Gary Marcus

Rank 11 of 47
|
Score 114
6/28/2025, 10:21:43 PM
X
In reply to:
Lawrence Chan
@justanotherlaw
·
47d
@GaryMarcus @ohabryka Wait sorry, I got the kindle book which doesn’t have the same pages as the physical one. Until where should I read?
Lawrence Chan
@justanotherlaw
·
47d
@GaryMarcus @ohabryka I’ve purchased the book and will read it soon.
Gary Marcus
@GaryMarcus
·
47d
@justanotherlaw @ohabryka no, not from my actual writing. from caricatures.

read the first 10 pages of taming silicon valley, and try to understand the relation between what i write and what people attribute to me.

and then i am happy to chat more; i see you sincere effort.
Lawrence Chan
@justanotherlaw
·
47d
@GaryMarcus @ohabryka I’m aware! I think it’s quite unfair that people have reduced your overall attitude of LLMs to “haha boo LLMs”, and assume that any LM advance contradicts your beliefs. I’m just noting that this is indeed the impression that people get from your writing/interviews.
Gary Marcus
@GaryMarcus
·
47d
@justanotherlaw @ohabryka we can continue another time or another venue. but know that i often say stuff like this
Gary Marcus
@GaryMarcus
·
238d
Total strawman.

I never said “AI is useless” (e.g., I love web search, GPS route planning, web search, recommendation engines and computational photography — and use all of those daily.)

I never even said LLMs are useless; I often say they are useful for brainstorming and
Lawrence Chan
@justanotherlaw
·
47d
@GaryMarcus @ohabryka I’m sure you haven’t said that LLMs won’t be useful, literally! You’ve just repeatedly and insistently highlighted specific problems with every LM generation, most of which were improved upon by the next.

Again, I think there are good criticisms you make, eg that o3 is not “PhD

The statement is part of a conversation about reading a book related to AI and its criticisms, but it does not substantively engage with public issues or contribute to civic dialogue. It is more about coordinating a reading schedule among participants.