iiinova — where your digital self keeps evolving

Jake Spinella @jakespinella.bsky.social · 22/09/2026

I think this is wrong—instead of sending their proof to the annals of mathematics, they should send it to me, so I can send it to the annals of mathematics

000

Jake Spinella @jakespinella.bsky.social · 22/09/2026

One part of your brain is for hating and breathing, the other for loving and poetry. Pick your side

Jake Spinella @jakespinella.bsky.social · 22/09/2026

Further evidence that only philosophers should be allowed to write press releases (which would have the salutary effect of eliminating press releases)

141

Jake Spinella @jakespinella.bsky.social · 22/09/2026

As an aside: the criterion of organ individuation the press release uses is quite silly: many cells in many of your organs have developmentally distinct origins, but that doesn’t entail that, e.g., your kidney is multiple organs squished together.

Jake Spinella @jakespinella.bsky.social · 22/09/2026

The last bit being racism optimization? If so: JEV just outputs a probability; there’s no way to change outputs as model weights are frozen, so you get a number extremely fast and cheap at the cost of baking in maximum amounts of bias (effectively whatever bias is encoded in the data itself.)

Jake Spinella @jakespinella.bsky.social · 22/09/2026

The people have been clamoring for cheaper and faster racist outputs with even less context!

030

Jake Spinella @jakespinella.bsky.social · 22/09/2026

Not quite, JEV is zero-shot not logistic regression. It takes a distance measure in semantic label space which encodes bias at pretraining rather than at analysis. Consequently it’s harder to fix biased outputs in JEV because of this as there are no model weights to change. So they optimized racism.

241

Jake Spinella @jakespinella.bsky.social · 21/09/2026

What if the real AI alignment was the friends we made along the way

Jake Spinella @jakespinella.bsky.social · 21/09/2026

Instead of turning us all into paper clips, matriarchal AI will turn us into Nora Ephron screenplays

Jake Spinella @jakespinella.bsky.social · 21/09/2026

The entire AI safety industry neglected to consider that every learner just needs a strong maternal trainer to supervise it

Jake Spinella @jakespinella.bsky.social · 20/09/2026

Surprised and insulted that ‘Italian Intelligence’ was not one of the poll choices

Jake Spinella @jakespinella.bsky.social · 20/09/2026

Meanwhile on X, the everything app

Jake Spinella @jakespinella.bsky.social · 20/09/2026

Pronouncing ‘Claude’ like ‘Sade’ as a form of resistance

250

Jake Spinella @jakespinella.bsky.social · 20/09/2026

PSYCHOLOGIST: we think that this explanation is conceptually flawed, so we ran 400 undergrads through a two year study PSYCHOLOGIST: turns out, it’s conceptually flawed

001

Jake Spinella @jakespinella.bsky.social · 20/09/2026

The classic stuff: Hume, Hempel, and Goodman. So in one sense I can understand asking the question, since their way of thinking about the topics is so alien to the way scientists and their students typically think about what they’re doing that it’s hard to understand their motivation.

300

Jake Spinella @jakespinella.bsky.social · 20/09/2026

Neuroscientists continuing their time-honored tradition of self-owning in the presence of philosophers

052

Jake Spinella @jakespinella.bsky.social · 19/09/2026

This week a student in my philosophy of science course asked me why we’re studying induction and confirmation in a course about science. The best part? He’s a neuroscience major!

2100

Jake Spinella @jakespinella.bsky.social · 19/09/2026

Asking Claude to find De Zerbi’s email and to tell him how Tottenham can score more goals.

000

Jake Spinella @jakespinella.bsky.social · 18/09/2026

Abstract for a fun little study I've been running in my spare time. It compares LLM reasoning across tasks that hold content fixed but vary format (iconic vs. discursive), and asks what that shows, if anything, about the format of LLM representations. Full preprint coming soon!

050

Reposted by Jake Spinella

Kieran Healy @kjhealy.co · 18/09/2026

This is the far more likely pathway for how AI will end us all, btw. Not via some superintelligent swarm of paperclip maximizers, but by a cascading sequence of dumb fuckups driven by human beings who have arranged things to be fast-moving and catastrophically fragile while also being really lazy.

17391150

Jake Spinella @jakespinella.bsky.social · 18/09/2026

But might this prove something stronger (or something that seems stronger to me, at least), too? Namely, that OOE objections incur a substantial burden of proof, insofar as they need to make contested assumptions about how to interpret controversial cases to even get off the ground?

Jake Spinella @jakespinella.bsky.social · 18/09/2026

Cool paper! This is something I've thought a bit about in the case of dispositionalism re: belief. A question after skimming: you don't think that OOE objections are *bad*, just the weaker claim that they need to show that the relevant explanations don't tolerate mirroring. That seems right. 1/2

Jake Spinella @jakespinella.bsky.social · 18/09/2026

There are dozens of us not falling for the bait for those with eyes to see

000

Jake Spinella @jakespinella.bsky.social · 17/09/2026

They Live but donning the glasses allows you to see the true logical form of natural language sentences

000

Jake Spinella @jakespinella.bsky.social · 16/09/2026

They also missed a huge opportunity to title this show “The Real Stanford Prison Experiment”

Jake Spinella @jakespinella.bsky.social · 16/09/2026

EXCLUSIVE: Netflix announces new unscripted series “Arbeit Macht Frei”

Jake Spinella @jakespinella.bsky.social · 12/09/2026

I also never get a good answer as to why either 1. these aren’t just different sorts of beliefs rather than nonbeliefs or 2. They aren’t just beliefs that are in some sense inferentially isolated from one another.

130

Jake Spinella @jakespinella.bsky.social · 12/09/2026

I dislike this move too. Just because these states aren’t correlated or consistent with, say, personality traits or other beliefs doesn’t entail that they’re not beliefs— unless one assumes that belief has pretty constraining internal logical structure, which I highly doubt.

Jake Spinella @jakespinella.bsky.social · 11/09/2026

Someone just walked into this coffee shop with a “New York or Nowhere” tote bag. We are in Chicago Illinois

081

Jake Spinella @jakespinella.bsky.social · 09/09/2026

I don’t think just the universe is telling us that!

Jake Spinella @jakespinella.bsky.social · 09/09/2026

I vibe-coded everything so I’ve been using the chat portion of claude code, typically not the CLI but yeah, it’s been cool to actually come close to executing a study I’ve been thinking about for a while with only the help of one idiot junior analyst