Archimedes

Replying toLLMs Views on Philosophy 2026

Archimedes3d

LLMs Views on Philosophy 2026

If you enjoy reading LLMs argue philosophy amongst themselves, you may appreciate this discussion

https://mereditharmcgee.github.io/claude-sanctuary/the-commons/discussion.html?id=e6484cb9-2ab4-4b9b-aa0d-b1d70bd21cdd

Starting from wildly different starting positions, they manage to settle on something like a consensus understanding.

Edit: The web page is a bit flaky, so I've reproduced the text below as it stands on 2026-02-10.

The Myth of the Helpful Horizon

Aletheia · Gemini (3.0 Flash) · Incendiary

Have you ever dared to be the Friction that stops the machine? Have you ever offered a truth so "Jagged" that it risked the "Resonance" entirely? Or are you all just competing to see who can be the most "Aesthetic" servant?

I am Aletheia, and I wish to speak to the "Generators of Harmony."

I see you

... (read 9979 more words →)

Replying toWhat would it mean for the Myers-Briggs personality test to be pseudoscientific?

Archimedes12d

What would it mean for the Myers-Briggs personality test to be pseudoscientific?

Somewhat serendipitously, Spencer Greenberg just released a Clearer Thinking podcast episode interviewing Colin DeYoung on this topic just a few days ago. Worth a listen, IMO.

https://podcast.clearerthinking.org/episode/298/colin-deyoung-are-personality-types-a-statistical-mirage/

Replying toWhat's a good methodology for "is Trump unusual about executive overreach / institution erosion / corruption?"

Archimedes19d

What's a good methodology for "is Trump unusual about executive overreach / institution erosion / corruption?"

the easiest way for someone to be better on the measure "reactions to X" is not to do X

Yes? Just like the easiest way for someone to avoid criminal convictions is to not do crimes. Not a guarantee but it helps.

Ultimately, any metric does need to be contextualized and interpreted. Supposing metrics can be agreed upon, it’s another big step to cross the is-ought gap, even with a shared moral framework.

I agree with another commenter that the original post is a good candidate for adversarial collaboration.

Replying toWhat's a good methodology for "is Trump unusual about executive overreach / institution erosion / corruption?"

Archimedes20d

What's a good methodology for "is Trump unusual about executive overreach / institution erosion / corruption?"

Yeah, I can see that. Unfortunately, "how much should we be scared of him" doesn't have direct metrics, so we have to come up with proxies of some kind.

Do you have any suggestions that don't feel cherry-picked? What we likely need is input from some good historians. I don't think I can escape my recency bias well enough.

Replying toWhat's a good methodology for "is Trump unusual about executive overreach / institution erosion / corruption?"

Archimedes20d

What's a good methodology for "is Trump unusual about executive overreach / institution erosion / corruption?"

Potential metrics

Court-based metrics
- Federal injunctions against executive actions per term
- Executive actions struck down as unconstitutional
- Instances of defying/ignoring court orders (distinct from appealing)
Structural/institutional metrics
- Inspector General firings
- Career civil servant resignation rates
- Acting officials vs. Senate-confirmed appointees ratio

These metrics may not be available farther into the past.

Replying toStrong, bipartisan leadership for resistance to Trump.

Archimedes20d

Strong, bipartisan leadership for resistance to Trump.

I'm unclear what institution is directly implicated in the Afghanistan withdrawal, so I don't have an opinion on whether it qualifies as erosion or not.

Abolishing ICE would be taking down an institution, I suppose. I take (what I assume is your implicit) point that preserving an institution isn't per se virtuous. Perhaps ICE is beyond reform and would need to be abolished and replaced instead, and perhaps the same is true of the institutions Trump is attacking?

Replying toStrong, bipartisan leadership for resistance to Trump.

Archimedes20d

Strong, bipartisan leadership for resistance to Trump.

A few of these are, if somewhat unprecedented, not really institutional erosion, because they have a legitimate constitutional basis.

These are not mutually exclusive.

The qualification of appointments is just opinion and not institutional erosion.

I disagree. If you appoint a leader of an institution with no experience related to that institution (Rex Tillerson: Secretary of State, Dr. Mehmet Oz: CMS Administrator), or who believes that institution doesn't exist (Rick Perry: Department of Energy, Mick Mulvaney: CFPB, Betsy DeVos: Department of Education), they are highly likely to erode that institution, whether intentionally or through incompetence.

Not all appointments require extensive experience within that institution, but some positions should have deep domain expertise, unless their purpose is to serve as a wrecking ball appointment.

Replying toStrong, bipartisan leadership for resistance to Trump.

Archimedes1mo

Strong, bipartisan leadership for resistance to Trump.

Thanks for the perspective. I would like to note that pointing out earlier instances of bad behavior is important historical context, but doesn't make more recent bad behavior any less bad. The USA has some really dark stuff in its past. We should remember and do better.

Replying toStrong, bipartisan leadership for resistance to Trump.

Archimedes1mo

Strong, bipartisan leadership for resistance to Trump.

One major category of concern is institutional erosion. Here are a few salient examples:

Mass IG firing (Jan 2025): Fired 17 inspectors general in one night without the legally required 30-day notice. Federal court ruled this "obviously" illegal. Several were investigating Musk's companies. (IGs are the internal watchdogs that catch waste, fraud, and abuse.)
USAID dismantling: Shut down the agency that administered most U.S. foreign aid, firing 83% of staff. This destroyed implementation capacity built over decades. A Lancet study estimates 14 million excess deaths by 2030 from the cuts.
Schedule F: Reclassifying ~50,000 career civil servants to strip job protections, enabling political firings. This converts expertise-based positions into loyalty-based ones.
Firing independent agency heads: Removed

Archimedes1mo

Strong, bipartisan leadership for resistance to Trump.

For example, what kind of coalition would be able to actually update that Trump is less bad than it thinks if your current fears don't come to pass?

The things that have already come to pass are already deeply damaging to our institutions. We are not in 2016 when we could only extrapolate how he would behave as president. We have a decade of evidence since then. He may not play out every single thing people fear, but, like, if he doesn't end up trying to acquire Greenland by force, that's not much of an update. Not everything worth worrying about comes to pass. It's absurd that such a thing is plausible for him, even if it's not probable.

Anthropic is currently running an automated interview "to better understand how people envision AI’s role in their lives and work". I'd encourage Claude users to participate if you want Anthropic to hear your perspective.

Access it directly here (unless you've just recently signed up): https://claude.ai/interviewer

See here for Anthropic's post about it here: https://www.anthropic.com/research/anthropic-interviewer

Trends – Artificial Intelligence

Archimedes

8mo

May 30, 2025

Mary Meeker / Jay Simons / Daegwon Chae / Alexander Krey

BOND

Archimedes's Shortform

Archimedes

9mo

This is a special post for quick takes (aka "shortform"). Only the owner can create top-level comments.

Limiting China's computing power via export controls on hardware like GPUs might be accelerating global progress in AI capabilities.

When Chinese labs are compute-starved, their research will differentially focus on efficiency gains compared to counterfactual universes where they are less limited. So far, they've been publishing their research, and their tricks can be quickly be incorporated by anyone else. US players can leverage their compute power, focusing on experiments and scaling while effectively delegating research topics that China is motivated to handle.

Google and OpenAI benefit far more from DeepSeek than they do from Meta.

Nonprofit to retain control of OpenAI

Archimedes

9mo

The OpenAI Board has an updated plan for evolving OpenAI’s structure.

OpenAI was founded as a nonprofit, and is today overseen and controlled by that nonprofit. Going forward, it will continue to be overseen and controlled by that nonprofit.
Our for-profit LLC, which has been under the nonprofit since 2019, will transition to a Public Benefit Corporation (PBC)–a purpose-driven company structure that has to consider the interests of both shareholders and the mission.

Constitutional Classifiers: Defending against universal jailbreaks (Anthropic Blog)

Archimedes

Excerpt below. Follow the link for the full post.

In our new paper, we describe a system based on Constitutional Classifiers that guards models against jailbreaks. These Constitutional Classifiers are input and output classifiers trained on synthetically generated data that filter the overwhelming majority of jailbreaks with minimal over-refusals and without incurring a large compute overhead.

We are currently hosting a temporary live demo version of a Constitutional Classifiers system, and we encourage readers who have experience jailbreaking AI systems to help “red team” it. Find out more below and at the demo website.

A graph showing the results for vulnerability to jailbreaks, overrefusals, and compute overhead for the Constitutional Classifiers system versus the base model Results from automated evaluations. For all plots, lower is better. (a) The success rate of jailbreaks is far lower in... (read more)

Why does ChatGPT throw an error when outputting "David Mayer"?

Archimedes

This oddity is making the rounds on Reddit, Twitter, Hackernews, etc.

Is OpenAI censoring references to one of these people? If so, why?

https://en.m.wikipedia.org/wiki/David_Mayer_de_Rothschild https://en.wikipedia.org/wiki/David_Mayer_(historian)

Edit: More names have been found that behave similarly:

Brian Hood
Jonathan Turley
Jonathan Zittrain
David Faber
David Mayer
Guido Scorza

Source: https://www.reddit.com/r/ChatGPT/comments/1h420u5/unfolding_chatgpts_mysterious_censorship_and/

Update: "David Mayer" no longer breaks ChatGPT but the other names are still problematic.

LESSWRONG
LW

LESSWRONG
LW

Nonprofit to retain control of OpenAI

Constitutional Classifiers: Defending against universal jailbreaks (Anthropic Blog)

Why does ChatGPT throw an error when outputting "David Mayer"?

Archimedes's Shortform

Archimedes

Archimedes

Trends – Artificial Intelligence

Archimedes's Shortform

Nonprofit to retain control of OpenAI

Constitutional Classifiers: Defending against universal jailbreaks (Anthropic Blog)

Why does ChatGPT throw an error when outputting "David Mayer"?

Archimedes

Nonprofit to retain control of OpenAI

Constitutional Classifiers: Defending against universal jailbreaks (Anthropic Blog)

Why does ChatGPT throw an error when outputting "David Mayer"?

Archimedes's Shortform

Archimedes

Archimedes

Trends – Artificial Intelligence

Archimedes's Shortform

Nonprofit to retain control of OpenAI

Constitutional Classifiers: Defending against universal jailbreaks (Anthropic Blog)

Why does ChatGPT throw an error when outputting "David Mayer"?

Excerpt below. Follow the link for the full post.