Mike Knoop
speaker
927 appearances
4 recordings
1 series
first heard Mar 2025
last heard 18 Nov
Mike Knoop’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 1 in all, peaking in Nov 2025 with 1.
Appearances
They have a big model as a component of the system, but they have some sort of
Sort of synthesis engine on top that's allowed allows them to like reuse their existing pre-training knowledge at test time to recombine the knowledge.
And that allows them to make significant progress towards ARC.
And we saw it with three in December, you know, score a pretty high degree of efficiency, solve uh ARC A which you want.
It's 75%.
So not quite at the 85%.
Um, and then you know, open AI tested an even higher uh compute a version of it that probably was probably a couple million dollars uh to test.
got eighty seven percent, I think.
crazy.
Um and so that's kind of where we we wrapped up uh last year.
Uh we've been part of work over the actual couple years afterwards.
Right now I think ARC AGI two that we just introduced this week on Monday uh is pretty direct evidence that we need new ideas still to get to AGI.
Um we shared I don't know if we can throw it up or we can put it in the show notes or something.
Sure.
Um, there's a chart I shared on Twitter that shows the scaling performance of even frontier AI systems like O and Pro and O three against the version one of the data set that we've been using for the last five years and then version two that we just uh introduced this week.
And basically it resets the baseline back to zero percent.
Language models, uh pure LLM systems are scoring like zero percent now uh again on uh on Arc B2.
Um single COT systems like R1 and O one score like one percent.
And uh the sort of estimates we have right now for the frontier systems that are actually adding pretty sophisticated, you know, uh synthesis engines on top, like O1 Pro and O three, single digit percentages.
on their efficient compute settings.
Showing 681–700 of 927 · page 35 of 47
← Previous
Next →