.org
Skircle
by @danabra.mov
StreamPlacePlaylist
by @katherine.computer
+ new component
ProfilePosts









Loading...
There are, also contrary to some popular understandings, continuous empirical and conceptual debates about what exactly LLMs have "learned" (and this presumes "learning" is the right concept; kind of an unnoticed assumption in popular talk recently, minus LeCun et al.). It's not remotely settled.
Best time to do your taxes in the US is the first week of February
7h
13h
Vincent Carchidi
Vincent Carchidi
A lot hinges on the analogy, and I'm never entirely sure what people mean by it. You could get rid of the scams, slop, etc. that we deal with today, and what are you left with that has comparable effects to the industrial revolution? Talking only in terms of actually existing tech here.
7h
I also think we should be wary of the (sometimes implicit, sometimes explicit) idea that being bullish on this or that set of uses means buying into some very specific scientific or philosophical assumptions. Not good!
Custom feeds? Nah, I'm good. I prefer foraging for my follows like my ancestors.
This is actually not a shitpost, believe it or not. Get that shit done early next time.
Moderately bullish on the medium- and long-term, but I am not actually sure this is the way to look at what is called "AI" today. And anything that would cause this probably should not share the same name ("AI").
Like maybe tomorrow someone releases a new type of thing that cures cancer, but it isn't a neural network and isn't any of AI's earlier techniques, but still calls it "AI." In any event, "doing existing things faster" vs. "changing the game altogether" remains the relevant distinction IMHO.
Useful - responding appropriately to a user's intent, often at least - just does not necessarily entail the cognitive capacities we associate with human "understanding" (even if we cannot precisely define that term, it's not meaningless, and we rightly believe there is something to it).
13h
14h
Gotta start putting a "Not an Anti" when I say stuff like this, but I think - vaguely going off these results - there is a possibility that LLMs have learned something that isn't quite "language," despite appearances, and our ability to impose meaning on anything they output is essentially blinding.
7h
7h
13h
7h
Vincent Carchidi
13h
Vincent Carchidi
Vincent Carchidi
Vincent Carchidi
Vincent Carchidi
Vincent Carchidi
Vincent Carchidi
Vincent Carchidi