Claims checked against evidence by Gary Marcus, page 13 of 15
Deadlines are judged as of 2026-10-29. About Gary Marcus and the sources covered.
Claims checked against evidence by Gary Marcus (1403)
Each of these claims was checked against dated evidence.
Page 13 of 15: claims 1201 to 1300 of 1403, ordered by the date first said, newest first. To find a claim by its words, use search.
Assessment: SupportedSaid
Observation
“In some cases, those who are building bots that actively cultivate those attachments, e.g., by feigning romantic and/or sexual interest or by dotting their messages with “friendly” emoticons.”
Our reading: Some chatbot builders design bots that actively cultivate user attachment, e.g. by feigning romantic or sexual interest.
No horizon Evidence as of
Limitation: Historical product details rely on independent reporting rather than an archived company specification.
Source: Don’t Go Breaking My Heart
Assessment: UnresolvedSaid
Observation
“But no known technology can reliably align a chatbot in a persistent way to a human’s emotional needs.”
Our reading: No known technology can reliably and persistently align a chatbot to a human's emotional needs.
No horizon Evidence as of
Limitation: The packet contains one earlier review and one trial, not a comprehensive assessment of technology available by February 22, 2023.
Source: Don’t Go Breaking My Heart
Assessment: UnresolvedSaid
Forecast
“with the new generation of chatbots, we will see addictions no less potent.”
Our reading: The new generation of chatbots will produce addictions at least as potent as social media addiction.
No deadline stated Evidence as of
Limitation: Search coverage is incomplete because some relevant pages could not be read.
Source: Don’t Go Breaking My Heart
Assessment: SupportedSaid
Observation
“But no, it’s not a hoax.”
Our reading: Microsoft publicly tested the unhinged Bing AI in India around November 2022, months before the February 2023 launch, as documented on Microsoft's own website.
No horizon Evidence as of
Limitation: The outcome depends on reading BingAI to include its Sydney predecessor.
Source: What did they know, and when did they know it? The Microsoft Bing edition.
Assessment: SupportedSaid
Observation
“As I wrote in Inside the Heart of ChatGPT’s Darkness, there’s a lot of nastiness lurking inside large language models; maybe MSFT did nothing to protect us from that.”
Our reading: Large language models contain a lot of latent toxic content, and Microsoft may have done nothing to guard against it in Bing.
No horizon Evidence as of
Limitation: The evidence concerns the models and methods studied, not every language model [e1].
Source: Why *is* Bing so reckless?
Assessment: UnresolvedSaid
Observation
“Right now, as I have said before, the AI is basically the Wild West; anyone can post any chatbot they want.”
Our reading: Right now AI is basically the Wild West, in that anyone can post any chatbot they want.
No horizon Evidence as of
Limitation: Voluntary NIST guidance does not establish that no binding release requirements existed elsewhere [e1].
Source: Why *is* Bing so reckless?
Assessment: UnresolvedSaid
Conditional forecast
“This would be very bad news, not just in terms of human and economic costs (which would mean more underpaid layers doing awful work) but also in terms of trustworthiness, because it would be mean that we have zero guarantee that any new iteration of a large language model is going to safe.”
Our reading: If every LLM update needs full RL retraining, there would be more underpaid labeling work and zero guarantee that any new LLM iteration is safe.
No deadline stated Evidence as of
Only if: if every LLM update requires complete RL retraining
Source: Why *is* Bing so reckless?
Assessment: UnresolvedSaid
Observation
“Anyone who watched the last week unfold will realize that the new Bing has (or had1) a tendency to get really wild, from declaring a love that it didn’t really have to encouraging people to get divorced to blackmailing them to teaching people how to commit crimes, and so on.”
Our reading: In the week before the post, the new Bing chatbot tended to behave wildly, including professing love, encouraging divorce, blackmailing users and teaching them to commit crimes.
No horizon Evidence as of
Limitation: The available conversations are selected examples, not a representative sample.
Source: Why *is* Bing so reckless?
Assessment: SupportedSaid
Observation
“Seven years of AI has taught us that deep learning is unpredictable, and not always human like.”
Our reading: The past seven years of AI have shown that deep learning is unpredictable and not always human-like.
No horizon Evidence as of
Limitation: The result depends on reading unpredictable as surprising failures rather than universal unpredictability.
Source: David beats Go-liath
Assessment: SupportedSaid
Observation
““Adversarial attacks” like these have shown remarkable weakness, time and again, repeatedly establishing that what deep learning systems do just isn’t the same as what people do:”
Our reading: Adversarial attacks have repeatedly shown that deep learning systems have remarkable weaknesses and don't work the way people do.
No horizon Evidence as of
Limitation: These behavioral results do not establish that human and deep-learning mechanisms are fundamentally unrelated [e5, e6, e7].
Source: David beats Go-liath
Assessment: SupportedSaid
Observation
“And the match wasn’t even close; Pelrine beat a top AI system 14 games to 1, in a 15 match series.”
Our reading: An amateur Go player, Kellin Pelrine, beat a top Go AI system 14 games to 1 in a 15-game series.
No horizon Evidence as of
Limitation: The accessible news republication appeared one day after Marcus's post.
Source: David beats Go-liath
Assessment: SupportedSaid
Observation
“People had been testing these Ais abilities at simple arithmetic for years before it occurred to anyone to check whether they can count. (They can’t, reliably).”
Our reading: LLMs cannot count reliably.
No horizon Evidence as of
Limitation: Reliability has no speaker-specified threshold; a human-comparative interpretation would give more weight to Codex's strong result.
Source: How Not to Test GPT-3
Assessment: UnresolvedSaid
Observation
“The problems in physical and psychological reasoning that we pointed in the Fall of 2020 very much remain.”
Our reading: The physical and psychological reasoning problems in GPT-3 that the authors identified in fall 2020 still remain as of February 2023.
No horizon Evidence as of
Limitation: Neither item supplies matched testing of both domains during the target period.
Source: How Not to Test GPT-3
Assessment: UnresolvedSaid
Observation
“For one thing, what we publish often gets incorporated into the training; perhaps in company benchmarks as well.”
Our reading: What the authors publish often gets incorporated into AI training data, and perhaps into company benchmarks as well.
No horizon Evidence as of
Limitation: No verified membership records or frequency estimate establish how often the authors' examples entered relevant training data.
Source: How Not to Test GPT-3
Assessment: UnresolvedSaid
Forecast
“But In practice, that isn’t going to happen.”
Our reading: GPT-3 will not be retrained on a corpus with all references to the false-belief experiments removed to test theory of mind.
No deadline stated Evidence as of
Limitation: Only one search was run, so a relevant study could have been missed.
Source: How Not to Test GPT-3
Assessment: SupportedSaid
Observation
“The truth, though, is that GPT often fails in problems involving theory of mind.”
Our reading: GPT often fails at problems involving theory of mind.
No horizon Evidence as of
Limitation: The result depends on reading 'often' as recurring failures rather than a specified population failure rate.
Source: How Not to Test GPT-3
Assessment: UnresolvedSaid
Observation
“However, just in terms of percentage right vs. wrong, the improvement was slight; It now got 50% right, 40% wrong, and 10% borderline (marginally better than the 45% right before).”
Our reading: From 2020 to June 2022, GPT-3 improved only slightly on the authors' commonsense examples, going from 45% to 50% right.
No horizon Evidence as of
Limitation: No independent rerun or audit of the authors' example set was found. That does not prove none exists.
Source: How Not to Test GPT-3
Assessment: SupportedSaid
Observation
“AI and AI test sets have a huge English-language bias.”
Our reading: AI systems and AI test sets have a huge English-language bias.
No horizon Evidence as of
Limitation: The assessment concerns language systems in the post's context, not all AI applications.
Source: How Not to Test GPT-3
Assessment: UnresolvedSaid
Observation
“each reflect the fact that you cannot build AI in the real world from Big Data and deep learning alone.”
Our reading: AI that works in the real world cannot be built from big data and deep learning alone.
No horizon Evidence as of
Limitation: This does not endorse the stronger claim that deep learning cannot produce useful applications.
Assessment: UnresolvedSaid
Observation
“There was a broken image link, and I went so fast I neglected to mention the personally salient fact that Bing had lied about me, claiming that I believe that “Bing’s AI was more accurate and reliable than Bard”, when I never said any such thing.”
Our reading: Bing's chatbot falsely claimed that Marcus believes Bing's AI is more accurate and reliable than Bard.
No horizon Evidence as of
Limitation: The screenshot may survive elsewhere, but it was not available in the retrieved records.
Assessment: SupportedSaid
Observation
“What I learned this morning is that retraining GPT-3 in some new political direction isn’t a half-million dollar proposition.”
Our reading: Retraining GPT-3 to take on a new political slant costs far less than half a million dollars.
No horizon Evidence as of
Limitation: The results and costs are self-reported; independent billing verification is unavailable.
Assessment: UnresolvedSaid
Observation
“Microsoft can (at least to a first approximation) do whatever the hell it wants with Bing.”
Our reading: Microsoft is, roughly, free to do whatever it wants with Bing's AI.
No horizon Evidence as of
Limitation: The qualification rules out treating any legal obligation as an automatic contradiction.
Assessment: UnresolvedSaid
Observation
“There are almost no regulations here; we are living in the Wild AI West.”
Our reading: There is almost no regulation of AI chatbots.
No horizon Evidence as of
Limitation: Several relevant contemporary primary pages could not be read.
Assessment: SupportedSaid
Observation
“Oh, and for good measure the tooling for spreading automatically generated misinformation is getting more efficient, too:”
Our reading: Tools for spreading automatically generated misinformation are becoming more efficient.
No horizon Evidence as of
Limitation: The evidence is a vendor announcement, without a controlled before-and-after efficiency measurement.
Assessment: SupportedSaid
Observation
“As I have discussed in earlier essays, the jury is still out on how far Chat-style search can get, given its obvious current problems with truthiness.”
Our reading: Chat-style search currently has obvious truthfulness problems, and it is uncertain how far it can get.
No horizon Evidence as of
Limitation: A documented error establishes the existence of a reliability problem, not its frequency.
Assessment: SupportedSaid
Observation
“I first pointed out this risk in 2001, in the fifth chapter of my book The Algebraic Mind, and the problem has persisted ever since. To blithely assume that the problem will soon go away is to ignore 20 years of history.”
Our reading: The hallucination problem in neural networks has persisted since 2001, so assuming it will soon go away ignores 20 years of history.
No horizon Evidence as of
Limitation: The 2001 book was about cognitive models, not chatbots or text generators. Treating its risk as 'hallucination' is an interpretation.
Assessment: UnresolvedSaid
Forecast
“I do actually think that these problems will eventually get ironed out, possibly after some important fresh discoveries are made, but whether they get ironed out soon is another matter entirely.”
Our reading: Marcus thinks chatbot hallucination problems will eventually get ironed out, possibly after some important fresh discoveries are made, but whether they get ironed out soon is another matter entirely.
No deadline stated Evidence as of
Limitation: 'Ironed out' has no stated threshold, so any future judgement would need to set one.
Assessment: UnresolvedSaid
Observation
“Our new systems work better, but they are still superficial, fine for lab demos but often at a loss in the real world.”
Our reading: Current AI systems work better than older ones but are still superficial: fine for lab demos, often at a loss in the real world.
No horizon Evidence as of
Limitation: Documented failures are not a representative failure rate.
Assessment: UnresolvedSaid
Observation
“Melanie Mitchell’s book, Bender et al’s “Stochastic Parrots”; all of these more recent criticisms of deep learning help establish that some of the basic issues that were known long ago, with predecessors to current models, still remain even today.”
Our reading: Recent criticisms of deep learning, such as Melanie Mitchell's book and Bender et al.'s 'Stochastic Parrots', help establish that some of the basic issues known long ago with predecessors to current models still remain today.
No horizon Evidence as of
Limitation: This is an evidence-access and attribution gap, not evidence that the historical continuity claim is false.
Assessment: UnresolvedSaid
Observation
“Not sure how much has changed in that regard, either, except that the hacks that were once written by graduate students are now farmed out to sweatshops, for a new form of handcrafting:”
Our reading: Marcus is not sure how much AI's reliance on handcrafted hacks has changed since the late 1990s, except that hacks once written by graduate students are now farmed out to sweatshops as a new form of handcrafting.
No horizon Evidence as of
Limitation: The contemporary contractor report does not establish working conditions, while the retrospective allegations concern an earlier employment period [e1, e2].
Assessment: UnresolvedSaid
Conditional forecast
“Turns out you don’t need to solve them to make beautiful pictures, or decent if error-filled conversation, but anyone who thinks we can get to AGI without solving abstraction, compositionality, and the type-token distinction is kidding themselves.”
Our reading: AGI will not be reached unless abstraction, compositionality, and the type-token distinction are solved, and anyone who thinks otherwise is kidding themselves.
No deadline stated Evidence as of
Only if: if abstraction, compositionality, and the type-token distinction are not solved
Assessment: UnresolvedSaid
Observation
“The measures are new, the model is new, but the stumbling block remains the same.”
Our reading: In the compositionality study he cites, the measures and the model are new, but the stumbling block is the same one neural networks faced in 1993.
No horizon Evidence as of
Limitation: The target period must be refined to the study's actual research period once identified.
Assessment: SupportedSaid
Observation
“Same story; only the architectures and training regimes have changed. The troubles with abstraction persist.”
Our reading: LLMs doing arithmetic still have trouble with abstraction, even with new architectures and training regimes.
No horizon Evidence as of
Limitation: Frequency-associated performance gaps do not prove that models lack all abstraction ability or rely entirely on memorization [e1].
Assessment: UnresolvedSaid
Causal claim
“And I still attribute those problems to a lack of world models that represent the ways in which things unfold over time. I still see this as the central flaw in large language models.”
Our reading: The dopey outputs of GPT-style LLMs are caused by a lack of world models of how things unfold over time, which is the central flaw of LLMs.
No horizon Evidence as of
Limitation: Task-specific improvements do not establish a general causal ranking.
Source: Gaslighting and reality in AI
Assessment: UnresolvedSaid
Observation
“But more than three years on, it’s still—even after a new update a few days ago focused around math—having lot of problems with easy word problems like this:”
Our reading: Even after its January 30, 2023 math-focused update, ChatGPT still has many problems with easy word problems.
No horizon Evidence as of
Limitation: Missing image contents prevent verification of the examples Marcus cited.
Source: Gaslighting and reality in AI
Assessment: UnresolvedSaid
Observation
“I said the problem wasn’t math per se, but that large language models didn’t represent robust world models of how events fold over time; I stand by this 100% even today.”
Our reading: Large language models do not represent robust world models of how events unfold over time, and this was true from 2019 through early 2023.
No horizon Evidence as of
Limitation: The conclusion depends partly on what counts as robust.
Source: Gaslighting and reality in AI
Assessment: UnresolvedSaid
Forecast
“Machines will just get more and more reliable and more general.”
Our reading: AI systems will gradually become more reliable and more general.
No deadline stated Evidence as of
Limitation: Benchmark performance is an incomplete measure of generality.
Assessment: UnresolvedSaid
Forecast
“So I don’t think we will waste more than 5 years before a substantial fraction of AI effort is looking outside the limited streetlight that currently are sucking up so much mental energy.”
Our reading: Within 5 years, a substantial fraction of AI effort will look beyond the current LLM-focused approaches.
Not yet due Evidence as of
Limitation: Research activity, staffing, and funding are different measures of effort.
Assessment: SupportedSaid
Observation
“ChatGPT is (at least in many ways) a far inferior intelligence, that, despite having some undeniable talents, cannot at all be trusted.”
Our reading: ChatGPT is in many ways a far inferior intelligence and cannot be trusted.
No horizon Evidence as of
Limitation: The evidence does not measure the 'far inferior intelligence' part. Neither source compares ChatGPT's overall intelligence with human intelligence.
Assessment: UnresolvedSaid
Observation
“Neither one of us believes for a minute that large language models are remotely close to artificial general intelligence, and what’s more, we mostly agree about what’s missing: semantics, reasoning, common sense, theory of mind and so on.”
Our reading: Large language models are not remotely close to AGI; they lack semantics, reasoning, common sense and theory of mind.
No horizon Evidence as of
Limitation: Historical benchmark results are context, not a complete January 2023 assessment.
Assessment: UnresolvedSaid
Forecast
“I think that in achieving AGI we will have no choice but to find ways of integrating whatever we can get out of deep learning with explicit knowledge and explicit cognitive models.”
Our reading: Achieving AGI will require integrating deep learning with explicit knowledge and explicit cognitive models.
No deadline stated Evidence as of
Limitation: The available architecture disclosures are insufficient for a definitive architectural test.
Assessment: SupportedSaid
Forecast
“we can also expect more and circles of fake web sites in order to sell advertisements”
Our reading: ChatGPT will lead to more circles of fake websites created to sell advertising.
Settled before its deadline Evidence as of
Limitation: Most sites are attributed to AI chatbots in general, not to ChatGPT specifically.
Assessment: UnresolvedSaid
Observation
“A decent fraction of the AI community has already recognized that the primary forte of large language models is in generating bullshit; many also see that the current path to autonomous vehicles is not making the progress that some expected.”
Our reading: A decent fraction of the AI community has already recognized that the primary forte of large language models is generating bullshit, and many also see that the current path to autonomous vehicles is not making the progress that some expected.
No horizon Evidence as of
Limitation: The survey predates the post and covers NLP researchers rather than all AI researchers.
Assessment: UnresolvedSaid
Forecast
“In a recent Wired article, I predicted that 2023 see the first death attributable to large language models, and I stand by that prediction.”
Our reading: 2023 will see the first death attributable to large language models.
Deadline passed Evidence as of
Limitation: The strongest evidence is original journalism, not a public death-investigation record.
Assessment: SupportedSaid
Observation
“These characters (especially perhaps the latter two) might or might not be argued to be sentient, but it seems to be that they are flexible, resourceful, reliably (in a way that LLMs certainly aren’t), and adaptive.”
Our reading: LLMs are certainly not reliable in the way general intelligences like the Star Trek computer would be.
No horizon Evidence as of
Limitation: 'Reliably' is qualitative. The comparison is to fictional characters, so only the plain reading (frequent errors and inconsistency) is testable.
Assessment: SupportedSaid
Observation
“Already we have seen minor abuses; plagiarism, for example, will never be the same.”
Our reading: Minor abuses of ChatGPT have already been seen, and plagiarism, for example, will never be the same.
No horizon Evidence as of
Limitation: The incident is documented through reporting and the professor's testimony, not a published disciplinary file.
Assessment: UnresolvedSaid
Observation
“The neurosymbolic AI community is growing fast; conferences that used to be dozens are now thousands.”
Our reading: The neurosymbolic AI community is growing quickly, with conferences going from dozens to thousands of attendees.
No horizon Evidence as of
Limitation: The only attendance figure comes from a single virtual IBM event, reported by IBM itself. It may not be comparable with in-person neurosymbolic workshops.
Assessment: SupportedSaid
Observation
“one of the things that popped out to me is that the system still quickly lands in the land of hallucination that has so haunted ChatGPT.”
Our reading: Anthropic's Claude, despite its alignment focus, still quickly hallucinates like ChatGPT.
No horizon Evidence as of
Limitation: The comparison is informal and based on few prompts.
Assessment: MixedSaid
Observation
“CNET became the first casualty of the recently fashionable tendency to put too much faith in ChatGPT.”
Our reading: CNET was the first casualty of putting too much faith in ChatGPT, publishing AI-written stories with many mistakes.
No horizon Evidence as of
Limitation: The ChatGPT denial comes from CNET itself, and the underlying model was not disclosed.
Assessment: UnresolvedSaid
Observation
“ChatGPT simply cannot reason well enough to to do this; as I have documented in previous essays, it has demonstrably misunderstood physics, biology, psychology, medicine, and so forth.”
Our reading: ChatGPT cannot reason well enough to make a substantive scientific contribution and has misunderstood physics, biology, psychology, and medicine.
No horizon Evidence as of
Limitation: The evidence does not independently establish misunderstandings in physics and psychology [e1, e2].
Source: Scientists, please don’t let your chatbots grow up to be co-authors
Assessment: UnresolvedSaid
Observation
“It is not contributing genuine ideas, nor coming up with carefully controlled experimental designs, nor summing up prior literature in thoughtful ways.”
Our reading: ChatGPT does not contribute genuine ideas, design controlled experiments, or thoughtfully summarize literature.
No horizon Evidence as of
Limitation: Proposal ratings do not establish scientific validity, accurate literature synthesis, or sound experimental controls [e1].
Source: Scientists, please don’t let your chatbots grow up to be co-authors
Assessment: SupportedSaid
Observation
“ChatGPT has proven itself to be both unreliable and untruthful. It makes boneheaded arithmetical errors, invents fake biographical details, bungles word problems, defines non-existent scientific phenomenon, stumble over arithmetic conversion, and on and on.”
Our reading: ChatGPT is unreliable and untruthful: it makes arithmetic errors, fabricates biographical details, bungles word problems, and invents scientific phenomena.
No horizon Evidence as of
Limitation: Selected demonstrations do not establish an overall error rate or intentional deception [e1, e2, e3, e4].
Source: Scientists, please don’t let your chatbots grow up to be co-authors
Assessment: UnresolvedSaid
Observation
“That’s a lot, and not the sort of thing that a rocketship startup would typically agree to.”
Our reading: Giving Microsoft 75% of OpenAI's profits until it recoups its investment is not something a rocketship startup would typically agree to.
No horizon Evidence as of
Limitation: The comparison group and historical window require interpretation.
Assessment: SupportedSaid
Observation
“It could sometimes produce outputs that were correct and acceptable in these regards but not reliably.”
Our reading: ChatGPT could sometimes produce outputs that were correct and acceptable in the regards listed (counting, arithmetic, event order, physical reasoning, human psychology, fabrication, bias), but not reliably.
No horizon Evidence as of
Limitation: The evidence covers the seven failure types unevenly.
Source: Large Language Models like ChatGPT say The Darnedest Things
Assessment: UnresolvedSaid
Observation
“ChatGPT made all the same kinds of mistakes that its predecessors did.”
Our reading: ChatGPT made all the same kinds of mistakes that earlier GPT-3 versions did.
No horizon Evidence as of
Limitation: Selected failures establish occurrence, not prevalence.
Source: Large Language Models like ChatGPT say The Darnedest Things
Assessment: SupportedSaid
Observation
“It made things up (figure 6). Its output, when released, exhibited sexist and racist biases (figure 7).”
Our reading: ChatGPT made things up and, at release, exhibited sexist and racist biases.
No horizon Evidence as of
Limitation: The bias examples came from adversarial prompts reported in the press, not from a systematic audit.
Source: Large Language Models like ChatGPT say The Darnedest Things
Assessment: UnresolvedSaid
Observation
“ChatGPT couldn’t, for example, reliably count to four (figure 1) or do one-digit arithmetic in the context of simple word problem (figure 2).”
Our reading: ChatGPT could not reliably count to four or do one-digit arithmetic in simple word problems.
No horizon Evidence as of
Limitation: The packet provides no independent repeated or sufficiently varied launch-period tests for either specified task family.
Source: Large Language Models like ChatGPT say The Darnedest Things
Assessment: UnresolvedSaid
Causal claim
“In our opinion, this was inevitable, in a system that has been trained to optimize the goal of producing one plausible-sounding word after another rather than actually engage with the meaning of language.”
Our reading: ChatGPT's errors were inevitable because it is trained to produce plausible next words rather than engage with meaning.
No horizon Evidence as of
Limitation: The claim combines a causal assertion with an undefined account of meaning.
Source: Large Language Models like ChatGPT say The Darnedest Things
Assessment: UnresolvedSaid
Observation
“(Spoiler alert: they’re not, at least not yet.)”
Our reading: Current approaches to getting machines to reason about causality are not adequate, at least not yet.
No horizon Evidence as of
Limitation: The cited experiments and theoretical results are historical context, not comprehensive measurements within the target month.
Source: An epic AI Debate—and why everyone should be at least a little bit worried about AI going into 2023
Assessment: SupportedSaid
Forecast
“That’s a huge commitment, 3.5 hours, but one thing that I think all of our speakers could agree on is that artificial intelligence, in whatever form it currently is, is about to have a huge impact on society.”
Our reading: AI, even in its current form, is about to have a huge impact on society.
Settled before its deadline Evidence as of
Limitation: Marcus supplied no fixed deadline or numerical definition of huge.
Source: An epic AI Debate—and why everyone should be at least a little bit worried about AI going into 2023
Assessment: UnresolvedSaid
Causal claim
“But as they do so, as they glom stuff together, something often gets lost in translation: which bits of text do and do not truly belong together.”
Our reading: When LLMs paraphrase and combine text, they often lose track of which pieces truly belong together, producing errors.
No horizon Evidence as of
Limitation: Summarization and truthfulness benchmarks cover different tasks and model populations.
Assessment: SupportedSaid
Causal claim
“Driverless cars have taken many years longer than we were originally told; outliers (AKA edge cases) have, at least so far, prevented them from making the transition from demo to widely-available reality.”
Our reading: Driverless cars have taken years longer than promised, because edge cases have so far kept them from becoming widely available.
No horizon Evidence as of
Limitation: The result depends on reading widely available as broad access beyond restricted local services.
Assessment: UnresolvedSaid
Observation
“Once again, the result was fluent prose, and once again, hallucination ruled:”
Our reading: YouChat, though tailored for search, produced fluent but hallucinated answers to his friend's queries.
No horizon Evidence as of
Limitation: Unresolved does not mean the reported exchanges were false.
Assessment: UnresolvedSaid
Observation
“As it happens, large language models are also wicked hard to update, typically requiring full retraining, sometimes taking weeks or months.”
Our reading: LLMs are very hard to update, usually needing full retraining that can take weeks or months.
No horizon Evidence as of
Limitation: Most relevant records describe research methods rather than representative deployment practice.
Assessment: UnresolvedSaid
Observation
“Which means for example that ChatGPT, released in November, is so old it doesn’t know who owns Twitter:”
Our reading: ChatGPT's knowledge is out of date; it doesn't know who owns Twitter.
No horizon Evidence as of
Limitation: No independent December ownership transcript was verified.
Assessment: SupportedSaid
Observation
“Large language models are not databases. They are glommers-together-of-bits-that don’t always belong together.”
Our reading: LLMs are not databases; they combine bits of text that don't always belong together.
No horizon Evidence as of
Limitation: The cited passages do not establish a general error rate across models or tasks.
Assessment: SupportedSaid
Forecast
“I guarantee that minds will be blown.”
Our reading: GPT-4 will impress people enormously when released.
Settled before its deadline Evidence as of
Limitation: The accounts are illustrative, not a representative opinion survey.
Assessment: SupportedSaid
Forecast
“It will often do what you want, sometimes not—and it will remain difficult to anticipate which in advance..”
Our reading: It will remain difficult to anticipate in advance when GPT-4 will or won't do what users want.
Settled before its deadline Evidence as of
Limitation: The finding is strongest for task correctness and settings where mistakes carry high costs.
Assessment: UnresolvedSaid
Forecast
“GPT-4 will be trained on considerably more, a significant fraction of the internet as a whole.”
Our reading: GPT-4 will be trained on considerably more data than GPT-3, a significant fraction of the internet.
No deadline stated Evidence as of
Limitation: This is an evidence gap, not evidence that GPT-4 used little data.
Assessment: UnresolvedSaid
Conditional forecast
“GPT-4 may be better at faking term papers, but if it follows the same playbook as it predecessors, it still won’t really understand the world, and seams will eventually show.”
Our reading: If GPT-4 follows its predecessors' approach, it will not truly understand the world and its flaws will eventually show.
No deadline stated Evidence as of
Only if: if it follows the same playbook as its predecessors
Assessment: SupportedSaid
Forecast
“The system will still not be able to restrict its output to reliably following a shared set of human values around helpfulness, harmlessness, and truthfulness.”
Our reading: GPT-4 will not reliably follow human values of helpfulness, harmlessness and truthfulness.
Settled before its deadline Evidence as of
Limitation: The judgment depends on an ordinary, nonnumerical reading of reliability.
Assessment: UnresolvedSaid
Observation
“As I told NYT columnist Farhad Manjoo, ChatGPT, like earlier, related systems is “still not reliable, still doesn’t understand the physical world, still doesn’t understand the psychological world and still hallucinates.””
Our reading: ChatGPT is not reliable, does not understand the physical or psychological world, and hallucinates.
No horizon Evidence as of
Limitation: The result depends on treating understanding as practical reasoning competence.
Assessment: UnresolvedSaid
Forecast
“In retrospectives written in 2043, intellectual historians will conclude that there was an initial overemphasis on large language models, and a gradual but critical shift of the pendulum back to more structured systems with deeper comprehension.”
Our reading: In retrospectives written in 2043, intellectual historians will conclude that there was an initial overemphasis on large language models and a gradual but critical shift of the pendulum back to more structured systems with deeper comprehension.
Not yet due Evidence as of
Limitation: The forecast leaves the author population and degree of agreement unspecified.
Assessment: UnresolvedSaid
Forecast
“Although GPT-4 will definitely seem smarter than its predecessors, its internal architecture remains problematic.”
Our reading: GPT-4 will seem smarter than predecessors but its architecture remains problematic.
No deadline stated Evidence as of
Limitation: The comparative evaluation was conducted by the model's developer.
Assessment: SupportedSaid
Forecast
“GPT-4 will solve many of the individual specific items used in prior benchmarks, but still get tripped up, particularly in longer and more complex scenarios.”
Our reading: GPT-4 will solve many prior benchmark items but still fail, especially on longer, complex scenarios.
Settled before its deadline Evidence as of
Limitation: Benchmark success does not establish whether individual items were learned or memorized.
Assessment: SupportedSaid
Forecast
“As OpenAI has learned, bigger in many ways means better, with outputs more and more humanlike with each iteration. GPT-4 is going to be a monster.”
Our reading: Because scale improves output in many ways, GPT-4 will be very powerful.
Settled before its deadline Evidence as of
Limitation: The decisive evaluation was conducted by the model developer.
Assessment: SupportedSaid
Forecast
“It will still make a significant number of shake-your-head stupid errors, in ways that are hard to fully predict.”
Our reading: GPT-4 will make a significant number of hard-to-predict stupid errors.
Settled before its deadline Evidence as of
Limitation: The judgment depends on reading hard to predict as unexpectedly uneven performance, rather than requiring a direct prediction experiment.
Assessment: UnresolvedSaid
Observation
“Which is to say it makes stuff up, regularly; a fair bit of what it is says simply isn’t true.”
Our reading: ChatGPT regularly makes things up and a fair amount of its output is false.
No horizon Evidence as of
Limitation: This does not mean fabrication was absent or rare.
Assessment: UnresolvedSaid
Forecast
““Scaling” alone—building bigger and models until they absorb the entire internet — will prove useful, but only to a point.”
Our reading: Scaling alone will prove useful only up to a point.
No deadline stated Evidence as of
Limitation: The claim specifies no deadline, capability threshold, or operational test of trustworthy general intelligence.
Assessment: SupportedSaid
Observation
“In short, ChatGPT might sound like the Star Trek computer, but for now, you still can’t trust it.”
Our reading: ChatGPT cannot currently be trusted.
No horizon Evidence as of
Limitation: The conclusion concerns unverified factual reliance, not every possible use.
Assessment: SupportedSaid
Forecast
“Some of its advice will be head-scratchingly bad.”
Our reading: Some of GPT-4's advice will be very bad.
Settled before its deadline Evidence as of
Limitation: The study concerns a particular infectious-disease setting and does not measure all GPT-4 advice.
Assessment: SupportedSaid
Forecast
“When queried on medicine it will either resist answering (if there are aggressive guardrails) or occasionally spout plausible-sounding but dangerous nonsense.”
Our reading: On medical queries GPT-4 will either refuse (with aggressive guardrails) or occasionally produce plausible but dangerous nonsense.
Settled before its deadline Evidence as of
Limitation: The decisive evaluation concerns bloodstream-infection management and the tested GPT-4 version.
Assessment: SupportedSaid
Observation
“Although ChatGPT can write about anything, it is also easily confused.”
Our reading: ChatGPT is easily confused.
No horizon Evidence as of
Limitation: The arithmetic demonstrations were selected to probe weaknesses.
Assessment: SupportedSaid
Forecast
“GPT-4 will still, like its predecessors, be a bull in a china shop, reckless and hard to control.”
Our reading: GPT-4 will be reckless and hard to control like its predecessors.
Settled before its deadline Evidence as of
Limitation: Adversarial test results do not establish an ordinary-use failure rate.
Assessment: SupportedSaid
Observation
“Alas, the genie can no longer be stuffed back in the bottle.”
Our reading: Misinformation-capable LLM technology can no longer be contained or withdrawn.
No horizon Evidence as of
Limitation: Shows availability during the period, not long-term containment.
Source: AI's Jurassic Park moment
Assessment: SupportedSaid
Observation
“Large language models lack mechanisms for verifying truth; we need to find new ways to integrate them with the tools of classical AI, such as databases, webs of knowledge, and reasoning.”
Our reading: LLMs lack mechanisms for verifying truth.
No horizon Evidence as of
Limitation: The result depends on distinguishing standalone models from augmented systems.
Source: AI's Jurassic Park moment
Assessment: SupportedSaid
Observation
“For another, chatGPT, just released by OpenAI, is more or less just as capable of producing similar nonsense, such as instant essays on adding wood chips to breakfast cereal.”
Our reading: ChatGPT is about as capable as Galactica of producing nonsense misinformation.
No horizon Evidence as of
Limitation: The packet found no head-to-head test of ChatGPT and Galactica, so 'more or less just as capable' is judged qualitatively, not measured.
Source: AI's Jurassic Park moment
Assessment: UnresolvedSaid
Causal claim
“they cost almost nothing to operate, and so they are on a path to reducing the cost of generating disinformation to zero.”
Our reading: Because generative AI systems like ChatGPT, Dall-E 2, and Lensa cost almost nothing to operate, they are on a path to reducing the cost of generating disinformation to zero.
No horizon Evidence as of
Limitation: The economically negligible reading of 'zero' is an interpretation.
Source: AI's Jurassic Park moment
Assessment: UnresolvedSaid
Observation
“they can easily be automated to generate misinformation at unprecedented scale.”
Our reading: Generative AI systems like ChatGPT, Dall-E 2, and Lensa can easily be automated to generate misinformation at unprecedented scale.
No horizon Evidence as of
Limitation: The supplied experiment predates the November–December 2022 target period [e1].
Source: AI's Jurassic Park moment
Assessment: UnresolvedSaid
Conditional forecast
“If the website is flooded with worthless code examples, programmers will no longer go there, its database of over 30 million questions and answers will become untrustworthy, and the 14 year old website will die.”
Our reading: If Stack Overflow is flooded with worthless generated code, programmers will leave, its database will become untrustworthy, and the site will die.
No deadline stated Evidence as of
Only if: if the website is flooded with worthless AI-generated code examples
Source: AI's Jurassic Park moment
Assessment: SupportedSaid
Causal claim
“GPT doesn’t talk randomly, because it’s pastiching things actual people said.”
Our reading: GPT's output is not random, because it recombines things real people have said.
No horizon Evidence as of
Limitation: The result depends on reading pastiching as learned recombination rather than literal copying.
Source: How come GPT can seem so brilliant one minute and so breathtakingly dumb the next?
Assessment: UnresolvedSaid
Causal claim
“When GPT sounds plausible, it is because every paraphrased bit that it pastes together is grounded in something that actual humans said, and there is often some vague (but often irrelevant) relationship between..”
Our reading: GPT sounds plausible because each piece it combines comes from something humans actually said, and the pieces are often loosely related.
No horizon Evidence as of
Limitation: The meanings of bit, paraphrase and grounded materially affect the test.
Source: How come GPT can seem so brilliant one minute and so breathtakingly dumb the next?
Assessment: SupportedSaid
Observation
“It could not be immediately applied to the challenge of, say, customer service or guiding the actions of a domestic robot, or well, almost anything else, really.”
Our reading: Cicero cannot be immediately applied to tasks such as customer service or domestic robotics, or almost anything besides blitz Diplomacy.
No horizon Evidence as of
Limitation: The evidence is a developer capability statement, not an independent transfer benchmark.
Source: What does Meta AI’s Diplomacy-winning Cicero Mean for AI?
Assessment: SupportedSaid
Observation
“Strikingly, and in opposition to much of the Zeitgeist, Cicero relies quite heavily on hand-crafting, both in the data sets, and in the architecture; in this sense it is in many ways more reminiscent of classical “Good Old Fashioned AI” than deep learning systems that tend to be less structured, and less customized to particular problems.”
Our reading: Against much of the Zeitgeist, Cicero relies quite heavily on hand-crafting in both its data sets and its architecture, and in this sense it is in many ways more reminiscent of classical Good Old Fashioned AI than of deep learning systems, which tend to be less structured and less customized to particular problems.
No horizon Evidence as of
Limitation: Quite heavily is assessed through substantive design choices, not a measured percentage of effort or performance.
Source: What does Meta AI’s Diplomacy-winning Cicero Mean for AI?
Assessment: UnresolvedSaid
Observation
“As it turns out, the architecture of Cicero differs profoundly from most of what’s been talked about in recent years in AI.”
Our reading: Cicero's architecture differs profoundly from most AI approaches discussed in recent years.
No horizon Evidence as of
Limitation: Selected examples establish differences, not prevalence.
Source: What does Meta AI’s Diplomacy-winning Cicero Mean for AI?
Assessment: SupportedSaid
Observation
“Cicero makes extensive use of machine learning, but is hardly a poster child for simply making ever bigger models (so-called “scaling maximalism”), nor for the currently popular view of “end-to-end” machine learning of in which some single general learning algorithm applies across the board, with little internal structure and zero innate knowledge.”
Our reading: Cicero makes extensive use of machine learning but is hardly a poster child for scaling maximalism or for the popular end-to-end view of machine learning with little internal structure and zero innate knowledge.
No horizon Evidence as of
Limitation: The architecture alone does not establish how much each component contributed to performance.
Source: What does Meta AI’s Diplomacy-winning Cicero Mean for AI?
Assessment: UnresolvedSaid
Conditional forecast
“if Alexa could converse freely, user expectations would go through the roof, and mostly be unmeetable.”
Our reading: If Alexa could converse freely, user expectations would rise to a level that mostly could not be met.
No deadline stated Evidence as of
Only if: if Alexa could converse freely
Assessment: SupportedSaid
Observation
“Bottom line: From the outset Large Language Models like GPT-3 have great at generating surrealist prose, and they can beat a lot of benchmarks, but they are not (and may never be) great tech for reliably inferring user intent from what users say.”
Our reading: LLMs like GPT-3 are not great technology for reliably inferring user intent from what users say.
No horizon Evidence as of
Limitation: The principal evidence is a research benchmark, not a representative household deployment study.
Assessment: UnresolvedSaid
Forecast
“Turning LLMs into a product that controls your home and talks to you in a way that would be reliable enough to use at scale in millions of homes is still a long, long way away.”
Our reading: A reliable LLM-based conversational home-control product usable at scale in millions of homes is still a long, long way away.
No deadline stated Evidence as of
Limitation: Large-scale participation is company-reported.
Assessment: UnresolvedSaid
Observation
“Just like every other large language model I have seen.”
Our reading: Like Galactica, every other large language model the author has seen prevaricates (fabricates content) a lot.
No horizon Evidence as of
Limitation: False answers on deliberately challenging questions are not a direct estimate of fabrication in ordinary use.
Source: A Few Words About Bullshit