Mauve 👁💜 @mauve@mastodon.mauve.moe

Ok so nvidia fell for the old ways but you gotta wonder what state actors are in anthropic and which country is the best at polyamory tradecraft

Was re-reading Melanie Mitchel's 2019 book, "Artificial Intelligence: A Guide for Thinking Humans", https://melaniemitchell.me/aibook/ and was relieved to (re)read how she refers to "units" instead of "neurons" in artificial neural networks. Even adds a note about doing so to avoid the dismay of neuroscientists. Indeed.

Got reminded again today of how powerful biological neurons are by Mickey London's lab latest preprint:

"What can a neuron compute", Aizenbud et al. 2026 (Segev, London)
https://www.biorxiv.org/content/10.64898/2026.06.08.730984v1.full

The secret sauce of biological neurons? Mostly dendritic morphology:

"Bounds on the computational complexity of neurons due to dendritic morphology"
Anamika Agrawal, Michael A. Buice 2025
https://www.biorxiv.org/content/10.1101/2025.07.11.664448v1

A point spectacularly well made by Jones and Kording a while back:

"Can single neurons solve MNIST? the computational power of biological dendritic trees", Jones and Kording 2020
https://arxiv.org/abs/2009.01269

... in line with plenty of prior work, even by the same lab, like:

"On Functions Computed on Trees", Farhoodi et al. 2019
https://direct.mit.edu/neco/article/31/11/2075/95669/On-Functions-Computed-on-Trees

The key ingredient? That a presynaptic neuron delivers multiple inputs across the postsynaptic neuron's dendritic tree. That enables a lot of the above's computational properties. To reproduce that, a whole multi-layer ANN is needed to match the power of a single cortical pyramidal neuron:

"Single cortical neurons as deep artificial neural networks", Beniaguev, Segev, London 2021.
https://www.cell.com/neuron/fulltext/S0896-6273(21)00501-8

What's more, adding biological constraints to ANNs improve the latter:

"Do Biological Constraints Impair Dendritic Computation?", Jones and Kording 2021
https://www.sciencedirect.com/science/article/pii/S0306452221003900

#neuroscience #dendrites #ANN #AI

@i0null

If you’re confused about what’s happening with Nvidia right now…

Basically Jensen Huang, the guy who runs Nvidia, which is the company that makes the chips basically every AI company on earth needs, just sent an email to his entire company admitting he cheated on his wife, and that came right after the New York Times reported that US intelligence believes a woman who got close to him was secretly passing information to the Chinese government, and to understand how we got here you have to go back to June 2024, when this Chinese businesswoman named Alice Huang, who also goes by Huang Le and isn't related to Jensen at all, they just have the same last name, gets photographed at a party in Taipei sitting right next to him, and then about a year later her company, a small Singapore company called Megaspeed that almost nobody had ever heard of, reportedly sets up a deal to buy around $2 billion worth of Nvidia's chips, which is a crazy amount of money for a company that small, and the Times even pointed out it was an unusual amount of business for a firm like that, and people already had questions about where all those chips were really going, but Nvidia looked into it and said everything checked out, so for a while that's kind of where things sat, until October 29, when the Times reported that US intelligence had concluded Alice was passing information to Beijing, and that it wasn't some new discovery either, because Trump had reportedly been briefed on it about a year earlier, and the officials involved apparently believed Jensen didn't know what she was up to, but here's the strange part, nobody ever told him, because there's this thing called a defensive briefing, where counterintelligence people basically pull someone aside and say hey, a foreign government might be targeting you, be careful, and Jensen never got one, even though the whole time the government was still watching her, so you have the head of the most important AI chip company in the world apparently being targeted, and the people who knew about it just let it keep going, and then everyone starts denying it, Nvidia says the Megaspeed allegations were already investigated and debunked last year, China's foreign ministry says the whole report is baseless, and Jensen says the business with Megaspeed was totally normal, just a customer buying chips like any other customer, but Congress doesn't let it go, and three House Republicans who run national security committees ask Jensen to come in and explain himself in a closed-door briefing, which means a private meeting away from the public and the cameras, and then on a Sunday, out of nowhere, Jensen sends that email to every single employee saying he had been unfaithful to his wife, that it hurt her and their children, and that nothing excused what he did, and he doesn't say who it was with, but you can imagine where everyone's mind went, and right after that, Nvidia's board, which is the group of people who are supposed to oversee the CEO, announces it's setting up an independent committee with outside lawyers to investigate, but at the same time says it has no reason to believe any national security secrets or company secrets were compromised and that it still has confidence in him, and then a company spokesperson comes out and says none of this is going to slow Nvidia down and Jensen is going to keep running the company without any interruption, so basically you have a suspected Chinese intelligence asset who got close to the CEO, a $2 billion chip deal, a president who reportedly knew about, a CEO who was never warned, a confession about an affair, an investigation, and a company insisting everything is fine, so that's pretty much where Nvidia is tonight.

One thing that still shapes how I think of the world is this introductory linguistics class where the instructor have explained the difference between prescriptivist and descriptivist linguists.

Prescriptivists think they know language better, or at least think that one can know “the language” “better.” For them, language is governed by rules and patterns, and everyday language is merely a manifestation of these structures.

Descriptivists take a totally different stance. They believe that everyday language is “the language” and that language is continuously reshaped by its users in unpredictable ways. Which means: there is no structure, and no rules set in stone. Only the language itself, an entity in constant flux.

Being a prescriptivist is comforting while quite involved in the justifying change part. Being a descriptivist is FOMO, but freeing in that you know there’s no universal unattainable truth.

This one class made me a descriptivist. Not only in linguistics, but in many other areas and systems. Before I rush to change something, I think: is this situation a consequence of how people interact, systemically? Is there benefit to it? Should I understand it better before trying to impose my rules on it?

I should.