iiinova — where your digital self keeps evolving

Mike Caulfield @mikecaulfield.bsky.social · 14h

Kind of amazing General Conference on Weights and Measures adopted the "leap hour" unit just today and we've already spent it

120

Reposted by Mike Caulfield

Matthew Salganik @msalganik.bsky.social · 14/10/2026

Happy to share a new working paper: "Machine Learning Models Trained on a Longitudinal Survey Dataset Poorly Predict Hundreds of Life Outcomes" arxiv.org/abs/2610.17115 You can hear Emily Cantrell present it Thursday Oct 15 at Oxford: metrics-and-models.github.io (online participation possible)

arxiv.org

Machine Learning Models Trained on a Longitudinal Survey Dataset Poorly Predict Hundreds of Life Outcomes

Understanding the predictability of life outcomes is important for public policy and science, but existing studies focus on a limited number of outcomes and therefore struggle to describe the broad pa...

06829

Reposted by Mike Caulfield

Mike Caulfield @mikecaulfield.bsky.social · 14/10/2026

Face/Off is a film where two men swap identities followed by both actors doing a perfect imitation of Nicolas Cage

1345

Mike Caulfield @mikecaulfield.bsky.social · 14/10/2026

Face/Off is a film where two men swap identities followed by both actors doing a perfect imitation of Nicolas Cage

1345

Mike Caulfield @mikecaulfield.bsky.social · 14/10/2026

Little did they know it was just another red earring

1201

Mike Caulfield @mikecaulfield.bsky.social · 11/10/2026

Please just stop, AI Overview, you're out of your depth here

1120

Mike Caulfield @mikecaulfield.bsky.social · 11/10/2026

But it all goes through a lot of safety testing anyway, I'm just confused by this I guess

120

Mike Caulfield @mikecaulfield.bsky.social · 11/10/2026

Maybe the core for him is "a bunch of chemicals no one has seen before" I mean if it looked like something from another planet, sure I'd pass.

120

Mike Caulfield @mikecaulfield.bsky.social · 11/10/2026

Also what would it mean for "one person that understands the mechanism" to exist? Outside math the problem of science is many people purportedly "know" things but those things differ and conflict and science is a process of dealing with that.

120

Mike Caulfield @mikecaulfield.bsky.social · 11/10/2026

I think it seems to miss in a really incredible way that there are stakes on both sides of this which is what makes policy hard. If risk could be construed as something only novel things produce or as something never about lost opportunity policy would be quite simple.

130

Mike Caulfield @mikecaulfield.bsky.social · 11/10/2026

I don't know about cancer cures but for half the psych meds people take for off label uses no one really knows the mechanism for certain. It's not ideal, but if someone is suffering from addiction and Ozempic works to stop it that seems good enough for personal choice and not enough for broad rec.

180

Mike Caulfield @mikecaulfield.bsky.social · 10/10/2026

Operating LLMs as connection machines is driven ultimately by descriptions, which now create the sort of hyperlink-like networks we used to only dream of. This in turn, if we learn to use it, can broaden our minds instead of narrowing them. These are good things and we should learn to do them.

3456

Mike Caulfield @mikecaulfield.bsky.social · 07/10/2026

Also if this is true for a lot of these (probably all?) we weren't that far away from avg mathematicians doing this at home. 3 hours! We will see how it shakes out but if true that changes things

000

Mike Caulfield @mikecaulfield.bsky.social · 07/10/2026

Runaway is so bad and so unwatched that if you go on niche Prime channel Howdy and select Runaway it will say "customers also watched The Postman" and I'm pretty sure that's because watched The Postman and Runaway last week

040

Mike Caulfield @mikecaulfield.bsky.social · 07/10/2026

That is not most of the critics' assumption! But to truly understand what's going on one does need to comprehend that assumption is *real* for the people advancing it.

130

Mike Caulfield @mikecaulfield.bsky.social · 07/10/2026

People are also missing that if you assume (like the labs) that the advancement of AI is going to produce significant GDP and quality of life gains greater than anything since the industrial revolution the idea that you wouldn't advance your engine using these problems seems morally wrong too.

120

Mike Caulfield @mikecaulfield.bsky.social · 07/10/2026

I love how escape from the planet of the apes starts off like a Gilligan's island episode then ends up a gritty noir with a twilight zone ending

0101

Mike Caulfield @mikecaulfield.bsky.social · 06/10/2026

But yeah, "pass laws to make mail in voting difficult and point to the decreased numbers as evidence of a coming Democratic rigged election" is one reason why this is not a good field for your mental health.

171

Mike Caulfield @mikecaulfield.bsky.social · 06/10/2026

North Carolina is also down and they added a requirement that you photocopy your id or get exception form and a flurry of reqs as the ballots went out with the inside envelopes etc. It's not even trust, it's just gotten much more complex.

150

Mike Caulfield @mikecaulfield.bsky.social · 06/10/2026

You turn people who were waiting for their ballot and didn't receive one because of changes to the new law into a story about how there's really a conspiracy to cover up bad Democratic numbers. It's shameless and shameful but what's new.

130

Mike Caulfield @mikecaulfield.bsky.social · 06/10/2026

There's been a ton of laws that have made all this harder, and if we think about the vicious cycle this is just another way you delegitimize elections to get rules and you use the results of those rules to further delegitimize

130

Mike Caulfield @mikecaulfield.bsky.social · 06/10/2026

I think the easier explanation is just structural - Orange County is down, but Florida also changed renewal rules so that people that got one for the general didn't auto renew for the midterms which is HUGE.

140

Mike Caulfield @mikecaulfield.bsky.social · 05/10/2026

The AI got it right actually. Wikipedia is wrong, as is imdb. In some research I found that this error can be found as far back as a 1993 issue of TV Guide.

010

Mike Caulfield @mikecaulfield.bsky.social · 04/10/2026

Futurism is cope as a service (CAAS) and it's best to find another link.

0120

Mike Caulfield @mikecaulfield.bsky.social · 04/10/2026

It helps that it was the name of that book and I am not kidding

020

Mike Caulfield @mikecaulfield.bsky.social · 04/10/2026

Paramount put out Reds and they also put out Top Gun. It's hard enough to figure out what film is going to get your money back without throwing political selection into the mix. Unless the tax code makes you rich enough not to care.

0142

Mike Caulfield @mikecaulfield.bsky.social · 04/10/2026

The thing is this is one way capitalism works pretty well when capitalists have to worry about losing money. It really does. But our tax policy has screwed that up.

1213

Mike Caulfield @mikecaulfield.bsky.social · 03/10/2026

I think they think they are making it fancy but it just looks like keebler fudge stripes

010

Mike Caulfield @mikecaulfield.bsky.social · 03/10/2026

All these cookies had issues my guess is zero of those problems were underbaking

240

Mike Caulfield @mikecaulfield.bsky.social · 03/10/2026

Also every time the judges said oh this is underbaked I did not in fact believe them

150

Mike Caulfield @mikecaulfield.bsky.social · 03/10/2026

Can't help wondering if this episode is why Pru left, I don't think she could have made it through

020

Mike Caulfield @mikecaulfield.bsky.social · 03/10/2026

These people do not know what a cookie is

3130

Reposted by Mike Caulfield

Jo @whatsjo.bsky.social · 03/10/2026

me: Hey, I can’t sleep. Do you think it’s because of stress? hand puppet: *rolls over on my pillow to look at me* you have had a lot on your plate recently.

439987

Mike Caulfield @mikecaulfield.bsky.social · 02/10/2026

Me, after eating all the doughnuts I brought to work for my colleagues "man, those top pot donuts sure go fast huh"

1170

Mike Caulfield @mikecaulfield.bsky.social · 01/10/2026

Well that one is going to be next year at the latest with the off the shelf and is certainly functional in the lab by now

010

Mike Caulfield @mikecaulfield.bsky.social · 30/09/2026

Kind of wild, the agentic film fact-checker I built can now use computer vision to check if a film summary accurately orders a sequence of events. Wasn’t there a prediction by someone that LLMs wouldn’t be able to watch a film and adequately explain it by 2030? We are not there yet but headed there.

19 contact sheets streamed from the public-domain Internet Archive copy. Each tile is stamped with its time in the film (h:mm:ss); the sheet names are the checker's own labels. Click a sheet for full size.

The claim notes cite times in the film rather than sheet names; the 12 source-audit claims that cite a time are listed at the bottom.

2100

Mike Caulfield @mikecaulfield.bsky.social · 29/09/2026

Honestly replace two of the three debates with quizzes

47615

Mike Caulfield @mikecaulfield.bsky.social · 29/09/2026

Most of the increased accuracy on niche questions comes through search, not corpora, and it is good to remember that. mikecaulfield.substack.com/p/yes-when-l...

mikecaulfield.substack.com

Yes, when LLMs don’t use tools they still suck at niche questions

And that’s important for nonobvious reasons

0114

Mike Caulfield @mikecaulfield.bsky.social · 28/09/2026

New Couch to 4k episode on how bad the Neflix thriller The Ice Road is www.youtube.com/watch?v=Z1oP...

youtube.com

The Ice Road (2022): "A bagged salad of tropes"

YouTube video by Couch to 4K

140

Mike Caulfield @mikecaulfield.bsky.social · 26/09/2026

Honestly I've been burned by so many eBay discs I use amazon because if vendors are under their return policy I can just drop a bad disc off at whole foods. I also order direct from kino lorber etc, but for standard discs amazon

100

Mike Caulfield @mikecaulfield.bsky.social · 25/09/2026

This isn't to let other models off the hook, but I'd do think the release of GPT-4o was a uniquely bad decision, and we will be sorting through its impacts for a while.

1162