4o has caused a lot of harm, on this we can agree.
@garymarcus
-
Neurosymbolic Necessary but Not Sufficient, Says Marcus
By
–
neurosymbolic is necessary but not sufficient, see eg my 2020 arxiv The Next Decade in AI and the 2018 arxiv with innateness in the title
-
Marcus agrees on need for intrinsic spatial and temporal logic
By
–
oh i certainly agree on the need for intrinsic spatial and temporal logic, see eg my 2018 essay with innateness in the title, rebooting AI, etc
-
Mythos surpasses GPT 5.5 but raises security concerns
By
–
1. Agree with @scaling01: Mythos seems to be a better GPT 5.5 on many indicators.
2. Mythos is definitely a major call for vigilance regarding security, and will pose problems for real-world systems that are not well protected. As @scaling01 -
Gary Marcus: Deep Learning hitting a wall, possible Substack essay
By
–
almost all is clearly explained in my article Deep Learning is hitting a wall, in Nautilus: https://
nautil.us/deep-learning-
is-hitting-a-wall-238440
… but maybe i will go line by line in a substack essay? -
Gary Marcus highlights the irony of critics as the field evolves
By
–
It is truly astonishing to see the entire field evolving in the neurosymbolic and world model directions that I advocated in 2019 and 2020, and simultaneously to observe people claiming (inevitably without specific evidence) that I am still wrong.
-
Global cognitive models have finally come of age
By
–
six years ago world (cognitive) models were the centerpiece of my essay The Next Decade in AI.
— Gary Marcus (@GaryMarcus) 23 mai 2026
their time is finally coming. https://t.co/VdnPWajZckSix years ago, global (cognitive) models were at the heart of my essay *The Next Decade in AI*. Their time has finally come.
-
Anthropic rises fast while OpenAI loses lead and reputation
By
–
some yes some no. anthropic seems to be more efficient, and is rising fast; openAi has lost their lead and has developed a reputation problem.
-
Questioning the Generalization Capabilities of AI in Mathematics
By
–
can’t believe people assume that success on highly verifiable problems in math (where we don’t even know how many tests were performed and how many might have failed) iautomatically generalize to everything else when there is not a shred of evidence that they do.