Gary Marcus

Rank 12 of 47
|
Score 83
5/18/2025, 1:42:20 PM
X
In reply to:
Dagan Shani
@daganshani1
路
248d
@GaryMarcus @HermopolisPrime @DKokotajlo And also as far as the x risk goes, Daniel thinks that this is a very plausible scenario, even by 2027 if I understood him right(now he updated the AGI possibility to 2028), while you talk about catastrophic risk, not total human inhalation.
Gary Marcus
@GaryMarcus
路
248d
@daganshani1 @HermopolisPrime @DKokotajlo Llms might be a component but certainly not the whole answer.

Another difference is I think there is a huge range of unsolved problems that can鈥檛 plausibly be solved by 2027.
Dagan Shani
@daganshani1
路
248d
@HermopolisPrime @GaryMarcus @DKokotajlo I think that the main difference is that Kokotajilo's timeline for AGI shows that he thinks AGI would come out of LLMs, while Marcus pretty much denys the possibility of AGI coming from LLMs.
Hermopolis Prime 馃嚦馃嚳
@HermopolisPrime
路
248d
@GaryMarcus @daganshani1 @DKokotajlo I don't think the disagreement is as great as it first appears. I mean they both think 'AGI' is a valid concept, and could appear in the near-term future, but they do differ in terms of the timeline and how much intelligence is in current systems perhaps
Gary Marcus
@GaryMarcus
路
248d
@daganshani1 @DKokotajlo I like him and would be glad to do a podcast discussing with him?
Dagan Shani
@daganshani1
路
248d
It's really perplexing when you have smart, well-intentioned, not profit-motivated AI experts like @GaryMarcus and @DKokotajlo who have such polar opposite positions about AGI.

The statement is a lighthearted correction of a typo in a discussion about AGI timelines and risks. It does not engage substantively with the public issue being discussed, which is the potential risks and timelines associated with AGI development.