Jed Brown
banner
jedbrown.org
Jed Brown
@jedbrown.org
Prof developing fast algorithms, reliable software, and healthy communities for computational science. https://hachyderm.io/@jedbrown

https://PhyPID.org | aspiring killjoy | against epistemicide | he/him
"Value" rides exponentials when it's the same company manufacturing the counterfeit legal objections as is "reviewing" those objections.

It's a denial of service attack on public process, and gov adoption of these products is a defiance of public trust.
November 9, 2025 at 7:59 PM
That may be acceptable for the data product (and it's certainly important for provenance), but downstream users of the data product are generally not preserving that. Moreover, professional norms of conduct usually require epistemically-precise citation, which these systems cannot do.
November 9, 2025 at 5:45 AM
Striking parallels with George Washington surrounded by a halo and swastikas at Madison Square Garden in 1939.
m.youtube.com/watch?v=NC1M...
A Night at the Garden | An American Nazi Rally in 1939 | POV
YouTube video by PBS
m.youtube.com
November 9, 2025 at 3:46 AM
Oklo's license application for the Aurora micro-reactor design was denied by the Nuclear Regulatory Commission in 2022.
www.nrc.gov/docs/ML2135/...

They are now in a DOE pilot program waiting for DOE authorization to begin fabrication without needing NRC approval.
www.ans.org/news/2025-09...
November 6, 2025 at 12:20 AM
Publishing to arXiv, Zenodo, or similar with a Creative Commons license is insurance (and wholesome). Either the whole document or a report that at least contains relevant figures, etc. If forced to later sign a copyright assignment/exclusive license (can fight), the CC license remains irrevocable.
November 4, 2025 at 8:27 PM
Note that training can be fair use (as Judge Alsup has ruled) and some uses of a trained model can be fair use (e.g., phrase embeddings for conference scheduling) while other uses would not be (e.g., reproducing an entire book from its first line, arxiv.org/abs/2505.12546).
Extracting memorized pieces of (copyrighted) books from open-weight language models
Plaintiffs and defendants in copyright lawsuits over generative AI often make sweeping, opposing claims about the extent to which large language models (LLMs) have memorized plaintiffs' protected expr...
arxiv.org
November 2, 2025 at 3:23 AM
Reproducing source content (or obfuscations thereof) violates even CC-BY because it doesn't honor the requirement of attribution. That's non-consensual and plagiarism (misconduct in some professions). The legality hinges on whether the output can be considered fair use.
November 2, 2025 at 3:18 AM
Fox has since changed the headline, but they initially ran with those videos as being real.
So there were some AI-generated videos that purported to show Black women ranting about SNAP benefits, and... Fox News is reporting on this like these are real people. They're not! They're AI!

This is truly insane. Total unreality. www.foxnews.com/media/snap-b...
November 2, 2025 at 1:20 AM
Their old definition of AGI was strictly financial ($100B/year profit), which feels pretty unattainable given that everyone is losing money. This may be a way to make it easier to declare AGI to juice waning investor confidence.
October 29, 2025 at 2:38 PM
This is FOMO-driven university administration. It also appears to be abdication of shared governance, as most impacts of this choice will be on matters that faculty are granted principal authority over, yet the article doesn't mention a shared governance process.
October 28, 2025 at 1:28 PM