“But you don't make clear that much of the bill's requirements are limited to models with training runs of $100 million+. Companies that can afford that, presumably valued in the billions, are not exactly "little-tech."”
Our reading, the sentence that was checked:
Much of SB-1047's requirements are limited to models with training runs of $100 million or more, and companies that can afford such runs, presumably valued in the billions, are not exactly little tech.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, An open letter to Fei-Fei Li concerning California’s proposed AI regulation, SB-1047, paragraph 8.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-an-open-letter-to-fei-fei-li-concerning-d34f41-a3
We recorded 6 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 2024-08-10. 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)
Conditions
- much of the bill's requirements (scope, stated by the speaker)
- presumably valued in the billions (qualifier, stated by the speaker)
From the speaker
- much of the bill's requirements
- models with training runs of $100 million+
- Companies that can afford that
- presumably valued in the billions
- not exactly "little-tech."
Our interpretation
- Assess the proposed bill current when the post appeared, rather than later amendments or its eventual disposition.
- Read 'much' as the main developer safety, assessment, testing and reporting obligations, without requiring every provision to share the threshold.
- Read '$100 million+' as an approximate description of the statutory training-compute cost threshold, rather than a claim about treatment of a model costing exactly $100 million.
- Read the company-size sentence as a qualitative inference about the resources needed to afford such training. 'Presumably' does not firmly assert a verified billion-dollar valuation for every affected developer.
- The target period covers the July 3 version through the post date. This is an observation about a pending legislative text, not a financial reporting period.
- Metric scope is null because the testable monetary threshold defines legislative coverage; no named company's financial results or valuation are being measured.
Counts as supported if: The contemporaneous bill ties its main developer obligations to covered models whose initial training exceeds approximately $100 million in compute costs, supporting the ordinary description of that training as financially substantial. Separate provisions or derivative-model rules do not by themselves defeat 'much.'
Counts as contradicted if: The contemporaneous bill lacks the claimed initial-training cost threshold, or its main developer obligations generally apply to independently trained models below it. Evidence that some smaller firms face indirect effects would not alone contradict the narrower wording.
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 Appropriations analysis for the August 7, 2024 hearing identifies the July 3 version and ties principal safety, testing and compliance duties to covered models [e3]. Foley's July 9 commentary on that version identifies $100 million in training cost as the key threshold [e2]. The preceding Judiciary analysis specifies initial-training thresholds above $100 million and 10^26 operations, alongside a separate fine-tuning route [e1]. Together, these support the factual core that much of the bill concerned models with approximately $100 million-plus initial training costs [e1] [e2] [e3]. The separate fine-tuning route limits any inference that every covered developer must itself afford such an initial run [e1].
Limitations
- The complete July 3 bill text was not directly verified; the finding combines legislative analyses and contemporaneous legal commentary.
- The preceding Judiciary analysis calculates cost using average cloud-compute prices, which does not establish actual cash expenditure [e1].
- The evidence does not verify billion-dollar company valuations, define “little tech,” or establish a company-size exemption. The qualified valuation conjecture is not assessed.
- The evidence does not rule out indirect effects on smaller firms.
Evidence (3 items)
Supports the claim: Foley & Lardner LLP,
training cost of US$100,000,000.
Why it matters here: The article expressly discusses the July 3 amendment and confirms that the $100 million training-cost threshold remained central. Its current page also contains a later veto notice, which is not used here.
- Publisher
- Foley & Lardner LLP
- Kind of source
- Secondary summary
- Describes events of
- Public from
Supports the claim: California Assembly Committee on Appropriations; copy hosted by FastDemocracy,
This bill establishes requirements the developer of an artificial intelligence (AI) model trained using a specified level of computing power (“covered model”) must meet before training and releasing the model, establishes a state entity to provide oversight and guidance regarding covered model compliance, provides related whistleblower protections, grants enforcement authority to the Attorney General and Labor Commissioner, and creates CalCompute, a public cloud computing cluster.
Why it matters here: The committee's account of the relevant version ties the principal developer duties to covered models. Its ensuing list includes assessment, testing, audits, certification and incident reporting.
- Publisher
- California Assembly Committee on Appropriations; copy hosted by FastDemocracy
- Kind of source
- Regulatory filing
- Describes events of
- Public from
Context only: California Assembly Committee on Judiciary; copy hosted by FastDemocracy,
b) An artificial intelligence model created by fine-tuning a covered model using a quantity of computing power equal to or greater than 3 times 10^25 integer or floating-point operations.
Why it matters here: This preceding committee record specifies both the initial-training cost and compute thresholds. Its separate fine-tuning branch cautions against interpreting Marcus as proving that every regulated developer must itself spend $100 million.
- Publisher
- California Assembly Committee on Judiciary; copy hosted by FastDemocracy
- Kind of source
- Regulatory filing
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, An open letter to Fei-Fei Li concerning California’s proposed AI regulation, SB-1047
But you don't make clear that much of the bill's requirements are limited to models with training runs of $100 million+. Companies that can afford that, presumably valued in the billions, are not exactly "little-tech."
- Where
- paragraph 8
- Published
- 2024-08-10
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-an-open-letter-to-fei-fei-li-concerning-d34f41-a3: 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-an-open-letter-to-fei-fei-li-concerning-d34f41-a3.
HeldBy this site
The final-text check failed: citations_correct.
About assessment marcus-an-open-letter-to-fei-fei-li-concerning-d34f41-a2, which was not published.