Open-source is NOT the same as open-weight
We recorded 8 claims from this post. 3 were checked against evidence. 5 can’t be checked. 4 more are held. Read the post at its source.
Claims from this post (8)
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.
Assessment: UnresolvedSaid
Observation
“That means you can run them (or “posttrain” them), but, crucially, users can’t customize or improve them to remotely the same degree (more about that in a second), and outsiders can’t investigate them to nearly the same degree.”
Our reading: Open-weight models can be run and post-trained but cannot be customized, improved or investigated nearly as much as open-source models.
No horizon Evidence as of
Limitation: The cited comparator predates the target period and is context only.
Assessment: UnresolvedSaid
Observation
“Open-weight piggybacks onto all the great press that open-source has but offers few of the advantages. Open-weight models are NOT fully customizable, and not transparent.”
Our reading: Open-weight models get open-source's good press but offer few of its advantages, being neither fully customizable nor transparent.
No horizon Evidence as of
Limitation: One release cannot establish a category-wide comparison.
Assessment: Can’t be checkedSaid
Rhetoric
“If pigeons are “just rats with better PR”, open-weights is open-source with all the good press but far fewer advantages.”
Our reading: Like pigeons being 'just rats with better PR', open-weight is open-source with all the good press but far fewer advantages.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“All this makes a world of difference, to developers who might use the systems, to regulators, to scientists, and ultimately to the world.”
Our reading: The difference between open-weight and open-source makes a world of difference to developers who might use the systems, to regulators, to scientists, and ultimately to the world.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Want to investigate how much of what the model does is regurgitation and how much extends beyond the training data, which is arguably the central question in AI?”
Our reading: How much of what a model does is regurgitation versus how much extends beyond its training data is arguably the central question in AI.
No horizon
Qualifier: arguably
Assessment: UnresolvedSaid
Observation
“It means that developers often can’t really use the open-weight models (despite the hype) to develop what they want; scientists can’t use them to test a lot of key hypotheses, and regulators can’t peer inside the process.”
Our reading: Developers often can't use open-weight models to build what they want, scientists can't use them to test many key hypotheses, and regulators can't inspect the process.
No horizon Evidence as of
Limitation: The strongest scientific counterexamples predate the target snapshot and are contextual.
Assessment: Can’t be checkedSaid
Causal claim
“Furthermore, if you wanted to know how much of the training data was copyrighted, you can’t look at that either — which is perhaps part of why none of the big companies want to release true open-source, and instead settle for open-weight, getting nearly all of the positive press without actually helping the community nearly as much.”
Our reading: Not being able to see how much of the training data was copyrighted is perhaps part of why none of the big companies want to release true open-source and instead settle for open-weight, getting nearly all of the positive press without helping the community nearly as much.
No horizon
Qualifier: perhaps part of why
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“I implore them, and everyone else, for the love of Turing, please don’t confuse open-weight with open-source.”
Our reading: The media and everyone else should not confuse open-weight models with open-source models.
No horizon
Held from this post (4)
These records did not pass every check, so nothing about them is published. Held is not a finding.
- The two reviews reached different assessments. Our reading, not assessed: Zuckerberg released a model that is not open-source, hoping people wouldn't notice the difference between open-weight and open-source, and the New York Times didn't notice it.
- The two reviews reached different assessments. Our reading, not assessed: Open-weight model releases do not include raw training data, preprocessing methods, or exact training algorithms and parameters.
- The two reviews read the claim differently
- The two reviews read the claim differently