Jed Brown

"AI" is a speculative financial instrument.

A wealthy company founded on uncompensated labor. The could easily pay for subjects matter experts to write definitions for terms of art, but that would spoil the investors' magic spell.

Jitsi Meet Join a WebRTC video conference powered by the Jitsi Videobridge

I'm also distrustful of the usual suspects, even if it isn't direct recording. Jitsi (meet.jit.si) is dead simple to use, open source, and is probably the best privacy you can get for its usability. Signal has great privacy if your group is already there, albeit limited conferencing features.

AI and Fraternity, Abeba Birhane, AI Accountability Lab I envision a future where human dignity, justice, peace, kindness, care, respect, accountability, and rights and freedoms serve as the north stars that guide AI development and use. Realising these ideals can’t happen without intentional tireless work, dialogues, and confrontations of ugly realities – even if they are uncomfortable to deal with. This starts with deciphering hype from reality. Pervasive narratives portray AI as a magical, fully autonomous entity approaching a God-like omnipotence and omniscience. In reality, audits of AI systems reveal a consistent failure to deliver on grandiose promises and suffer from all kinds of shortcomings, issues often swept under the rug. AI in general, and GenAI in particular, encodes and exacerbates historical stereotypes, entrenches harmful societal norms, and amplifies injustice. A robust body of evidence demonstrates that — from hiring, welfare allocation, medical care allocation to anything in between — deployment of AI is widening inequity, disproportionately impacting people at the margins of society and concentrating power and influence in the hands of few. Major actors—including Google, Microsoft, Amazon, Meta, and OpenAI—have willingly aligned with authoritarian regimes and proactively abandoned their pledges to fact-check, prevent misinformation, respect diversity and equity, refrain from using AI for weapons development, while retaliating against critique. The aforementioned vision can’t and won’t happen without confrontation of these uncomfortable facts. This is precisely why we need active resistance and refusal of unreliable and harmful AI systems; clearly laid out regulation and enforcement; and shepherding of the AI industry towards transparency and accountability of responsible bodies. "Machine agency" must be in service of human agency and empowerment, a coexistence that isn't a continuation of modern tech corporations’ inequality-widening,

so I am one of the 12 people (including the “god-fathers of AI”) that will be at the Vatican this September for a two full-day working group on the Future of AI

here is my Vatican approved short provocation on 'AI and Fraternity' for the working group

Two years later, this holds up very well.

Yup, and this is why they need a surveillance state, receiverships, "AI" capturing agency from qualified people, and a rebirth of HUAC-style innuendo-driven tactics to compel voluntary compliance with their revanchist ideology. A courageous opposition would insist on alternate power structures.

Artificial Intelligence and Academic Professions Executive SummaryEducational technology, or ed-tech, including artificial intelligence (AI), continues to become more integrated into teaching and research in higher education, with minimal oversight.

I feel this greatly amid the pandering to an overtly corrupt and hostile agenda. Does your institution have a union? I've found that to be a community with common cause, offering both catharsis and an alternate power center. And this is a worthy statement to rally behind www.aaup.org/reports-publ...

One study in particular showed that student evaluations are good predictors of learning. But there’s a twist. The correlation between student evaluations and quality of learning is negative. The higher the instructor’s score in the student evaluations, the worse the learning; and the lower the evaluation score, the better the learning. (Feel free to read that again. There is no typo.)

Indeed, there is evidence from randomized controlled trials of large introductory classes, that student course evaluations are negatively correlated with how well students learned the material.

maa.org/math-values/...

At this point we're basically talking about a special-purpose search engine for queries worded as false claims. Whether an "LLM" helps or harms that embedding/query would need to be tested, but is quite in the weeds.

Yes, that's a way to have instant refutation of misinformation (not necessarily matching every detail of the misinformed claim). The reach of a bot firing off community notes on social media is unclear. Maybe better to give a journalist leads for framing in real-time, and who to seek quotes from.

Countering the Gish gallop Mehdi Hasan, a British journalist, suggests using three steps to beat the Gish gallop:[5] Because there are too many falsehoods to address, it is wise to choose one as an example. Choose the weakest, dumbest, most ludicrous argument that the galloper has presented and tear that argument to shreds ("the weak point rebuttal"). Do not budge from the issue or move on until having decisively destroyed the nonsense and clearly made the counter point. Call out the strategy by name, saying: "This is a strategy called the 'Gish Gallop'—do not be fooled by the flood of nonsense you have just heard." Generally, it is more difficult to use the Gish gallop in a structured debate than a free-form one.[6] If a debater is familiar with an opponent who is known to use the Gish gallop, the technique may be countered by pre-empting and refuting the opponent's commonly used arguments before the opponent has an opportunity to launch into the Gish gallop.[7]

In other words, we cannot counter the Gish Gallop by systematically and precisely refuting each false claim. To do that faster and at greater volume (which is the most an LLM can offer, albeit with a higher error rate) is the opposite of effective strategy.
en.wikipedia.org/wiki/Gish_ga...

We face a denial of service attack on the epistemic foundations of knowledge, but I don't think it can be won by turning to the tools of fraud and disinformation to mass-produce low-quality refutation that is nominally "aligned" to the science. This is an existential public communication challenge.

It takes more effort to revise LLM output to meet scientific standards of integrity than to write it yourself. The risks and consequences for errors are high. The technosolutionism makes it harder to communicate epistemic soundness and creates a race to the bottom with both sides flooding the zone.

I spent seven years writing a book on this and she just… said it in three succinct paragraphs to politico.

And if you think a rough draft that seems plausible gets you most of the way there, consider that this sort of thing keeps happening. Scientists putting their name on LLM output even after diligent review will lead to such errors, and undermine public confidence in science.
bsky.app/profile/jedb...

That's a fine standard for a botnet arguing on social media, but not for a scientist of integrity to put their name on. The authors of the HHS report (disinformation on gender-affirming care) that was full of LLM slop did not face consequences.
bsky.app/profile/jedb...

No. LLMs may get the gist right (by remixing the form of scholarship in the training corpus), but will have mistakes and improper citations that make low-hanging fruit for counter-attacks. LLMs amplify Brandolini's Law by orders of magnitude because of the disparate burdens of proof and integrity.

Scientists Say New Government Climate Report Twists Their Work A new Department of Energy report “fundamentally misrepresents” climate research and leaves out key context, multiple scientists cited in the report tell WIRED.

NEW: i spoke to nine (!) scientists across several disciplines whose work was cited in the new Department of Energy report that downplays the severity of climate change. all of them say their work was misrepresented, cherry picked, and/or lacked context —

B. Prohibited Use of Proxies for Protected Characteristics l. \ilhat Constitutes Unlawful Proxies? Unlawful proxies occur when a federally funded entity intentionally uses ostensibly neutral criteria that function as substitutes for explicit consideration of race, sex, or other protected characteristics. While these criteria may appear facially neutral, they become legally problematic under any of the following circumstances: They are selected because they correlate with, replicate, or are used as substitutes for protected characteri stics. o They are implemented with the intent to advantage or disadvantage individuals based on protected characteristics. 2. Examples of Potentially Unlawful Proxies "Cultural Competence" Requirements: A federally funded university requires job applicants to demonstrate'ocultural competence," "lived experience," or "cross-cultural skills" in ways that effectively evaluate candidates' racial or ethnic backgrounds rather than objective qualifications. This includes selection criteria that advantage candidates who have experiences the employer associates with certain racial groups. For instance, requiring faculty candidates to describe how their "cultural background informs their teaching" may ñrnction as a proxy if used to evaluate candidates based on race or ethnicity. Geographic or Institutional Targeting: A federally funded organization implements recruitment strategies targeting specific geographic areas, institutions, or organizations chosen primarily because of their racial or ethnic composition rather than other legitimate factors. "Overcoming Obstacles" Narratives or ttDiversity Statements": A federally funded progr¿ìm requires applicants to describe o'obstacles they have overcome" or submit a "diversity statement" in a manner that advantages those who discuss experiences intrinsically tied to protected characteristics, using the narrative as a proxy for advantaging that protected characteristic in providing benefits.

I recall an entire field of study examining ways in which ostencibly colorblind policies use proxies for race to perpetuate racial discrimination. I recall it being critical of these power structures. Yet conservatives were not fond of this theory. Seems important.
www.justice.gov/ag/media/140...

The House UnAmerican Activities Committee is one parallel. It operated for 30 years — and was most effective *when its targets cowered*.

If you are FOIA’d you have options: you can say “fk you” to these 🤡s by posting your syllabus yourself. Don’t give them the thrill of a ‘secret documents’ story!

Age verification is censorship and a denial of privacy rights.

“Univ. of Illinois report shows that a large humanities department like English produces substantial net profit; units like engineering and agriculture run at a loss. The widely respected Delaware Study of Instructional Costs and Productivity shows the same pattern.” /7 thi.ucsc.edu/the-humaniti...

"I think it is only honor that will save us. It is honor that causes people to do the irrational things that are good—that separates the fighters from those who flee in the face of danger. Unfortunately, this sort of honor is not something on which we select for our institutional leaders."

This was also true of the House Unamerican Activities Committee, which relied on voluntary obedience by universities to inflict its desired suppression of First Amendment rights and academic freedom.

www.lawfaremedia.org/article/proc...

Possibly assign partners and have students discuss in-person/in-class how lenses from class apply to their chosen books. Actual human interaction in which you are asserting to have read and thought about the material feels a lot more fake when it's based on LLM output/summaries.

I don't see the pedagogical value of such an assignment beyond students learning that LLM output is very mid. Perhaps once. I disagree with "the only way out is through" -- that transfers way too much agency away from students.

It also means the student isn't learning to frame their own arguments, degrading their ability to persuade and to challenge power. The results across the class will be more formulaic, reflecting less creativity and individuality.

As in previous EOs, "AI" in education/gov is about power to implement segregation, to erase people's humanity. You can't just tell all teachers & fed workers/contractors to do this bigotry, but LLM owners will try. "Quality" will suffer, but it'll still cause harm.
www.whitehouse.gov/presidential...

The issue is that one needs transparency in process to achieve a high standard of proof that obfuscated plagiarism is © infringement. If you hire 1000 people to plagiarize, there will be ample records and testimony available. A proprietary plagiarism machine hides that in quasi-religious debate.

Do not let any LLM serve as a form of epistemic grounding for society.