Joe Bak-Coleman

It’s amazing seeing @holdenthorp.bsky.social say the obvious part clearly. www.science.org/doi/10.1126/...

science.org

Tips for journalists covering "AI in space": -Ask about heat dissipation *at scale* -Ask "Doesn't the sun come down here too?" -Ask "y tho?" www.nytimes.com/2026/09/24/t...

nytimes.com

Oh there’s like a pretty big body of literature that gets shunted to the banner of “critical Metascience” that questions all of the replication crisis narratives they undergird things like LLM reproducibility or predicting replicability. Stuff like this: arxiv.org/abs/2604.26268

The Difference Between "Replicable" and "Not replicable" is not Itself Scientifically Replicable arxiv.org

Some background here. He was a philosopher in the NYT joebakcoleman.com/blog/2026-09...

Anatomy of an AI Interest Story · Joe Bak-Coleman joebakcoleman.com

For example this report "Beyond P(doom)", which throws 1970s statistical thinking at Yud's existential risk and summarizes it neatly. You wind up with this doozy of a recommendation; but note that it is all couched in LessWrong thinking. cset.georgetown.edu/publication/...

Beyond P(doom) for AI Risk: Quantifying Uncertainty Without Probability | Center for Security and Emerging Technology Georgetown AI cset.georgetown.edu

The counter-perspective would be one that focuses on a world in which AI *isn't* Boom or Doom; Here's a report I worked on that pushed back on boom and doom perspectives. Check out the executive summary. hdr.undp.org/content/huma...

Human Development Report 2025 | Human Development Reports hdr.undp.org

Lots of people have been around longer than these companies but they also work with them. Chalmers for example has ties to Anthropic and Eleos AI. eleosai.org/post/taking-...

eleosai.org

Lots of people have been around longer than these companies but they also work with them. Chalmers for example has ties to Anthropic and Eleos AI. eleosai.org/post/taking-...

New report: Taking AI Welfare Seriously eleosai.org

I don’t get the “I’m so flooded with peer reviews I simply must use AI” take. Just uhh turn them down? Click decline.

I got a *remarkably quick* rejection email from a tenure track job that I applied to. Rejections are part of the process! However, if you’re going to screen candidates with ai, please don’t ask for more than a cv and let us know! It’s a lot of time learning about your dept, writing and tailoring.

I’ll write a longer blog post about it soon, but I’ll believe OpenAI agents have formed a hacking collective and it’s not just a PR stunt when they ransomware a Swiss bank, not another AI outfit about to be acquired by a hardware providider; when Interpol not MIRI investigates.

Had my first email arising from what seems to be an LLM-guided spite-check of my published code. Anyone else experienced this yet? I'm pulling the repo and verifying the possible error, will happily correct if need be but this seems like a great way to DDoS scientists. #metascience #AcademicSky

One of the more common responses I got yesterday was the argument from philosophy of mind that roughly asserts simpler organisms than ourselves must also have minds (lest we be human chauvinists); perhaps even groups of organisms therefore LLMs must have minds too. It's a rhetorical switcheroo tho.

It's wild to me that journalists at storied institutions can't smell an IPO pump and instead wind up sanity-laundering the weird AI Death cult.

Setting aside the AI debate on hand, it’s wild to assert that reading @emilymbender.bsky.social and @edzitron.com could produce a tragedy right up there next to all wars, pandemics, massacres, etc.. that happened in the last 25 years.

Now that we're all learning about who the effective altruists are through the AI Doom/Sex Cult discourse; can we have a discussion about the intense amount of money they're pushing into the hands of anyone using AI in metascience?

The serendipity of @faineg.bsky.social introducing @jamellebouie.net to the rationalist cult running ai policy via the Aella alluvial diagram made Bluesky feel like good twitter.

If you really wanna watch AI bullshit you, get diagnosed with an exceedingly rare disorder for which the literature is a cobweb filled closet with a couple of expired cans of beans, a hammer, and (for some reason), three new iPhones.