“The trouble is that the bigger the model, the higher the internal costs. And current-sized models, already expensive, aren’t reliable enough for many jobs.”
Our reading, the sentence that was checked:
Bigger models have higher costs, and current-sized models are already expensive and not reliable enough for many jobs.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, When Will the GenAI Bubble Burst?, paragraph 15.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-when-will-the-genai-bubble-burst-2b9be3-a1
We recorded 19 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2024-03-31. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Observation
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No horizon
- Deadline state on 2026-10-29
- No horizon (the release’s reference date)
From the speaker
- the bigger the model, the higher the internal costs
- current-sized models, already expensive
- aren’t reliable enough for many jobs
Our interpretation
- I use the first quarter of 2024 through the post date as the current-model observation window; this is not a company fiscal period.
- The ordinary engineering reading is that increasing model size generally increases computing requirements when architecture, workload, hardware and implementation are comparable.
- Already expensive means substantial training or operating resource requirements, not a specified dollar amount.
- Many jobs means multiple practical task classes with consequential accuracy requirements, rather than all jobs or a majority of occupations.
- Insufficient reliability means that unverified outputs cannot safely be relied upon; it does not exclude useful supervised assistance.
- This is a general technical cost claim, not a claim about a named company's financial accounts.
Counts as supported if: Contemporaneous technical evidence establishes substantial resource requirements that generally increase with comparable model size, while practical evaluations identify material reliability limitations across multiple task classes.
Counts as contradicted if: Comparable size-cost measurements undermine the general relationship, or broad practical evaluations establish adequate reliability without verification across the task classes at issue.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Supported
The dated evidence matches what the claim committed to, under the criteria frozen before grading.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- No horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
The ordinary, qualified reading is supported. Early-2024 technical analysis links larger comparable models to substantial computing demands. Contemporaneous legal and clinical assessments identify practical reliability limits requiring caution and verification. This does not establish that every larger model costs more than every smaller one, or that supervised LLM applications cannot be useful.
Limitations
- The result depends on comparing otherwise similar implementations and interpreting reliability as suitability for unverified use.
- Sparse architectures and optimization prevent a universal ordering by total parameter count.
- The evidence establishes resource burdens, not a particular company's dollar costs.
- The task studies do not estimate the proportion of all occupations affected.
- Suitability-assessment dates are distinguished from the dates of underlying experiments.
Evidence (4 items)
Supports the claim: arXiv,
large parameter count
Why it matters here: The contemporaneous engineering assessment supports the general size-resource relationship and describes substantial deployment demands. It also explains that sparse architectures can weaken the relationship between total parameters and computation.
- Publisher
- arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Supports the claim: arXiv,
legal tasks.
Why it matters here: This January assessment identifies reliability barriers to unsupervised legal research and related tasks. The date refers to the authors' contemporaneous suitability assessment, not a claim that every underlying experiment occurred that day.
- Publisher
- arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Supports the claim: Nature Communications,
some weaknesses
Why it matters here: The March assessment identifies remaining weaknesses across clinical decision-support tasks, while also reporting improvements. Its suitability warning provides a second practical domain; it does not establish that all medical assistance is unreliable.
- Where
- Abstract, conclusion
- Publisher
- Nature Communications
- Kind of source
- Peer-reviewed paper
- Describes events of
- Public from
Context only: arXiv,
13B active parameters
Why it matters here: The architecture described in January shows why total parameter count is not a universal operating-cost measure. This qualifies the size claim without reversing the comparison between otherwise comparable models.
- Where
- Abstract
- Publisher
- arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, When Will the GenAI Bubble Burst?
The trouble is that the bigger the model, the higher the internal costs. And current-sized models, already expensive, aren’t reliable enough for many jobs.
- Where
- paragraph 15
- Published
- 2024-03-31
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-when-will-the-genai-bubble-burst-2b9be3-a1: Current. Outcome: Supported. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.
About assessment marcus-when-will-the-genai-bubble-burst-2b9be3-a1.