Pundit ClaimsAI claims by Gary Marcus and Ed Zitron, with the evidence

Ed Zitron Where's Your Ed At Published

Godot Isn't Making it

We recorded 33 claims from this post. 23 were checked against evidence. 10 can’t be checked. 22 more are held. Read the post at its source.

Claims from this post (33)

In the order they appear in the post. Each quotation is the speaker’s words in this post. A label such as “Supported” or “Can’t be checked” is the claim’s current assessment; its page has the evidence and the reasons.

  1. Assessment: UnresolvedSaid

    “Throughout these pieces I have repeatedly made the point that — separate to any lack of a core value proposition, training data drought, or unsustainable economics — generative AI is a dead end due to the limitations of probabilistic models that hallucinate, where they authoritatively state things that aren't true.”

    What we tested: Generative AI is a dead end because probabilistic models hallucinate, stating false things authoritatively.

  2. Assessment: MixedSaid

    “The hallucination problem is one that is nowhere closer to being solved — and, at least with the current technology — may never go away, and it makes it a non-starter for a great many business tasks, where you need a high level of reliability.”

    What we tested: The hallucination problem is nowhere closer to being solved, may never go away at least with current technology, and makes generative AI a non-starter for a great many business tasks that need high reliability.

    What held: Testing on May 23–27 found frequent hallucinations in a specialized legal research product. This supports the barrier to unverified, high-reliability legal work under the stated interpretation, but does not establish the number of business tasks excluded. (arXiv, )

    What did not: Table 2 reports SimpleQA hallucination rates of 0.61 for GPT-4o and 0.44 for o1-preview, with improvements on the other listed tests. The same paragraph reports contrary anecdotal feedback and cautions against generalizing beyond these evaluations. (OpenAI, )

  3. Assessment: ContradictedSaid

    “Generative AI's products have effectively been trapped in amber for over a year.”

    What we tested: Generative AI products have not meaningfully progressed for over a year.

    What happened: Developers could use the API to have Claude inspect screens, move a cursor, click and type. This materially extends possible workflows, although Anthropic describes the release as experimental and error-prone. (Anthropic, )

  4. Assessment: Can’t be checkedSaid

    “Microsoft is cherry-picking a number based on non-specific criteria and claiming it's a big deal, when it's actually pretty pathetic considering its capital expenditures will likely hit over $60 billion in 2024.”

    What we tested: Microsoft's capital expenditures will likely exceed $60 billion in 2024, making its $10B AI revenue run-rate pathetic.

  5. Assessment: SupportedSaid

    “OpenAI currently spends $2.35 to make $1, and Microsoft CFO Amy Hood said that OpenAI would cut into Microsoft's profits this October, losing it a remarkable $1.5 billion, "mainly because of an expected loss from OpenAI" according to CNBC.”

    What we tested: OpenAI currently spends $2.35 for every $1 of revenue it makes.

    What happened: These estimates imply $8.7 billion of expenses and a ratio rounding to $2.35. CNBC also reports exclusions, including equity-based compensation, so this supports the cited estimate rather than comprehensive accounting expenses. (CNBC, )

  6. Assessment: UnresolvedSaid

    “Instead, big tech has found itself peddling increasingly-more-expensive iterations of near-identical Large Language Models — a direct result of them all having to use the same training data, which it’s now running out of.”

    What we tested: Big tech's LLMs are increasingly expensive, near-identical iterations because they all use the same training data, which is running out.

  7. Assessment: UnresolvedSaid

    “Generative AI is incredibly unprofitable.”

    What we tested: Generative AI is incredibly unprofitable.

  8. Assessment: UnresolvedSaid

    “OpenAI, the biggest player in the industry, is on course to lose more than $5 billion this year, with competitor Anthropic (which also makes its own transformer-based model, Claude) on course to lose more than $2.7 billion this year.”

    What we tested: Anthropic is on course to lose more than $2.7 billion in 2024.

  9. Assessment: ContradictedSaid

    “The entire tech industry has become oriented around a dead-end technology that requires burning billions of dollars to provide inessential products that cost them more money to serve than anybody would ever pay.”

    What we tested: Generative AI is a dead-end technology whose inessential products cost more to serve than anyone would pay.

    What happened: The preceding sentence identifies 2024 revenue of $4 billion and inference compute and hosting costs of $2.4 billion. This reported positive serving margin supplies a counterexample to the categorical assertion, without establishing company-wide profitability. (Epoch AI, )

  10. Assessment: Can’t be checkedSaid

    “While one can dazzle investors with buzzwords and charts, the laws of physics are a far harsher mistress, and if NVIDIA is struggling mere months before the first installations are to begin, it's unclear how it practically launches this generation of chips, let alone continues its yearly cadence.”

    What we tested: Given its cooling struggles, it is unclear NVIDIA can practically launch Blackwell, let alone maintain a yearly chip cadence.

  11. Assessment: Can’t be checkedSaid

    “The fact that NVIDIA is having such significant difficulties with thermal performance is very, very bad.”

    What we tested: It is very, very bad that NVIDIA is having such significant difficulties with thermal performance on its Blackwell racks.

  12. Assessment: UnresolvedSaid

    “If NVIDIA doesn’t solve its cooling problem — and solve it well — its customers will undoubtedly encounter thermal throttling, where the chip reduces speed in order to avoid causing permanent damage.”

    What we tested: If NVIDIA does not solve Blackwell's cooling problem well, customers will encounter thermal throttling.

  13. Assessment: Can’t be checkedSaid

    “While NVIDIA might be able to fix this problem in isolation within its racks, it remains to be seen how this works at scale as they ship and integrate hundreds of thousands of Blackwell GPUs starting in the front half of 2025.”

    What we tested: Even if NVIDIA fixes cooling within its racks, it is uncertain the fix will work at scale as hundreds of thousands of Blackwell GPUs ship from H1 2025.

  14. Assessment: UnresolvedSaid

    “Remember, the capital expenditures on generative AI and the associated infrastructure have gone over $200 billion in the last year.”

    What we tested: Capital expenditures on generative AI and infrastructure exceeded $200 billion in the last year.

  15. Assessment: Can’t be checkedSaid

    “While some have claimed that "agents are the next frontier," the reality is that agents may be the last generative AI product — multiple Large Language Models and integrations bouncing off of each other in an attempt to simulate what a human might do at a cost that won't be sustainable for the majority of businesses.”

    What we tested: Agents may be the last generative AI product, and their cost won't be sustainable for most businesses.

  16. Assessment: Can’t be checkedSaid

    “And worse still, nothing has really come out of this movement.”

    What we tested: Nothing has really come out of the generative AI movement.

  17. Assessment: UnresolvedSaid

    “If we're truly at the diminishing returns stage of transformer-based models, it will be extremely difficult to justify buying further iterations of NVIDIA GPUs past Blackwell.”

    What we tested: If transformer models are at diminishing returns, it will be extremely difficult to justify buying NVIDIA GPUs beyond Blackwell.

  18. Assessment: ContradictedSaid

    “The models aren't going to get better.”

    What we tested: Even with racks of Blackwell GPUs, generative AI models are not going to get better.

    What happened: The report identifies MAI-Thinking-1's base-model training on 8K GB200 GPUs and reports 84.2% GPQA Diamond pass@1, exceeding the pre-post results in e3. It uses the same 198-question Diamond subset; its large output budget and grading procedure limit exact comparability. (Microsoft AI, )

  19. Assessment: Can’t be checkedSaid

    “Sam Altman has grown rich and powerful lying about how GPT will somehow lead to AGI, but at this point, what exactly is OpenAI meant to do?”

    What we tested: Sam Altman's claims that GPT will lead to AGI are lies.

  20. Assessment: UnresolvedSaid

    “The constant refrain I hear from VCs and AI fantasists is that "chips will bring down the cost of inference," yet I don't see any proof of that happening, nor do I think it'll happen quickly enough for these companies to turn things around.”

    What we tested: Chips will not bring down inference costs quickly enough for AI companies to turn their finances around.

  21. Assessment: UnresolvedSaid

    “Putting that aside, OpenAI is also a terrible business that has to burn $5 billion to make $3.4 billion, with no proof that it’s capable of bringing down costs.”

    What we tested: OpenAI is a terrible business burning $5 billion to make $3.4 billion, with no proof it can reduce costs.

  22. Assessment: UnresolvedSaid

    “Even if OpenAI listens to its better angels, the reality still stands: ads won’t dam the rising tide of red ink that promises to eventually drown the company.”

    What we tested: Advertising will not stop OpenAI's losses, which will eventually drown the company.

  23. Assessment: Can’t be checkedSaid

    “Even if he ultimately fails, should Musk manage to drag the process out, or delay it temporarily, it could strike a terminal blow for OpenAI.”

    What we tested: If Musk delays OpenAI's for-profit conversion, it could be a terminal blow for OpenAI.

  24. Assessment: Can’t be checkedSaid

    “Outside of a miracle, we are about to enter an era of desperation in the generative AI space.”

    What we tested: Barring a miracle, the generative AI industry is about to enter an era of desperation.

  25. Assessment: ContradictedSaid

    “The products aren't coming.”

    What we tested: Meaningful generative AI products are not coming.

    What happened: The disclosure dates general availability to May 2025 and reports continued growth in weekly users and business subscriptions in February 2026. Its reported commercial scale supplies evidence that adoption extended beyond an experimental launch. (Anthropic, )

  26. Assessment: UnresolvedSaid

    “"Orion," OpenAI's next model, will underwhelm, as will its competitors' models, and at some point somebody is going to blink in one of the hyperscalers, and the AI era will be over.”

    What we tested: At some point a hyperscaler will pull back and the AI era will end.

  27. Assessment: UnresolvedSaid

    “The revenue isn't coming.”

    What we tested: The revenue expected from generative AI is not coming.

  28. Assessment: UnresolvedSaid

    “Once the AI bubble pops, there are no other hyper-growth markets left, which will in turn lead to a bloodbath in big tech stocks as they realize that they’re out of big ideas to convince the street that they’re going to grow forever.”

    What we tested: When the AI bubble pops, the lack of other hyper-growth markets will cause a bloodbath in big tech stocks.

  29. Assessment: UnresolvedSaid

    “There is no iPhone moment coming, I'm afraid.”

    What we tested: Generative AI will not have a transformative iPhone moment.

  30. Assessment: Can’t be checkedSaid

    “Generative AI isn’t a revolution, it’s an evolution of a tech industry overtaken by growth-hungry management consultant types that neither know the problems that real people face nor how to fix them.”

    What we tested: Generative AI is not a revolution but a product of a growth-obsessed tech industry.

  31. Assessment: ContradictedSaid

    “I also want to be clear that none of these companies ever had a plan.”

    What we tested: None of the big AI companies ever had a plan for generative AI beyond throwing GPUs at it.

    What happened: Microsoft specified concrete generative AI functions inside Word and other work applications. That is a product and deployment strategy beyond purchasing GPUs, regardless of whether the promised benefits were achieved. (Microsoft, )

  32. Assessment: UnresolvedSaid

    “And before then, you’re going to get a lot of desperate stories about how “AI gains can be found outside of training new models” to try and keep the party going, despite reality flicking the lights on and off and threatening to call the police.”

    What we tested: Before hyperscalers pull back, there will be many desperate stories claiming AI gains can come from outside training new models.

  33. Assessment: UnresolvedSaid

    “I believe all of this is going nowhere, and that at some point Google, Microsoft, or Meta is going to blink and pull back on their capital expenditures.”

    What we tested: Generative AI is going nowhere, and at some point Google, Microsoft or Meta will pull back on AI capital expenditures.

Held from this post (22)

These records did not pass every check, so nothing about them is published. Held is not a finding.