Gary Marcus

55.7K posts

Gary Marcus banner
Gary Marcus

Gary Marcus

@GaryMarcus

“In the aftermath of GPT-5’s launch … the views of critics like Marcus seem increasingly moderate.” —@newyorker

Katılım Aralık 2010
6.9K Takip Edilen213.2K Takipçiler
Sabitlenmiş Tweet
Gary Marcus
Gary Marcus@GaryMarcus·
Three thoughts on what really matters: 1. Fuck cancer 2. Friends are irreplaceable 3. The new "Marcus test" for AI is when AI makes a significant dent on cancer May that happen sooner, much sooner, rather than later. In memory of my childhood friend Paul.
English
136
139
2.3K
374K
Gary Marcus
Gary Marcus@GaryMarcus·
@AVMiceliBarone it’s not an intelligence test; it’s a way of deconfounding memorization
English
1
0
0
29
Antonio Valerio Miceli Barone
Antonio Valerio Miceli Barone@AVMiceliBarone·
@GaryMarcus It's not a fair evaluation, the task is too hard. How many humans can write a program that counts primes in Brainfuck? Can you?
English
1
0
0
32
Gary Marcus
Gary Marcus@GaryMarcus·
Pretty shocking result (that once again confirms what I wrote about the perils of distribution shift, 25 years ago): Translate coding benchmarks into languages LLMs can’t memorize and performance utterly falls apart.
Lossfunk@lossfunk

🚨 Shocking: Frontier LLMs score 85-95% on standard coding benchmarks. We gave them equivalent problems in languages they couldn't have memorized. They collapsed to 0-11%. Presenting EsoLang-Bench. Accepted to the Logical Reasoning and ICBINB workshops at ICLR 2026 🧵

English
31
41
287
28.4K
Gary Marcus retweetledi
Dirty Indy 🟥🟧🟨
Dirty Indy 🟥🟧🟨@cobracommanduhr·
maybe his credibility comes from his creating and selling a machine learning startup to Uber, or his senate testimony, or his six books, or the work where he anticipated the hallucinations and the problems of distribution shifts decades ago? It’s not that hard to research and I have seen multiple times people filling out Marcus apology forms and seems like you just might be doing that if pause gets green-lit. (Just my opinion as an independent wall st analyst) @kevinroose
English
1
1
14
5K
Gary Marcus
Gary Marcus@GaryMarcus·
Calling for a 6-month pause on AI journalism until we can realize that @kevinroose is not a credible journalist. An independent survey showed that 90%+ of my technical predictions are correct. How is that not credible? Calling on journalists to dismiss me is completely unprofessional.
Gary Marcus tweet media
English
12
12
91
9.3K
Amanda Askell
Amanda Askell@AmandaAskell·
Perhaps I should get married again so that the media has a more recent man they can reference any time they mention me or my work.
English
246
70
2.6K
278K
Scott Santens
Scott Santens@scottsantens·
@unusual_whales @vidadelunicorn Yes, but it does appear to be enabling at least partial replacement of entry level workers. Those in their early 20s are looking at a jobs market that looks like the Great Recession. Employers are looking to use AI before hiring.
Scott Santens tweet media
English
3
11
69
4.6K
unusual_whales
unusual_whales@unusual_whales·
"Massive investment in AI contributed basically zero to US economic growth last year," per Goldman Sachs
English
632
7.1K
44.2K
4M
Gary Marcus
Gary Marcus@GaryMarcus·
Scoop below. Get used to this kind of story. And get used to have your personal data compromised. Amazon last week; Meta this week. Not even the biggest companies can really handle the consequences of AI agents.
Jyoti Mann@jyoti_mann1

🚨Scoop: A rogue AI agent recently triggered a major security alert at Meta, by taking action without approval that led to the exposure of sensitive company and user data to Meta employees who didn't have authorization to access the data.

English
18
56
177
13.2K
Gary Marcus
Gary Marcus@GaryMarcus·
@chatgpt21 i really don’t take benchmarks seriously in this era
English
0
0
0
26
Chris
Chris@chatgpt21·
Arc AGI 1-2 have steadily shown that ai is getting better at reasoning. Do you think LLMS haven’t gotten better at reasoning. I understand it’s not parametric learning. But your original point can be demonstrated false by epochs math benchmark, swe pro, terminal bench, arc AGI 2 etc.
English
1
0
0
39
Chris Huber
Chris Huber@chubes4·
@GaryMarcus I still think your laughing emoji reply will age like milk. Because my opinion is still the same as my initial reply. For the record I agree with you that scaling alone isn't going to solve it. It will be scaling, context engineering, RL, and human creativity.
English
1
0
0
25
Gary Marcus
Gary Marcus@GaryMarcus·
@chubes4 then maybe you wanna retract your spoiled milk comment. but whatever.
English
1
0
0
37
Chris Huber
Chris Huber@chubes4·
@GaryMarcus I'm not in the business of making tech predictions, I stated my opinion. I think we can create something very close to AGI right now by combining the methods I described. I don't need you to agree with me. I also don't really appreciate the condescending tone. But whatever.
English
1
0
0
22
Vishnu
Vishnu@vishivishx·
@GaryMarcus But they're not just scaling anymore, right? Aren't they now improving the tooling layer and making this system more and more "Neurosymbolic" by strengthening the deterministic components of their Harness+LLM Agent Hybrids?
English
2
0
1
93
Gary Marcus
Gary Marcus@GaryMarcus·
@chubes4 my tech predictions over the last year have (per an independent survey) been 90% correct. how about yours?
English
1
0
0
40
Chris Huber
Chris Huber@chubes4·
@GaryMarcus AGI is going to be solved with context engineering and reinforcement learning.
English
1
0
0
34
Gary Marcus
Gary Marcus@GaryMarcus·
@chatgpt21 now plot losses, which have grown even faster but anyway that is $, not AGI, of which there is nothing to plot, because there is none. (just benchmarks, which often appear to be gamed).
English
1
0
0
36
Chris
Chris@chatgpt21·
@GaryMarcus “So far the results have not been promising”
Chris tweet media
English
1
0
2
38

Keşfet

@AVMiceliBarone @kevinroose @AmandaAskell @TechToonDaily @nathanhamiel @scottsantens @unusual_whales @vidadelunicorn