Alexander C. Kaufman
@kaufman
Same, except I'm using Perplexity AI. It's been really helpful, but at the same time it's made me somewhat skeptical of the maximalist claims about AI. It's a great improvement on the certain internet tools I have relied on for most of my adult life. But I'm not yet seeing it's value as some all-encompassing replacement for things I think benefit from a human touch.
2 replies
1 recast
7 reactions
š_š
@m-j-r.eth
give it a year.
1 reply
0 recast
1 reaction
Alexander C. Kaufman
@kaufman
a year of using it or a year of seeing it expand its capabilities
1 reply
0 recast
0 reaction
š_š
@m-j-r.eth
good to do both, the benchmarks are gamed for announcement. and the top three corporate services are in modal collapse, disruption arrives from the other teams assembling whichever insights from research papers. have you checked out https://huggingface.co/papers? most of it isn't truly commercial in any sense, but functionally lags by 6 months at most.
2 replies
0 recast
1 reaction
Josh Linder
@jcl
What do you think of "Claude plays Pokemon" as a more objective ongoing assessment of where the models are at?
1 reply
0 recast
0 reaction
š_š
@m-j-r.eth
these still have a lot of interventions by their creators, so the models are okay, but they haven't quite picked up transferable heuristics. metaphorically, an acolyte will eventually force a more experienced figure to say "here, you're doing this wrong, watch me and I'll let you know if you fail, then repeat so I know it's not a fluke" but that's kind of a mouthful and everyone wants the statistics to scale until that deliberate approach is moot.
0 reply
0 recast
0 reaction