Claims that can’t be checked, page 32 of 37
Claims by Gary Marcus and Ed Zitron. Deadlines are judged as of 2026-10-29.
Claims that can’t be checked (3643)
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 32 of 37: claims 3101 to 3200 of 3643, ordered by the date first said, newest first. To find a claim by its words, use search.
Assessment: Can’t be checkedSaid
Gary Marcus Conditional forecast
“I agree that if Alphabet pull the plug, there will likely be a self-driving winter across the industry (though for now players like Aurora and Waabi may still have plenty of cash).”
Our reading: If Alphabet shuts down Waymo, the self-driving industry will likely enter a 'winter', though players like Aurora and Waabi may still have plenty of cash for now.
No horizon
Only if: if Alphabet pulls the plug on Waymo
Source: Why driverless cars might—or might not—be at the end of the road
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“And while every large tech company is making some effort around AI safety at this point, the collective sum of those efforts hasn’t yielded all that much.”
Our reading: The combined AI safety efforts of large tech companies have not yielded much.
No horizon
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“Anthropic’s “Constitutional AI” idea seems to me to be clever and the best of the bunch, but still inadequate, riddled with the usual problems that LLMs face (reliability, excessive dependence on similarity to what is in the training set, hallucinations, unreliable reasoning, etc).”
Our reading: It seems to Marcus that Anthropic's Constitutional AI idea is clever and the best of the current AI safety ideas, but still inadequate, riddled with the usual LLM problems such as unreliability, excessive dependence on similarity to training data, hallucinations and unreliable reasoning.
No horizon
Qualifier: seems to me
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“What I noted in 2017 (at the time focusing on AI for medicine rather on AI safety per se) is that some problems in AI might be too complex for individual labs, and not of sufficient financial interest to the large AI companies like Google and Facebook (now Meta), ehich are after all, driven by profits rather than humanity. That’s still true.”
Our reading: Some AI problems may be too complex for individual labs and of insufficient financial interest to profit-driven large AI companies, which still holds true.
No horizon
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“Don’t listen to them; enormous money is at stake.”
Our reading: People should not listen to those who deny there is any AI risk, such as Marc Andreessen, because enormous money is at stake.
No horizon
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“We need to consider both current AI risk and AGI risk. Period.”
Our reading: We need to consider both current AI risk and AGI risk.
No horizon
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“We need to stop worrying (just) about Skynet and robots taking over the world, and think a lot more about what criminals, including terrorists, might do with LLMs, and what, if anything, we might do to stop them.”
Our reading: AI risk discussion should focus much more on misuse of LLMs by criminals and terrorists, not just on AI takeover scenarios.
No horizon
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“The world can likely benefit enormously from AI, in ways that extend far beyond what we have already seen in recent months with productivity, but anyone with eyes can see that there are risks, too.”
Our reading: The world can likely benefit enormously from AI, in ways extending far beyond the productivity gains seen in recent months, though there are clearly risks too.
No horizon
Qualifier: can (potential benefit, not a prediction that it will happen)
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“Count me among those who continue to favor a CERN for AI safety and Global Safety, so long as we balance the portfolio of risks we are trying to address.”
Our reading: A CERN for AI safety and global AI safety governance should be created, provided they address a balanced range of risks.
No horizon
Qualifier: balance the portfolio of risks addressed
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“There are over 60 elections around the world in 2024; the chance that at least one will be swayed by deep fakes is high.”
Our reading: It is likely that at least one of the 60+ elections worldwide in 2024 will be swayed by deepfakes.
No horizon
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“That seems like a sensible idea; indeed, it is a crude version of the kind of hybrid neuro-symbolic system that we have been advocating for many years.”
Our reading: Having GPT-4 use external tools such as the Code Interpreter and Wolfram Alpha plug-ins seems a sensible idea, and it is a crude version of the hybrid neuro-symbolic system the authors have advocated for many years.
No horizon
Qualifier: seems like a sensible idea
Source: Getting GPT to work with external tools is harder than you think
Assessment: Can’t be checkedSaid
Gary Marcus Conditional forecast
“Some of these problems may fixable. With enough resources OpenAI or Wolfram Alpha could fix some of the problems in the interfaces.”
Our reading: With enough resources, OpenAI or Wolfram Alpha could fix some of the interface problems.
No horizon
Only if: with enough resources
Source: Getting GPT to work with external tools is harder than you think
Assessment: Can’t be checkedSaid
Gary Marcus Conditional forecast
“All that said, much more fundamental problems remain and, until those problems are addressed, the ability of AI to apply mathematics to real world situation will remain limited.”
Our reading: Until fundamental problems are addressed, AI's ability to apply mathematics to real-world situations will remain limited.
No horizon Evidence as of
Only if: until fundamental problems are addressed
Source: Getting GPT to work with external tools is harder than you think
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“We don’t claim that this problem is unsolvable, but addressing it may require major innovation.”
Our reading: Getting LLMs to reliably use external tools may require major innovation, though it is not necessarily unsolvable.
No horizon
Source: Getting GPT to work with external tools is harder than you think
Assessment: Can’t be checkedSaid
Gary Marcus Causal claim
“A true understanding of elementary mathematics and its uses requires a deeper view.”
Our reading: True understanding of elementary math requires more than LLM token prediction or Wolfram-style symbol manipulation.
No horizon
Source: Getting GPT to work with external tools is harder than you think
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“It may get fixed, with a band-aid of some sort or another, but we are confident a zillion other pictures like it will probably yield similar errors.”
Our reading: Even if this specific kitchen example gets patched, many other similar pictures will probably still produce similar errors in multimodal LLMs.
No horizon
Qualifier: even if this specific example is patched
Source: Hello, Multimodal Hallucinations
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“GPT-v (the visual version of GPT) can use words like refrigerator and counter, but that doesn’t mean it actually knows what these things mean..”
Our reading: GPT-4V can use words like refrigerator and counter, but that does not show it knows what they mean.
No horizon
Source: Hello, Multimodal Hallucinations
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“The whole point of Rebooting AI was that the AI of 2018 didn’t actually understand how the world works, and that we couldn’t trust AI that didn’t. We stand fully by what we said.”
Our reading: The AI of 2018 did not actually understand how the world works, AI that does not understand how the world works cannot be trusted, and the authors fully stand by both points from Rebooting AI.
No horizon
Source: Hello, Multimodal Hallucinations
Assessment: Can’t be checkedSaid
Gary Marcus Rhetoric
“The old systems were lost in a sea of correlations between words; the new system are lost in a sea of correlations between words and pixels.”
Our reading: Multimodal models, like text-only ones before them, depend on correlations (now between words and pixels) instead of understanding.
No horizon
Source: Hello, Multimodal Hallucinations
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“And 70% would be great for free throws; for multiplication it sucks.”
Our reading: Around 70% accuracy is unacceptable for multiplication, which requires exact answers.
No horizon
Source: “Math is hard” — if you are an LLM – and why that matters
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“Some kind of hybrid may well work, but LLMs on their own remain stuck.”
Our reading: Some kind of hybrid system may well succeed at reliable math where pure LLMs fail.
No horizon
Source: “Math is hard” — if you are an LLM – and why that matters
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“That, in a nutshell, is why we should never trust pure LLMs; even under carefully controlled circumstances with massive amounts of directly relevant data, they still never really get even the most basic linear functions.”
Our reading: Pure LLMs should never be trusted.
No horizon
Source: “Math is hard” — if you are an LLM – and why that matters
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“It’s certainly plausible that some of the prerequisites of AGI may have been achieved, but considerable overreach to presume that all of “the most important ones” have already been invented.”
Our reading: It is certainly plausible that some prerequisites of AGI may have been achieved, but it is considerable overreach to presume that all of the most important ones have already been invented.
No horizon
Qualifier: certainly plausible that some ... may have been achieved
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“Maybe what’s required already exists, but more likely, we are still at least a few paradigm shifts away .”
Our reading: AGI is more likely than not still at least a few paradigm shifts away.
No horizon
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“The problem with the paper though isn’t just its hostility towards critics, or the way in which it lowers the bar, it’s that its key arguments don’t go through.”
Our reading: The problem with the Agüera y Arcas and Norvig paper is not just its hostility toward critics or the way it lowers the bar for AGI, but that its key arguments don't go through.
No horizon
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“The fact that LLMs trained on the task of next-token prediction can also do a pretty good job at a variety of natural language tasks, and, very recently, a considerably less good job of tasks that combine text with images, as emphasized by Agüera y Arcas and Norvig is certainly interesting, but it doesn’t justify the claim that they constitute “general intelligence”.”
Our reading: LLMs trained on next-token prediction doing a pretty good job at a variety of natural language tasks, and very recently a considerably less good job at tasks combining text with images, is certainly interesting but does not justify the claim that they constitute general intelligence.
No horizon
Qualifier: very recently
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“In AI, no single technique is likely to be a silver bullet, now or ever.”
Our reading: No single AI technique is likely ever to be a silver bullet.
No horizon
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“But there is zero justification for claiming that the current technology has achieved general intelligence.”
Our reading: There is zero justification for claiming that the current technology has achieved general intelligence.
No horizon
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“Current models trained on next-token prediction have remarkable abilities and remarkable weaknesses. The scope of those abilities and weaknesses of the current systems is not well understood.”
Our reading: Current next-token prediction models have remarkable abilities and weaknesses whose scope is not well understood.
No horizon
Assessment: Can’t be checkedSaid
Gary Marcus Conditional forecast
“If GPT-5 did fail to please, or even if it were just deferred indefinitely (like so many promises that have been about driverless cars), it could have a rapid, deflationary effect on sky-high valuations.”
Our reading: If GPT-5 disappoints or is indefinitely deferred, AI valuations could deflate rapidly.
No horizon
Only if: if GPT-5 disappoints or is deferred indefinitely
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“They might have decided that GPT-5, if it was simply a bigger version of GPT-4, would not meet expectations, and that it wasn’t worth spending the hundreds of millions of dollars required, if the outcome would only turn out to be disappointing, or even embarrassing.”
Our reading: OpenAI might have decided, perhaps after early proof-of-concept tests by May 2023, that GPT-5, if simply a bigger GPT-4, would not meet expectations and was not worth the hundreds of millions of dollars required if the outcome would be disappointing or embarrassing.
No horizon
Qualifier: if GPT-5 was simply a bigger version of GPT-4
Assessment: Can’t be checkedSaid
Gary Marcus Conditional forecast
“People might even start to realize that building truly reliable, trustworthy AI is actually really, really hard.”
Our reading: If GPT-5 failed to please or were deferred indefinitely, people might even start to realize that building truly reliable, trustworthy AI is really, really hard.
No horizon
Only if: if GPT-5 failed to please or were deferred indefinitely
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“Another theory is that OpenAI realized how expensive it would be either to train the model or to run it, and wasn’t sure they could make a profit given those costs. That still seems viable.”
Our reading: It is plausible that OpenAI was unsure it could profit from GPT-5 given training and inference costs.
No horizon
Assessment: Can’t be checkedSaid
Gary Marcus Conditional forecast
“If Yann LeCun and I are correct, AI will need genuinely new paradigms to get to the next level in AI, and OpenAI may be no closer to that than anyone else.”
Our reading: If Marcus and LeCun are correct, AI will need genuinely new paradigms to get to the next level, and OpenAI may be no closer to that than anyone else.
No horizon
Only if: if Yann LeCun and Marcus are correct
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“Few have risen as quickly, but there’s no guarantee that OpenAI can maintain their stratospheric trajectory.”
Our reading: OpenAI may not be able to maintain its rapid growth trajectory.
No horizon
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“It's mean *working* for a positive future, not just declaring that one is inevitable and hoping for the best.”
Our reading: A positive technological future has to be actively worked for, not just declared inevitable while hoping for the best.
No horizon
Source: Two kinds of Techno-Optimism
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“It means working to recognize the risks, and working to address them, so that we can reach the positive outcomes that are promised, not casting those who recognize those risks as the enemy.”
Our reading: Techno-optimists should work to recognize and address technology's risks to reach its promised benefits, rather than treating people who point out those risks as enemies.
No horizon
Source: Two kinds of Techno-Optimism
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“But being long-term optimistic about technology doesn't mean you have to ignore the short-term (or long-term) risks of technology.”
Our reading: Being long-term optimistic about technology does not mean one has to ignore the short-term or long-term risks of technology.
No horizon
Scope: for someone who is long-term optimistic about technology
Source: Two kinds of Techno-Optimism
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“Our focus should be on figuring on how to build AI that can represent and reason about *values*, rather than simply perpetuating past data”.”
Our reading: AI research should focus on building systems that can represent and reason about values instead of just perpetuating past data.
No horizon
Source: Race, statistics, and the persistent cognitive limitations of DALL-E
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“It’s that it can’t separate the world from the statistics of its dataset. DALL-E does not have a cognitive construct of a doctor or a patient or a human being or an occupation or medicine or race or egalitarianism or equal opportunity or any of that.”
Our reading: DALL-E 3 cannot distinguish the world from its training data statistics and has no cognitive concepts of doctors, patients, race, or similar.
No horizon
Source: Race, statistics, and the persistent cognitive limitations of DALL-E
Assessment: Can’t be checkedSaid
Gary Marcus Causal claim
“But the level of actual cognitive comprehension of here is low. The modest number of properly labeled examples is presumably swamped by all the billions of watch advertisements.”
Our reading: Multimodal ChatGPT has low real comprehension, and the few correctly labeled clock examples are probably drowned out by billions of watch ads showing the canonical time.
No horizon
Source: Race, statistics, and the persistent cognitive limitations of DALL-E
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“Blind fealty to the statistics of arbitrary data is not AGI. It’s a hack. And not one that we should want.”
Our reading: Blind fealty to the statistics of arbitrary data is not AGI; it is a hack, and not one we should want.
No horizon
Source: Race, statistics, and the persistent cognitive limitations of DALL-E
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“What one really needs, and no one yet knows how to build, is a system that can derive semantics of wholes from their parts as a function of their syntax.”
Our reading: AI needs systems that derive the meaning of wholes from their parts according to syntax, which no one yet knows how to build.
No horizon
Source: Race, statistics, and the persistent cognitive limitations of DALL-E
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“but we are in complete agreement that we are at a turning point with enormous uncertainty, and that we need to make the right choices now.”
Our reading: AI is at a turning point of enormous uncertainty, and society needs to make the right choices now.
No horizon
Source: What was 60 Minutes thinking, in that interview with Geoff Hinton?
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“To be fair, large language models aren’t just looking things up, but the idea that a good prediction of next word necessarily implies understanding is fallacious.”
Our reading: Good next-word prediction by LLMs does not necessarily imply understanding.
No horizon Evidence as of
Limitation: This does not reject Marcus's argument or endorse Hinton's.
Source: What was 60 Minutes thinking, in that interview with Geoff Hinton?
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“Of course, I don’t think these machines are conscious, either.”
Our reading: Current AI machines are not conscious.
No horizon
Source: What was 60 Minutes thinking, in that interview with Geoff Hinton?
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“But I don’t we are all that close to machines that are more intelligent than us”
Our reading: We are not very close to machines that are more intelligent than humans.
No horizon
Qualifier: partly definitional
Source: What was 60 Minutes thinking, in that interview with Geoff Hinton?
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“We are giving way too much authority to machines that we can’t control.”
Our reading: We are giving way too much authority to machines that we can't control.
No horizon
Source: What was 60 Minutes thinking, in that interview with Geoff Hinton?
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“Since they can’t reliably solve them, I don’t think we should attribute “understanding” to them, at least not in any remotely deep sense of the word understanding.”
Our reading: We should not attribute deep understanding to generative AI systems because they cannot reliably solve basic problems like family-relationship inference.
No horizon
Qualifier: because they cannot reliably solve basic inference problems
Source: What was 60 Minutes thinking, in that interview with Geoff Hinton?
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“this is one of the reasons laws should be written to make machines disclose the fact that they are machines.”
Our reading: Laws should require machines to disclose that they are machines.
No horizon
Source: What was 60 Minutes thinking, in that interview with Geoff Hinton?
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“I am much more worried about bad actors deliberately misusing AI, than machines deliberately wanting to take over. But Geoff’s right that we can’t fully rule it out either.”
Our reading: Marcus is much more worried about bad actors deliberately misusing AI than about machines deliberately wanting to take over, though the possibility of a takeover cannot be fully ruled out.
No horizon
Qualifier: cannot fully rule it out
Source: What was 60 Minutes thinking, in that interview with Geoff Hinton?
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“Finding that some stuff correlates with space or time doesn’t mean that stuff genuinely represents space or time.”
Our reading: The correlations with space and time in Gurnee and Tegmark's findings do not show that LLMs genuinely represent space or time.
No horizon
Source: Muddles about Models
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“Of course as Fei Fei Li correctly pointed out on X, it all depends on what you mean by a “model”, but here’s the crux, same as it ever was: correlations aren’t causal, semantic models.”
Our reading: Correlations found inside LLMs are not causal, semantic world models.
No horizon
Source: Muddles about Models
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“And here is why all this matters: safety, which is what Tegmark really cares about, is not going to emerge anytime soon, and probably ever, from hit-or-miss systems that substitute correlation for a factual, structured, interrogable representations of the world.”
Our reading: AI safety will not emerge anytime soon, and probably never, from correlation-based systems that lack structured, interrogable representations of the world.
No horizon
Qualifier: systems that substitute correlation for structured representations
Source: Muddles about Models
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“I would conjecture that you can’t say to ChatGPT, let’s play a game of chess and then (say) 37 moves in pause, and get it to reliably answer how many white bishops are left on the board, nor to explain what happened to some pawn that had been taken a few moves earlier.”
Our reading: Marcus conjectures that ChatGPT cannot reliably say how many white bishops remain, or what happened to a captured pawn, partway through a chess game.
No horizon
Qualifier: conjecture
Source: Muddles about Models
Assessment: Can’t be checkedSaid
Gary Marcus Conditional forecast
“To encapsulate the heart of what Ernie Davis and I said at length in our 2019 book Rebooting AI, AI is unlikely to ever be trustworthy if its makers continue to skip the step of having explicit cognitive models of the world.”
Our reading: AI is unlikely ever to be trustworthy if its makers keep skipping explicit cognitive models of the world.
No horizon
Only if: if makers continue to skip explicit cognitive models of the world
Source: Muddles about Models
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“The worst part is not that every single sentence contains at least one lie, but that the whole thing sounds detailed and real. To someone who wasn’t following matters closely, it might well all sound plausible.”
Our reading: The worst part of Bing's false answer is not that every sentence contains at least one falsehood but that the whole thing sounds detailed and real, so that to someone who wasn't following matters closely it might well all sound plausible.
No horizon
Scope: to someone who wasn't following matters closely
Source: Seven lies in four sentences
Assessment: Can’t be checkedSaid
Gary Marcus Rhetoric
“Ezra and I harped on their propensity to generate bullshit because bullshit was and is their fundamental product.”
Our reading: Bullshit was and still is the fundamental product of LLMs.
No horizon
Source: Seven lies in four sentences
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“we really ought to commit to building AI that can.”
Our reading: The field should commit to building AI that can distinguish facts from fiction.
No horizon
Source: Seven lies in four sentences
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“That projection probably comes from their two best quarters, and some data suggest their spectacular growth has already been a slow down after the amazing initial takeoff.”
Our reading: OpenAI's projected $1 billion in revenue this year probably comes from its two best quarters, and some data suggest its growth has already slowed after the initial takeoff.
No horizon
Qualifier: probably
Source: Top 5 reasons OpenAI is probably not worth 90 billion dollars
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“It might be OK in domains where 70% correct is OK, but unlikely to be acceptable in mission-critical applications.”
Our reading: LLMs may be acceptable where 70% accuracy suffices but are unlikely to be acceptable for mission-critical applications.
No horizon
Qualifier: mission-critical applications
Source: Top 5 reasons OpenAI is probably not worth 90 billion dollars
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“LLMs will likely never deliver on the promise that they are somehow close to AGI.”
Our reading: LLMs will likely never deliver on the promise of being close to AGI.
No horizon
Source: Top 5 reasons OpenAI is probably not worth 90 billion dollars
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“In the current state, even small changes to the underlying database might require complete retraining.”
Our reading: Currently, even small changes to an LLM's training data might require complete retraining.
No horizon
Qualifier: small changes to training data
Source: Top 5 reasons OpenAI is probably not worth 90 billion dollars
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“Students don’t as rule have a lot of money; they may quickly turn to free alternatives.”
Our reading: Students using OpenAI products may quickly switch to free alternatives.
No horizon
Source: Top 5 reasons OpenAI is probably not worth 90 billion dollars
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“Current models still have a lot of problems, necessitating future models. Future models will likely be bigger models, even more expensive to train, even more expensive to operate, further reducing profits.”
Our reading: OpenAI's future models will likely be bigger and more expensive to train and operate, further reducing profits.
No horizon
Source: Top 5 reasons OpenAI is probably not worth 90 billion dollars
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“If, after training on virtually the entire internet, you know Tom is Mary Lee‘s son, but can’t figure out without special prompting that Mary Lee therefore is Tom’s mother, you have no business running all the world’s software.”
Our reading: A system that, after training on virtually the entire internet, knows Tom is Mary Lee's son but can't figure out without special prompting that Mary Lee is therefore Tom's mother has no business running all the world's software.
No horizon
Qualifier: after training on virtually the entire internet
Source: Elegant and powerful new result that seriously undermines large language models
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“Computers should be expected to make optimal use of information given, regardless of human limitations.”
Our reading: Computers should be expected to make optimal use of the information given, regardless of human limitations.
No horizon
Qualifier: regardless of human limitations
Source: Elegant and powerful new result that seriously undermines large language models
Assessment: Can’t be checkedSaid
Gary Marcus Causal claim
“The will to believe in neural networks is frequently so strong that counterevidence is often dismissed or ignored, for much too long.”
Our reading: The will to believe in neural networks is frequently so strong that counterevidence is often dismissed or ignored for much too long.
No horizon
Qualifier: frequently
Source: Elegant and powerful new result that seriously undermines large language models
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“None of what I raised then has really been adequately addressed in the intervening decades.”
Our reading: The generalization problems Marcus raised in 2001 about neural networks have not been adequately addressed since.
No horizon
Source: Elegant and powerful new result that seriously undermines large language models
Assessment: Can’t be checkedSaid
Gary Marcus Causal claim
“My own guess is that between the release of GPT-4 in November 2022 and the May 2023 testimony OpenAI ran some preliminary tests and decided that the system would not meet expectations.”
Our reading: Marcus guesses that, between GPT-4's release (which he dates to November 2022) and the May 2023 testimony, OpenAI ran some preliminary tests and decided GPT-5 would not meet expectations, and that this explains why it is not training GPT-5.
No horizon
Qualifier: my own guess
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“We have far too little machinery in place to surveil or address such threats.”
Our reading: We have far too little machinery in place to surveil or address threats such as a new form of bioweapon attack that could arise if superintelligent AI were at some point imminent and posed a genuine threat.
No horizon
Scope: threats from hypothetically imminent superintelligent AI
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“But here’s someone else the AI industry is taking advantage of: you.”
Our reading: The AI industry is taking advantage of ordinary people, in this context through harvesting their personal data.
No horizon
Scope: in the context of collecting personal data (Meta, Google, Zoom, carmakers)
Source: The dirty underbelly of AI
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“My own view is that Cyc has been neither a success nor a failure, but somewhere in between: I see it as a ground-breaking, clarion experiment that never fully gelled.”
Our reading: In Marcus's own view, Cyc has been neither a success nor a failure but somewhere in between: a ground-breaking, clarion experiment that never fully gelled.
No horizon
Qualifier: in his own view
Source: Doug Lenat, 1950-2023
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“He didn’t fully succeed – we will need a different path – but he picked the critical mountain that we still must cross.”
Our reading: Capturing common sense is the critical problem AI still has to solve, but it will take a different path from Cyc's.
No horizon
Source: Doug Lenat, 1950-2023
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“But every single one of them should know something about Cyc.”
Our reading: Every young AI researcher should know something about Cyc.
No horizon
Source: Doug Lenat, 1950-2023
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“Not because Cyc will get used out of the box, as some sort of drop-in replacement for Large Language Models, but because what Lenat tried to do - to get machines to represent and reason about common sense — still must be done.”
Our reading: Getting machines to represent and reason about common sense still has to be done, though Cyc itself will not serve as a drop-in replacement for LLMs.
No horizon
Source: Doug Lenat, 1950-2023
Assessment: Can’t be checkedSaid
Gary Marcus Conditional forecast
“No, Cyc didn’t set the world on fire, but yes, it will seem more and more important in hindsight, as we eventually make real progress towards artificial general intelligence.”
Our reading: Cyc will seem more and more important in hindsight as real progress toward artificial general intelligence is eventually made.
No horizon
Only if: as real progress toward artificial general intelligence is eventually made
Source: Doug Lenat, 1950-2023
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“The biggest problem with Cyc from an academic perspective is that it’s proprietary.”
Our reading: From an academic perspective, Cyc's biggest problem is that it is proprietary.
No horizon
Qualifier: from an academic perspective
Source: Doug Lenat, 1950-2023
Assessment: Can’t be checkedSaid
Ed Zitron Recommendation or value judgment
“An industry where David Sacks, Chamath Palihapitiya, Jason Calacanis, Elon Musk, and Marc Andreessen lead the conversation is not one to be proud of, resembling the opening of a dystopian sci-fi more than any fanciful lie about utopias powered by AI.”
Our reading: An industry where David Sacks, Chamath Palihapitiya, Jason Calacanis, Elon Musk, and Marc Andreessen lead the conversation is not one to be proud of, and it resembles the opening of a dystopian sci-fi more than any fanciful lie about AI-powered utopias.
No horizon
Scope: an industry where David Sacks, Chamath Palihapitiya, Jason Calacanis, Elon Musk, and Marc Andreessen lead the conversation
Source: You Are Not The Man In The Arena
Assessment: Can’t be checkedSaid
Ed Zitron Recommendation or value judgment
“The argument, of course, is that AI is “in its early days,” the same noxious defense that was used for cryptocurrency — which had also been around for many years before its venture-backed boom.”
Our reading: The argument that AI is 'in its early days' is a noxious defense, the same one used for cryptocurrency, which had also been around for many years before its venture-backed boom.
No horizon
Source: You Are Not The Man In The Arena
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“If we are to get the right regulatory regime for AI – one that protects consumers, without stifling innovation – we are going to need government to step up.”
Our reading: Government must step up to achieve an AI regulatory regime that protects consumers without stifling innovation.
No horizon
Qualifier: to get the right regulatory regime for AI
Source: Could we all be doomed, without ranked-choice voting?
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“I rather doubt that will happen, but I am not sure that it won’t.”
Our reading: An AI-induced extinction event that kills everyone is unlikely, though not ruled out.
No horizon
Source: Could we all be doomed, without ranked-choice voting?
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“The goal of the companies is return to shareholders, not to return to humanity.”
Our reading: AI companies aim at returns to shareholders rather than benefit to humanity.
No horizon
Source: Could we all be doomed, without ranked-choice voting?
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“But I think ranked-choice-voting (or some similar alternative) is very much worth serious consideration.”
Our reading: Marcus thinks ranked-choice voting, or some similar alternative, is very much worth serious consideration.
No horizon
Qualifier: I think
Source: Could we all be doomed, without ranked-choice voting?
Assessment: Can’t be checkedSaid
Gary Marcus Conditional forecast
“What I am sure of, though, is that our chances to avoid serious consequences aren’t great if we control neither AI nor the companies that make AI.”
Our reading: If neither AI nor AI companies are controlled, society's chances of avoiding serious consequences are poor.
No horizon
Only if: if we control neither AI nor the companies that make AI
Source: Could we all be doomed, without ranked-choice voting?
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“We waited too long on climate change; to the extent that there may be any real risk, either of outright doom or “mere” catastrophe, let’s not wait too long on AI.”
Our reading: To the extent AI carries any real risk of doom or catastrophe, society should not wait too long to address it, unlike with climate change.
No horizon
Qualifier: to the extent there may be real risk of doom or catastrophe
Source: 𝐩(𝐝𝐨𝐨𝐦)
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“I am, at least for now, much more worried about the overapplication of mediocre AI (eg what happens if we put large language models into running military strategy or into hiring decisions that affect millions of people) than the use of AI that might be so smart it could catch its own errors.”
Our reading: Marcus says that, at least for now, he is much more worried about the overapplication of mediocre AI (e.g., putting large language models into running military strategy or into hiring decisions affecting millions) than about the use of AI that might be so smart it could catch its own errors.
No horizon
Qualifier: at least for now
Source: 𝐩(𝐝𝐨𝐨𝐦)
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“And what really gets the goat of people like Timnit Gebru and Abeba Birhane, not unreasonably, is that there is a good chance that any immediate effects will disproprotionately affect the dispossessed; they are very likely right about that.”
Our reading: Marcus says Gebru and Birhane are very likely right that there is a good chance any immediate effects of AI will disproportionately affect the dispossessed.
No horizon
Qualifier: a good chance
Source: 𝐩(𝐝𝐨𝐨𝐦)
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“Unlike Yudkowsky I don’t think the probabilities are high enough—or imminent enough- to warrant bombing data centers any time soon. (Or probably ever.)”
Our reading: AI doom probabilities are not high or imminent enough to justify bombing data centers any time soon, and probably never will be.
No horizon
Source: 𝐩(𝐝𝐨𝐨𝐦)
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“It’s therefore worth getting a rough idea of the number, even if we can’t achieve consensus on the exact number.”
Our reading: It is worth getting a rough estimate of p(doom) even if there is no consensus on the exact number.
No horizon
Qualifier: even without consensus on the exact number
Source: 𝐩(𝐝𝐨𝐨𝐦)
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“None of this means of course that I suddenly think that we are close to super-smart machines. (Spoiler alert: we are not).”
Our reading: We are not close to super-smart machines.
No horizon Evidence as of
Limitation: This is not a finding that superintelligence was near or far in 2023.
Source: 𝐩(𝐝𝐨𝐨𝐦)
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“no matter what you feel about that p(doom | superintelligence), whether you see it as sci-fi or not, you should be worried about p(catastrophe | widespread AI adoption)”
Our reading: Whatever one thinks about doom from superintelligence, people should worry about the probability of catastrophe given widespread AI adoption.
No horizon
Qualifier: regardless of views on superintelligence doom
Source: 𝐩(𝐝𝐨𝐨𝐦)
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“Perhaps extinction is vanishingly unlikely, but what I will call 𝐩(𝐜𝐚𝐭𝐚𝐬𝐭𝐫𝐨𝐩𝐡𝐞) – the chance that an incident that kills (say) one percent or more of the population, seems a lot higher.”
Our reading: Perhaps extinction is vanishingly unlikely, but the chance of an incident that kills, say, one percent or more of the population seems a lot higher.
No horizon
Qualifier: perhaps
Source: 𝐩(𝐝𝐨𝐨𝐦)
Assessment: Can’t be checkedSaid
Gary Marcus Causal claim
“As Englander well knows, we will never get perfect estimates, but even imperfect estimates might help guide policy decisions as we try to balance attention to short-term and long-term risks in AI.”
Our reading: We will never get perfect estimates of p(doom) or p(catastrophe), but even imperfect estimates might help guide policy decisions as we try to balance attention to short-term and long-term AI risks.
No horizon
Qualifier: even imperfect estimates
Source: 𝐩(𝐝𝐨𝐨𝐦)
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“LeCun’s conclusion —that humans will always set the agenda—might be true, but he hasn’t given a real argument.”
Our reading: LeCun has given no real argument that humans will always set the agenda over AI, though that conclusion might be true.
No horizon Evidence as of
Limitation: This does not endorse LeCun’s conclusion or establish that his argument is sound.
Source: 𝐩(𝐝𝐨𝐨𝐦)
Assessment: Can’t be checkedSaid
Gary Marcus Forecast
“It’s a story for another day, but clearly you can have catastrophe without mass death, and we are heading in that direction.”
Our reading: AI can bring catastrophe without mass death, and society is heading in that direction.
No horizon
Source: 𝐩(𝐝𝐨𝐨𝐦)
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“There are in fact already some scenarios that are faintly imaginable now (e.g. bad actors using machines to create novel toxins, in ways that go awry), and over time could be more.”
Our reading: Some faintly imaginable scenarios already exist in which machines could kill everyone, such as bad actors using machines to create novel toxins, and more such scenarios could appear over time.
No horizon
Source: 𝐩(𝐝𝐨𝐨𝐦)
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“But imprecision is not an excuse for inaction.”
Our reading: Imprecise risk estimates do not justify failing to act on AI risk.
No horizon
Source: 𝐩(𝐝𝐨𝐨𝐦)
Assessment: Can’t be checkedSaid
Gary Marcus Recommendation or value judgment
“Unfortunately, the fact that we can’t predict the course of technology with any precision even a decade out doesn’t magically immunize us from its effects, nor relieve us of a moral responsbility to try to take precautions.”
Our reading: Not being able to predict technology precisely does not excuse us from the moral responsibility to take precautions against its risks.
No horizon
Source: 𝐩(𝐝𝐨𝐨𝐦)
Assessment: Can’t be checkedSaid
Gary Marcus Observation
“But bluntly, uncontrollable superintelligence is by no means the only risk from AI, and from a policy perspective may not be the most important risk, given both the lowish (?) probability and the time horizon.”
Our reading: Uncontrollable superintelligence is by no means the only risk from AI and, from a policy perspective, may not be the most important risk, given both its lowish (?) probability and the time horizon.
No horizon
Qualifier: from a policy perspective
Source: 𝐩(𝐝𝐨𝐨𝐦)