Claims that can’t be checked by Gary Marcus, page 8 of 24
Deadlines are judged as of 2026-10-29. About Gary Marcus and the sources covered.
Claims that can’t be checked by Gary Marcus (2353)
Public evidence cannot settle these claims, so they are recorded and attributed, not scored. Most were screened out: both AI providers read the claim on its own and found it a value judgment, rhetoric, hedged, too vague to test, or about something no public evidence could show. The rest were found to be recommendations or too open to test in a full review.
Page 8 of 24: claims 701 to 800 of 2353, ordered by the date first said, newest first. To find a claim by its words, use search.
Assessment: Can’t be checkedSaid
Observation
“But it’s all still just an extended form of mimicry, not something deeper, as the persistent failures with parts and wholes keep showing us.”
Our reading: Image and video generation systems like these are still just an extended form of mimicry, not something deeper, as their persistent failures with parts and wholes keep showing.
No horizon
Scope: these systems (image and video generators discussed)
Source: What scaling does and doesn’t buy you: peeling back the hype surrounding Google‘s trendy Nano Banana
Assessment: Can’t be checkedSaid
Forecast
“The Klarna Effect may come back to haunt them.”
Our reading: Employers replacing employees with AI may end up reversing course as Klarna did.
No horizon
Assessment: Can’t be checkedSaid
Conditional forecast
“If we stop hiring junior employees, we risk destroying the pipeline to expertis.”
Our reading: Stopping hiring of junior employees in favor of AI risks destroying the pipeline to expertise.
No horizon
Only if: if we stop hiring junior employees
Assessment: Can’t be checkedSaid
Observation
“Without saying for sure that I know what will happen next — after all things like stock prices and the valuations AI startups are as much a matter of unpredictable crowd psychology as they are of (in this case, shaky) technical and economic fundamentals — I will leave this here, a cartoon I made (with ChatGPT’s help!) in mid-July.”
Our reading: The technical and economic fundamentals behind AI startup valuations are shaky.
No horizon
Source: Is this the moment when the Generative AI bubble finally deflates?
Assessment: Can’t be checkedSaid
Rhetoric
“Not that Sam is deterred; in his mind, a fuckup is just an excuse to ask for money:”
Our reading: Sam Altman is not deterred by a failure; in his mind, a failure is just an excuse to ask for money.
No horizon
Source: Is this the moment when the Generative AI bubble finally deflates?
Assessment: Can’t be checkedSaid
Conditional forecast
“Once the markets really understand this, enthusiasm may indeed collapse fairly quickly.”
Our reading: Once the markets really understand the point made in Marcus's cartoon, enthusiasm for generative AI may collapse fairly quickly.
No horizon
Only if: once the markets really understand 'this' (the point of the cartoon he shared, whose contents are not given in the text)
Source: Is this the moment when the Generative AI bubble finally deflates?
Assessment: Can’t be checkedSaid
Observation
“The technical foundation was, as I argued all the way back to the early days of GPT-2, never strong enough to support the hype.”
Our reading: The technical foundation of LLMs has never been strong enough to support the hype.
No horizon
Source: Is this the moment when the Generative AI bubble finally deflates?
Assessment: Can’t be checkedSaid
Observation
“the user has the option to accept a code change or not, but with so many changes to make at such a rapid pace, a developer in hurrry (which is almost every developer) could easily miss an attack, and ok the change or run a command”
Our reading: Even with manual approval, hurried developers could easily approve malicious changes suggested by coding agents.
No horizon
Qualifier: even with Auto-Run off
Assessment: Can’t be checkedSaid
Forecast
“(Nvidia advised, rightly, that everyone should disable Auto-Run mode if they’ve activated it, but we fear that many users may use it anyway, because it is so fast.)”
Our reading: Many Cursor users may keep using Auto-Run mode despite the security risk because it is fast.
No horizon
Qualifier: because it is so fast
Assessment: Can’t be checkedSaid
Observation
“As long as we have agents roaming the internet and otherwise incorporating data that they don’t fully understand – and LLMs don’t ever fully understand the data they are leveraging – there is enormous risk.”
Our reading: LLMs never fully understand the data they leverage.
No horizon
Assessment: Can’t be checkedSaid
Forecast
“Given current implementations of the technology, it's hard to imagine enough patches on the planet to thwart them all.”
Our reading: Given current implementations, it is unlikely that patches can thwart all data-injection attacks on generative AI systems.
No horizon
Qualifier: given current implementations
Assessment: Can’t be checkedSaid
Conditional forecast
“Sure the advice is good, as far is it goes, but if the goods inside are valuable enough, motivated thieves may well still find a way.”
Our reading: Even with mitigations, motivated attackers may still compromise valuable systems using coding agents.
No horizon
Only if: even with mitigations
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“The best defense would be not using agentic coding altogether.”
Our reading: The best security defense is to not use agentic coding at all.
No horizon
Assessment: Can’t be checkedSaid
Forecast
“But the tools are so seductive that we doubt many developers will resist.”
Our reading: Few developers are likely to refrain from using agentic coding tools.
No horizon
Assessment: Can’t be checkedSaid
Rhetoric
“There is a real sense in which GPT-5 is starting to look Altman’s Waterloo.”
Our reading: GPT-5 is starting to look like Altman's decisive defeat.
No horizon
Assessment: Can’t be checkedSaid
Observation
“Two years ago, many treated Altman like an oracle; in the eyes of many, Altman now looks more like a snake oil salesman.”
Our reading: Two years ago many people treated Altman like an oracle, and now, in the eyes of many, he looks more like a snake oil salesman.
No horizon Evidence as of
Limitation: No poll or systematic measure of views of Altman in 2023 or in August 2025 was found.
Assessment: Can’t be checkedSaid
Conditional forecast
“As people begin to recognize that truth (the bitter lesson about limits of Sutton’s The Bitter Lesson), scientists will start to charts new paths. One of the new paths may even give rise to what we so desperately need: AI that we can trust.”
Our reading: As the limits of scaling become recognized, scientists will pursue new approaches, one of which may produce trustworthy AI.
No horizon
Only if: as people recognize the limits of scaling LLMs
Assessment: Can’t be checkedSaid
Rhetoric
“But media coverage is not science, and braggadacio alone cannot yield AGI.”
Our reading: Hype and bragging alone cannot produce AGI.
No horizon
Assessment: Can’t be checkedSaid
Causal claim
“And you can’t make AI better if you drown out the critics and keep pouring good money after bad.”
Our reading: AI cannot be made better if you drown out the critics and keep pouring good money after bad.
No horizon
Only if: if you drown out the critics and keep pouring good money after bad
Assessment: Can’t be checkedSaid
Observation
“The LLM itself still isn’t inducing the rules. And this is a very special-purpose hack, not a remotely general solution. (And, e.g., it would likely fall apart if you changed the rules to play an alternative version of chess.)”
Our reading: A big company's added python code for chess legality is a special-purpose hack, the LLM still doesn't induce the rules, and it would likely fail on chess variants.
No horizon
Assessment: Can’t be checkedSaid
Observation
“But no matter how much LLMs mimic the patterns of human language, they are not like us. They sound like us, but they don’t think like us.”
Our reading: LLMs do not think like humans, however well they mimic human language.
No horizon
Qualifier: no matter how much they mimic human language
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Always regard them like the weird function approximators they are; never trust them.”
Our reading: People should never trust LLMs and should treat them as weird function approximators.
No horizon
Assessment: Can’t be checkedSaid
Observation
“We still feel that way. And the field still hasn’t put the time in.”
Our reading: The AI field still has not put the time in on giving time, space, and causality a solid computational footing.
No horizon
Assessment: Can’t be checkedSaid
Observation
“That rumor is bullshit.”
Our reading: The rumor that LLMs are basically just like humans is false.
No horizon
Assessment: Can’t be checkedSaid
Forecast
“GPT-5 is unlikely to be ahead of the pack for more than a couple months.”
Our reading: GPT-5 is unlikely to remain ahead of competing models for more than a couple of months.
No horizon
Source: GPT-5: Overdue, overhyped and underwhelming. And that’s not the worst of it.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Certainly, in a rational world, their valuation would take a hit.”
Our reading: In a rational world, OpenAI's valuation would certainly take a hit.
No horizon
Qualifier: in a rational world
Source: GPT-5: Overdue, overhyped and underwhelming. And that’s not the worst of it.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“By rights, Altman’s reputation should by now be completely burned.”
Our reading: Sam Altman's reputation should by now be completely burned given his AGI claims.
No horizon
Source: GPT-5: Overdue, overhyped and underwhelming. And that’s not the worst of it.
Assessment: Can’t be checkedSaid
Rhetoric
“In hindsight, that was all bullshit.”
Our reading: In hindsight, Altman's statements (his joke that AGI had been achieved internally, his claim that OpenAI knows how to build AGI, and his claim that talking to GPT-5 is like talking to a PhD-level expert) were all bullshit.
No horizon
Qualifier: in hindsight
Source: GPT-5: Overdue, overhyped and underwhelming. And that’s not the worst of it.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“After half a trillion dollars in that direction, it is obviously time to move on. The disappointing performance of GPT-5 should make that enormously clear.”
Our reading: After half a trillion dollars spent on scaling, the field should move on from the pure-scaling hypothesis.
No horizon
Source: GPT-5: Overdue, overhyped and underwhelming. And that’s not the worst of it.
Assessment: Can’t be checkedSaid
Causal claim
“And the worst part? Altman brought it all on himself. Had he not kept hinting at the moon, people might have been fine with just another incremental update.”
Our reading: Altman's overhyping caused the GPT-5 backlash; without it people might have accepted an incremental update.
No horizon
Source: GPT-5: Overdue, overhyped and underwhelming. And that’s not the worst of it.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“All I am saying is give neurosymbolic AI with explicit world models a chance.”
Our reading: The field should give neurosymbolic AI with explicit world models a chance.
No horizon
Source: GPT-5: Overdue, overhyped and underwhelming. And that’s not the worst of it.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“For all that, GPT-5 is not a terrible model.”
Our reading: GPT-5 is not a terrible model.
No horizon
Source: GPT-5: Overdue, overhyped and underwhelming. And that’s not the worst of it.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Nobody with intellectual integrity should still believe that pure scaling will get us to AGI.”
Our reading: No intellectually honest person should still believe pure scaling will reach AGI.
No horizon
Source: GPT-5: Overdue, overhyped and underwhelming. And that’s not the worst of it.
Assessment: Can’t be checkedSaid
Rhetoric
“We have been fed a steady diet of bullshit for the last several years.”
Our reading: Over the last several years, we have been fed a steady diet of bullshit about AI.
No horizon
Source: GPT-5: Overdue, overhyped and underwhelming. And that’s not the worst of it.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“He may not even be the right CEO for OpenAI anymore:”
Our reading: Sam Altman may no longer be the right CEO for OpenAI.
No horizon
Source: GPT-5: Overdue, overhyped and underwhelming. And that’s not the worst of it.
Assessment: Can’t be checkedSaid
Observation
“People are wising up that LLMs are not in fact AGI-adjacent.”
Our reading: LLMs are not close to AGI, and people are increasingly realizing this.
No horizon
Source: GPT-5: Overdue, overhyped and underwhelming. And that’s not the worst of it.
Assessment: Can’t be checkedSaid
Forecast
“Pricing is good, but profits may continue to be elusive; still no clear technical moat.”
Our reading: Although GPT-5 pricing is good, OpenAI's profits may continue to be elusive.
No horizon
Source: GPT-5 hot take
Assessment: Can’t be checkedSaid
Forecast
““AGI 2027” seems more and more remote by the day.”
Our reading: AGI arriving by 2027 looks increasingly unlikely.
No horizon
Source: GPT-5 hot take
Assessment: Can’t be checkedSaid
Causal claim
“Gary Marcus thinks that perhaps scaling is not in fact all you need.”
Our reading: Gary Marcus thinks that perhaps scaling is not in fact all you need.
No horizon
Qualifier: perhaps
Source: GPT-5 hot take
Assessment: Can’t be checkedSaid
Forecast
“Cesspools of automatically-generated fake websites, rather than ChatGPT search, may ultimately come to be the single biggest threat that Google ever faces. After all, if users are left sifting through sewers full of useless misinformation, the value of search would go to zero—potentially killing the company.”
Our reading: Cesspools of automatically-generated fake websites, rather than ChatGPT search, may ultimately become the single biggest threat Google ever faces, because if users are left sifting through sewers of useless misinformation the value of search would go to zero, potentially killing the company.
No horizon
Scope: rather than ChatGPT search
Source: Slopocalypse Now
Assessment: Can’t be checkedSaid
Observation
“Current systems can mimic the kinds of words people use in completing tasks, often in contextually relevant ways, but that doesn’t really mean that they understand the things that they are doing; they have no concept of what it means to delete a database or make up a fictitious calendar entry.”
Our reading: Current AI systems do not understand what they are doing and have no concept of actions like deleting a database.
No horizon
Assessment: Can’t be checkedSaid
Observation
“What drives LLMs (and what drives the current batch of agents) is mimicry, and not what Ernest Davis and I called in 2019 deep understanding.”
Our reading: LLMs and current agents are driven by mimicry rather than deep understanding.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“An important, publicly-visible system that can be beat in .001% of the times it is attacked is a mess.)”
Our reading: An important, publicly visible system that can be beaten in .001% of the times it is attacked is a mess.
No horizon
Scope: important, publicly-visible system
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Venture capitalists style themselves as brave funders of innovation, but in the current climate, they are anything but.”
Our reading: Venture capitalists style themselves as brave funders of innovation, but in the current climate they are anything but brave funders of innovation.
No horizon
Qualifier: in the current climate
Assessment: Can’t be checkedSaid
Forecast
“(In fairness, ChatGPT agent wasn’t yet out when Penrose ran the test, but I would be surprised if their results were wildly different).”
Our reading: ChatGPT agent is unlikely to produce wildly different results on Penrose's account-balance benchmark.
No horizon
Assessment: Can’t be checkedSaid
Forecast
“But I seriously doubt that LLMs will ever yield the substrate we need.”
Our reading: LLMs are unlikely ever to provide the substrate needed for working AI agents.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“In a rational world, generative AI would be penalized for failing to meet expectations around agents; in the actual world, investors and big tech companies are continuing to pour on the gas.”
Our reading: Generative AI should be penalized for failing to meet agent expectations, yet investors and big tech keep increasing investment.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Perhaps after enough failures of AI agents, which shouldn’t be trusted with your calendar let alone your credit cards, reality may finally start settle in.”
Our reading: AI agents should not be trusted with your calendar, let alone your credit cards.
No horizon
Assessment: Can’t be checkedSaid
Conditional forecast
“Perhaps after enough failures of AI agents, which shouldn’t be trusted with your calendar let alone your credit cards, reality may finally start settle in.”
Our reading: Perhaps, after enough failures of AI agents, reality may finally start to settle in.
No horizon
Only if: after enough failures of AI agents
Assessment: Can’t be checkedSaid
Forecast
“the “demo to reality” issue they mention is exactly what we should expect — probably for a years to come:”
Our reading: The gap between AI agent demos and real-world performance will probably persist for years to come.
No horizon
Qualifier: probably
Assessment: Can’t be checkedSaid
Forecast
“found a result that I suspect will be typical: AI errors tend to compound over time.”
Our reading: He suspects it will be typical that AI agent errors compound over time, as in Penrose's account-balance benchmark.
No horizon
Qualifier: I suspect
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Yet more good money is being poured after bad. Staggeringly large amounts of money.”
Our reading: Yet more good money, in staggeringly large amounts, is being poured after bad into generative AI.
No horizon
Assessment: Can’t be checkedSaid
Observation
“The “pushing the boundaries of what’s possible” that Roose promised us? Largely absent.”
Our reading: AI agents in 2025 have largely not delivered the boundary-pushing capabilities that Kevin Roose promised.
No horizon
Assessment: Can’t be checkedSaid
Forecast
“Someday, I predict, the failure of current agents to produce much of lasting value will be seen as a poignant reflection of a massive mistake, intellectually and economically speaking: the mistake of investing almost solely in LLMs, as a kind of hopeful shortcut.”
Our reading: Someday, current agents' failure to produce much of lasting value will be seen as a poignant reflection of a massive intellectual and economic mistake: investing almost solely in LLMs as a kind of hopeful shortcut.
No horizon
Qualifier: someday
Assessment: Can’t be checkedSaid
Conditional forecast
“When AGI (artificial intelligence) comes, large language models like GPT-5 may be seen in hindsight as part of the eventual solution, but only as part of the solution.”
Our reading: When AGI comes, large language models like GPT-5 may be seen in hindsight as part of the eventual solution, but only as part of it.
No horizon
Only if: when AGI comes
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“If all seven predictions prove correct, I hope that the field will finally realize that it is time to move on.”
Our reading: If his seven predictions prove correct, the AI field should move on from the LLM approach.
No horizon
Qualifier: if all seven predictions prove correct
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Absent any information of that kind, there is little reason to take on faith that whatever unknown techniques allowed them to do so well on IMO problems will generalize to any other kind of problem.”
Our reading: Absent disclosure, there is little reason to assume the techniques behind the IMO results will generalize to other kinds of problems.
No horizon
Qualifier: absent information explaining why they work
Source: DeepMind and OpenAI achieve IMO Gold. What does it all mean?
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Without question, these are impressive accomplishments.”
Our reading: The results from OpenAI and DeepMind, whose AI systems reached gold-medal level at the IMO, are without question impressive accomplishments.
No horizon
Qualifier: without question
Source: DeepMind and OpenAI achieve IMO Gold. What does it all mean?
Assessment: Can’t be checkedSaid
Observation
“The specific skills tested by IMO problems are sometimes useful in pursuing original mathematics, but they are far from the most important skills in original math research.”
Our reading: The specific skills tested by IMO problems are sometimes useful in pursuing original mathematics, but they are far from the most important skills in original math research.
No horizon
Qualifier: sometimes useful in pursuing original mathematics
Source: DeepMind and OpenAI achieve IMO Gold. What does it all mean?
Assessment: Can’t be checkedSaid
Rhetoric
“Overall, the new work certainly might be exciting. Until outside scientists get to dig for a serious review we won’t really know for sure what it all means.”
Our reading: The new work certainly might be exciting, but we won't really know for sure what it all means until outside scientists get to do a serious review.
No horizon
Qualifier: until outside scientists get to dig for a serious review
Source: DeepMind and OpenAI achieve IMO Gold. What does it all mean?
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“That’s simply preposterous. Being part of a 45-way tie for 27th place in a math contest for elite high school students is really not the kind of historic event that the moon landing was.”
Our reading: Bubeck's suggestion that the result may end up looking like a moon-landing moment for AI is simply preposterous, because being part of a 45-way tie for 27th place in a math contest for elite high school students is really not the kind of historic event the moon landing was.
No horizon
Scope: responding to Bubeck's remark that it 'may end up looking like a moon-landing moment'
Source: DeepMind and OpenAI achieve IMO Gold. What does it all mean?
Assessment: Can’t be checkedSaid
Observation
“Our point is just that the scratchy style is completely atypical of LLMs and suggests strongly that OpenAI-IMO is not simply a standard LLM, whatever it is.”
Our reading: OpenAI-IMO's atypical scratchy proof style strongly suggests it is not simply a standard LLM.
No horizon
Source: DeepMind and OpenAI achieve IMO Gold. What does it all mean?
Assessment: Can’t be checkedSaid
Observation
“This difference may reflect a limit in the OpenAI-IMO’s understanding. It certainly suggests that OpenAI-IMO would not make a very effective expositor of mathematical results.”
Our reading: The difference in OpenAI-IMO's answers (longer, algebraic, less well structured) may reflect a limit in its understanding, and certainly suggests that OpenAI-IMO would not make a very effective expositor of mathematical results.
No horizon
Qualifier: may reflect
Source: DeepMind and OpenAI achieve IMO Gold. What does it all mean?
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Awfully impressive. To be able to solve math problems at the level of the top 67 high school students in the world is to have really good math problem solving chops.”
Our reading: IMO gold-medal performance by an AI is awfully impressive, because solving math problems at the level of the world's top 67 high school students means having really good math problem-solving chops.
No horizon
Scope: at the level of the top 67 high school students in the world
Source: DeepMind and OpenAI achieve IMO Gold. What does it all mean?
Assessment: Can’t be checkedSaid
Rhetoric
“That said, some of the hype has reached completely absurd levels.”
Our reading: Some of the hype around the AI IMO gold results has reached completely absurd levels.
No horizon
Source: DeepMind and OpenAI achieve IMO Gold. What does it all mean?
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“We need serious regulation to prevent AI’s from running amok. Voluntary agreements between companies are not going to cut it.”
Our reading: Serious AI regulation is needed because voluntary company agreements will not suffice.
No horizon
Assessment: Can’t be checkedSaid
Observation
“What crystallized for me over the last few days is that we have such a person.”
Our reading: Marcus says it became clear to him over the last few days that Elon Musk is a person who matches his list of things one might need to really screw up the planet.
No horizon
Qualifier: you might need something like the following
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Forcing unreliable AI everywhere in the decision-making chain is not necessarily something we should want.”
Our reading: We should be wary of forcing unreliable AI throughout decision-making chains.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Maybe someone will come to their senses and stop putting LLMs into military systems.”
Our reading: LLMs should not be put into military systems.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“It is genuinely scary that Elon Musk went from being one of the first industry leaders warning about AI risks to being the most reckless of the AI leaders.”
Our reading: It is genuinely scary that Elon Musk went from being one of the first industry leaders to warn about AI risks to being the most reckless of the AI leaders.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Way out of line is right.”
Our reading: Marcus agrees with Samuel Marks's thread that xAI's conduct in releasing Grok 4 without following the industry's safety conventions was way out of line.
No horizon
Assessment: Can’t be checkedSaid
Forecast
“And industry-standard measures are unlikely to be enough, in any event.”
Our reading: Industry-standard AI safety measures are unlikely to be sufficient.
No horizon
Assessment: Can’t be checkedSaid
Forecast
“I am at maybe 3% now, much higher than a month ago, contemplating what a wealthy, reckless megalomaniac might in the worst circumstances do, But I am still betting on Team Human for the foreseeable future.”
Our reading: Marcus puts the probability of AI-caused doom at about 3%, up sharply from a month earlier, while still betting humans prevail for the foreseeable future.
No horizon
Assessment: Can’t be checkedSaid
Forecast
“ButI have seen enough to realize that there is a real risk, especially in the current anti-regulatory regime, that some exceptionally powerful person, Elon or otherwise, unconstrained in conventional ways, with a reckless disregard for humanity, could accidentally launch and spread an AI that deliberately or otherwise causes “significant harm to the world”.”
Our reading: There is a real risk, especially under the current anti-regulatory regime, that some exceptionally powerful person (Elon or someone else), unconstrained in conventional ways and recklessly disregarding humanity, could accidentally launch and spread an AI that, deliberately or otherwise, causes 'significant harm to the world'.
No horizon
Qualifier: especially in the current anti-regulatory regime
Assessment: Can’t be checkedSaid
Forecast
“But even so, I think most scenarios (e.g., people homebrewing biological weapons) could eventually be stopped, perhaps causing a lot of damage but coming nowhere near to literally extinguishing humanity.”
Our reading: Most AI misuse scenarios by bad actors could eventually be stopped, perhaps causing much damage but not extinguishing humanity.
No horizon
Qualifier: most scenarios
Assessment: Can’t be checkedSaid
Conditional forecast
“Absent systems cognitively rich enough to represent and reason about moral principles, I don’t how this can ever work.”
Our reading: Marcus does not see how building alignment through training data alone can ever work unless systems are cognitively rich enough to represent and reason about moral principles.
No horizon
Only if: absent systems cognitively rich enough to represent and reason about moral principles
Assessment: Can’t be checkedSaid
Observation
“Part of my reasoning then was that actual malice on the part of AI was unlikely, at least any time soon.”
Our reading: As of May 2025, Marcus believed actual malice on the part of AI was unlikely any time soon.
No horizon
Qualifier: his reasoning as of May 2025
Assessment: Can’t be checkedSaid
Observation
“Worse, it is clear, even from Elon’s own posts on X, that the company is in over its head.”
Our reading: xAI is in over its head.
No horizon
Assessment: Can’t be checkedSaid
Observation
“The fact that xAI can’t get its house together in even basic ways is disturbing.”
Our reading: xAI cannot get its house in order even in basic ways.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“We need liability, auditing, standards of malpractice, international treaties, too.”
Our reading: AI requires liability, auditing, malpractice standards, and international treaties.
No horizon
Assessment: Can’t be checkedSaid
Forecast
“Drones and even nuclear weapons may eventually be under LLM command.”
Our reading: Drones and even nuclear weapons may eventually be placed under LLM command.
No horizon
Assessment: Can’t be checkedSaid
Forecast
“Getting to AGI will likely take still more breakthroughs.”
Our reading: Reaching AGI will likely require further breakthroughs.
No horizon
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Causal claim
“Largely because of Hinton’s influence, the field still talks about neurosymbolic AI as if it were a dirty word. That’s likely holding back the science.”
Our reading: The stigma against neurosymbolic AI, largely due to Hinton, is likely holding back AI science.
No horizon
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Observation
“Massive infrastructure investments like Stargate are likely based on misleading impressions of what is actually driving progress.”
Our reading: Massive AI infrastructure investments like Stargate are likely based on a mistaken view of what drives AI progress.
No horizon
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Conditional forecast
“Admitting that they need to rely on neurosymbolic tools would pierce the scaling narrative.”
Our reading: If AI companies admitted that they need to rely on neurosymbolic tools, that would pierce the scaling narrative.
No horizon
Only if: if companies admitted they need to rely on neurosymbolic tools
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Causal claim
“Hinton’s long-time hostility against any role at all for symbols has, in my judgement, cost the field dearly.”
Our reading: Hinton's hostility to symbolic approaches has substantially harmed AI research.
No horizon
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Observation
“Where the field is getting stuck now is likely on how to scale these new, neurosymbolic tools and how to make models “understand” when to use what and how to logically combine the context.”
Our reading: The field is likely stuck on scaling neurosymbolic tools and getting models to know when and how to use them.
No horizon
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Causal claim
“It would appear that it is only recently, in desperation, as scaling started to reach diminishing returns, that mainstream machine learning began to broaden its vistas.”
Our reading: It would appear that mainstream machine learning only recently began to broaden its vistas, in desperation, as scaling started to reach diminishing returns.
No horizon
Qualifier: it would appear
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Observation
“There is likely still a great deal of work left to be done.”
Our reading: There is likely still a great deal of work left to be done.
No horizon
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Observation
“In my view, neither neural networks nor classical AI can really stand on their own.”
Our reading: In Marcus's view, neither neural networks nor classical symbolic AI can really stand on their own.
No horizon
Qualifier: in my view
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Causal claim
“We are drowning in a sea of mediocre prose precisely because LLMs, the backbone of virtually all current sytems, lack the underlying representational wherewithal to do better.”
Our reading: LLMs produce mediocre prose because they lack the representational capacity to do better.
No horizon
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Observation
“GPT-4 probably wasn’t (I suspect it included some symbolic filters in its guardrails), and o3, when it invokes code interpreter and symbolically-executed control structures certainly isn’t.”
Our reading: GPT-4 was probably not a pure LLM, and o3 using code interpreter is certainly not a pure LLM.
No horizon
Qualifier: o3 when it invokes code interpreter
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Fostering its further development may be among the best things that companies, researchers, and governments can do. Investors, take note.”
Our reading: Fostering the further development of neurosymbolic AI may be among the best things that companies, researchers, and governments can do, and investors should take note.
No horizon
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Observation
“None of which means we are close to AGI. There are many ways to put together symbols and neural nets. I am not at all convinced we have the right one.”
Our reading: These results do not mean we are close to AGI, and Marcus is not at all convinced the field has found the right way to combine symbols and neural nets.
No horizon
Qualifier: there are many ways to put together symbols and neural nets
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Observation
“After a thirty-year journey, I believe that neurosymbolic AI’s moment has finally arrived, in part from an unlikely place.”
Our reading: Neurosymbolic AI's moment has arrived as of mid-2025.
No horizon
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“We must find ways to bring them together.”
Our reading: The field must find ways to combine neural networks and symbolic AI.
No horizon
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Observation
“We have made some progress; we still likely have a long way to go.”
Our reading: We have made some progress toward AGI, but we likely still have a long way to go.
No horizon
Scope: in the mountain-climb metaphor with AGI as the highest peak
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Causal claim
“Hinton’s admonition against using symbolic AI was extraordinarily misguided, and probably delayed discovery of useful tools like Code Interpreter by years.”
Our reading: Hinton's opposition to symbolic AI was misguided and probably delayed tools like Code Interpreter by years.
No horizon
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Observation
“We still also (just as I argued in 2020) lack proper ways of building and inferring symbolic world/cognitive models on-the-fly.”
Our reading: AI still lacks proper ways to build and infer symbolic world models on the fly.
No horizon Evidence as of
Limitation: A stricter reading, such as 'no method works reliably', might be checkable. But that would add a threshold the speaker did not give.
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI
Assessment: Can’t be checkedSaid
Observation
“It’s likely that OpenAI, Anthropic and others are already doing more neurosymbolic AI than they actually make public. They still pretend it’s all “a model” but that model uses all kinds of tricks and mixtures with tools, code generation etc.”
Our reading: OpenAI, Anthropic and others are likely using more neurosymbolic techniques than they disclose.
No horizon
Source: How o3 and Grok 4 Accidentally Vindicated Neurosymbolic AI