iiinova — where your digital self keeps evolving

Reposted by Jordan Kay

I guess the standard parroter rebuttal to “How are large language models so effective (though not perfect) at performing arithmetic on arbitrary many-digit integers?” is “Secret calculator”

151136

Jordan Kay @jnatc.im · 23m

There are multiple authors on the paper, but she’s so defensive of the term as an accurate metaphor that one is forced to conclude it was her personal coinage.

010

Reposted by Jordan Kay

River_Tam @rivertam.bsky.social · 1h

BlueSky understand that balding is not a reflection of character challenge (impossible)

8493

Reposted by Jordan Kay

William B. Fuckley @opinionhaver.bsky.social · 29/09/2026

'you're cognitively offloading'. yes. correct. that is what i wish to do. i do not wish to expend cognition figuring out which pharmacist takes my insurance and is closest. i want to offload it & be told who that is. you are free to continue manually doing this task you evidently find so meaningful

1744544

Reposted by Jordan Kay

Colin @colin-fraser.net · 2h

I really wasn’t trying to start a fight with Prof. Bender but now that she has created a long thread subtweeting me and her disciples are all yelling at me, let me just say that this is the crux right here. She seems to respond as though I am describing something a priori impossible.

141709

Reposted by Jordan Kay

BeijingPalmer @beijingpalmer.bsky.social · 5h

everyone lucked the hell out on the terror attack today where the co-pilot stabbed the pilot, passengers took down the attacker *and* there seem to have been two off-duty pilots who were able to land the plane safely. Could have been one of the worst attacks for a while otherwise.

251072117

Reposted by Jordan Kay

dame @dame.is · 10h

in 5-10 years we’re all gonna look back on this current moment and realize AI caused a massive collective identity crisis that drove us all a little insane but it turned out fine and our worst fears didn’t come to fruition

101589

Reposted by Jordan Kay

Jordan Kay @jnatc.im · 29/09/2026

041

Reposted by Jordan Kay

Mikhail Gorbaechev @mikhailgorbaechev.bsky.social · 6h

Zohran is going to continue gaining popularity with Jews and that will continue marginalizing the forces of reaction among American Jewry.

1612

Reposted by Jordan Kay

eb99 @eb99.bsky.social · 6h

As I've noted before, this kind of action/handling is the difference between being divisive among your country's Jews (Zohran, whose approval with American Jews has increased recently) and being widely despised among your country's Jews (Corbyn eventually hit >85% disapproval with British Jews)

3353

Reposted by Jordan Kay

David Lebo @leboism.bsky.social · 7h

The Leo Frank stuff is especially horrifying. Bassem Youssef dipping into that was, to put it as mildly, disappointing

0231

Reposted by Jordan Kay

Alex @purplechrain.bsky.social · 17h

a few years ago, references to Simon of Trent or Leo Frank trutherism were a sign that someone was an intensely Christian and/or Neo-Nazi right-wing antisemite but now formerly-normal lefty politics influencers who stayed on Twitter casually reply to Mamdani with 15th century blood libel posts

8677141

Jordan Kay @jnatc.im · 4h

It’s not that the price of going out has changed, it’s that the price of staying in has changed (gone to zero, whereas it used to cost boredom). This is bad.

1125

Jordan Kay @jnatc.im · 4h

Not just that, but likely the most purely distilled example of the horseshoe of antisemitism in the modern age

0165

Jordan Kay @jnatc.im · 7h

Can’t really think of a time in the past ten days where it *wasn’t* playing in my head

020

Reposted by Jordan Kay

Colin @colin-fraser.net · 7h

My focus on arithmetic is basically for one reason: it’s an area that I can be completely confident that the facts are not directly in the training data. If the model correctly produces the sum of two 12 digit numbers I can be 100% confident that the model has not been trained on that sum.

6481

Reposted by Jordan Kay

Will Stancil @whstancil.bsky.social · 8h

I know people instinctively recoil at monocausal explanations and want important trends to have sufficiently weighty causes, but “We built a novel social environment that rewards anger and lurid pessimism and moved 95% of political discourse into it” is a pretty good explanation for a lot of stuff

1528931

Reposted by Jordan Kay

Elixir of Progress @elucidating.extradimensional.space · 29/09/2026

This thing where everyone is convinced that all tech workers are hyper libertarians has got to stop. It's one of the most liberal+left leaning workforce demographics in America.

41128

Reposted by Jordan Kay

William B. Fuckley @opinionhaver.bsky.social · 18h

The America.gov thing is a good general idea, you don’t have to hand it to the Trump admin at all because they’ve already fucked up the implementation, but a natural language interface to help people navigate bureaucracy is a good use case, hope he doesn’t poison the well.

81082

Reposted by Jordan Kay

Primordial Platypus @perfectclarity.bsky.social · 29/09/2026

It seems like at some point we decided "cynicism = truth" and that has enabled a lot of very miserable people to justify spreading that misery as a public service.

1298

Jordan Kay @jnatc.im · 10h

It does do advanced math.

140

Reposted by Jordan Kay

alias @antiali.as · 11h

Many people have studied the internal activations and found evidence of functional arithmetic units and rotational numerical representations that emerge in deep models during training, for example:

aclanthology.org

082

Reposted by Jordan Kay

Zoomer Antimillenarian @surcomplicated.bsky.social · 10h

Just absolute madness. AI denial is gonna drive you insane just as fast as thinking that AI is about to go full Roko's Basilisk. It's real, it's not about to take over the world and/or destroy humanity and/or become god. But it's sure gonna dramatically alter our world, mostly for the better.

4528

Jordan Kay @jnatc.im · 10h

This is all they have and will ever have now. Wishing them luck.

3595

Reposted by Jordan Kay

Rt Hon Peter Mannion MP @baldzera.bsky.social · 11h

Well, there are actually two rebuttals in the thread: Boxo McFoxo: everything LLMs do is totally expected and compatible with Stochastic Parrots and also I insist on everyone describing it using terminology I made up. Bender: impossible, the model probably secretly called a calculator.

3303

Reposted by Jordan Kay

Colin @colin-fraser.net · 29/09/2026

I guess I don't know the difference between "symbol manipulation" and "adding" here. If it has learned to do symbol manipulations which, when performed, produce correct sums, what's the difference between that and adding?

2182

Reposted by Jordan Kay

Orual @nonbinary.computer · 12h

this is hilarious when you can run an open weights model that can do this and you can verify that there is no secret calculator. how they seem to do math is actually super weird and cool!

7737

Reposted by Jordan Kay

jpicone @me.jamespicone.name · 12h

Notable at what point in the thread she stopped replying: bsky.app/profile/coli...

2513

Jordan Kay @jnatc.im · 12h

How are they not, you know. Maybe questioning a few things at this point?

3640

Reposted by Jordan Kay

Andy Masley @andymasley.bsky.social · 22h

This is the first part of a 2 part post on why I believe that current AI models meet my criteria for really "thinking" and "understanding," and that prohibitions on using these words because they "anthropomorphize" AI seem philosophically confused to me. blog.andymasley.com/p/why-i-beli...

blog.andymasley.com

Why I believe current AI models can really think and understand - Part 1 of 2

How the human mind is different from our conception of it

712217

Jordan Kay @jnatc.im · 12h

I guess the standard parroter rebuttal to “How are large language models so effective (though not perfect) at performing arithmetic on arbitrary many-digit integers?” is “Secret calculator”

151136

Reposted by Jordan Kay

Colin @colin-fraser.net · 29/09/2026

I'm curious how to understand within the framework of stochastic parrots the fact that LLMs can produce simple arithmetic facts (like "what is a + b?" for a and b 20 digits, let's say) with very high accuracy, when neither a nor b are in the training data. Do you have thoughts on this?

1352

Reposted by Jordan Kay

Nied [5'7"] @nied.bsky.social · 30/09/2026

Oh I understand exactly why they're outraged.

316524

Reposted by Jordan Kay

PMC Vanguardist @tylerboschert.bsky.social · 30/09/2026

Let's also not pretend that this person won't scream absolute bloody fucking murder as soon as a robot is able to wash the dishes.

612611

Reposted by Jordan Kay

First Wordle Problems @fwordleproblems.bsky.social · 29/09/2026

What the hell are they even talking about? That doesn’t describe ANYTHING happening in the real world! That damned hippo account is SO irritating.

1102

Reposted by Jordan Kay

William B. Fuckley @opinionhaver.bsky.social · 29/09/2026

Glonzo is real, and strong, and your friend. Masterful gambit sir.

521322

Reposted by Jordan Kay

Mayor Zohran Kwame Mamdani @mayor.nyc.gov · 29/09/2026

The Talmud teaches that anyone who sustains one soul, has sustained an entire world. For generations, Jewish New Yorkers have sustained our city. Today, we are launching the first-ever municipal Strategy to Combat Antisemitism.

293142612801

Reposted by Jordan Kay

Mayor Zohran Kwame Mamdani @mayor.nyc.gov · 29/09/2026

Moed Tov! Суккот муборак! Today I joined Rabbi Shlomo Nisanov in Kew Gardens to distribute free food to New Yorkers in honor of Sukkot. Thank you to Rabbi Nisanov for the reminder that community and service deepen our most joyful holidays.

514282661

Reposted by Jordan Kay

Hunter Dunn @hunterisntdunn.bsky.social · 29/09/2026

This is antisemitism. This isn’t leftism. This isn’t anti-Zionism. This isn’t progressive activism. This is just blatant antisemitism.

532264290

Reposted by Jordan Kay

Elizabeth Picciuto @epicciuto.bsky.social · 29/09/2026

I don’t say this often, but some people need to look at Bluesky more.

An anonymized tweet that says: When's the last time you saw someone on the left say, "I'm pissed at Democrats, so l'm voting Republican or staying home to teach my party a lesson

7648

Reposted by Jordan Kay

Thorne 🌸 @ens0.me · 29/09/2026

why do I even try?

someone being really grumpy a chatbot can't find her kids' shoes

202185

Jordan Kay @jnatc.im · 29/09/2026

Really weird thing to wonder imo

031

Reposted by Jordan Kay

Alt Bureau of Labor Statistics @domesticshoggoth.bsky.social · 28/09/2026

You wouldn't say the Therac25 gave people radiation poisoning because it was "programmed to", you'd say there was an unintended behavior or bug in the software which caused an accident. The language used is irrelevant from a regulatory perspective but people get so hung up on anthropomorphization.

1224

Reposted by Jordan Kay

Alt Bureau of Labor Statistics @domesticshoggoth.bsky.social · 28/09/2026

it's just weird that people are acting like computer programs, in general, have always acted exactly the way we expect them to

3253

Reposted by Jordan Kay

ellie lockhart (she/they) @eleanor.lockhart.contact · 29/09/2026

this is my absolute favorite rhetorical move that I want Democrats to pivot to, sincerely: make it clear that we don't support abolish the police/prisons by specifically advocating for the arrest of fascists

37817

Jordan Kay @jnatc.im · 29/09/2026

Frontier LLMs are almost always truthful and accurate.

010

Reposted by Jordan Kay

ellie lockhart (she/they) @eleanor.lockhart.contact · 29/09/2026

stop socially rewarding the performance of mental health distress and nihilism for clout. stop supporting creatives (including quirky shitposters) who send the message that the correct response to social challenges is nihilism, despair, or hatred. stop encouraging adults to be overgrown children.

05810