Gary Marcus

Rank 12 of 47
|
Score 83
12/23/2024, 1:33:27 PM
X
In reply to:
Petr Baudis
@xpasky
·
394d
@GaryMarcus @iruletheworldmo o3 is likely still just a fun prototype not useful for much autonomous work.

It shoukd make a damn powerful junior SWE or research math assistant, though. Which should qualify into your definition.

(But I bet Anthropic will present a more practical alternative soon.)
Gary Marcus
@GaryMarcus
·
394d
@iruletheworldmo When o3 is fully vetted by the community it will be clear that it is far from capable of completing “any intellectually knowledge sector work.”

(And also ridiculously expensive and damaging to the environment.)
🍓🍓🍓
@iruletheworldmo
·
394d
@GaryMarcus I'm allowed to have my own definition for agi. for me, it doesn't need to be embodied and sama's descriptions are closer to asi. likely to avoid breaking his deal with Microsoft.

something that can complete any intellectual knowledge sector work, where you only need a mouse and…
Gary Marcus
@GaryMarcus
·
394d
@iruletheworldmo Um (a) you are just an anonymous influencer and (b) you didn’t give a definition or argue for it whereas mine (see dear Elon musk essay) was derived from those who originated the term.

The point of the mountain biking isn’t the balancing, it’s the dynamic, intelligent…
🍓🍓🍓
@iruletheworldmo
·
394d
@GaryMarcus o3 is agi. not your mountain bike definition of agi. but my own definition.

in many ways it's closer to asi. you downplaying the intelligence of these systems plays into his hands so. not particularly concerned.
Gary Marcus
@GaryMarcus
·
394d
But Gary, nobody is saying o3 has anything to do with AGI.

The statement is a recommendation for using Wolfram for math, in the context of a discussion about AI capabilities. It does not substantively engage with public issues or contribute to civic dialogue, as it is more of a personal suggestion within a technical conversation.