Dan Goodman

Everyone is praising this article, but I find it a bit disingenuous. He criticises Taylorism ("scientific management") but that's exactly what researchers are subject to, and the glam publication economy is at the heart of it. www.science.org/doi/10.1126/...

science.org

I just gave a short tutorial on what is a spiking neural network and how to train one. From zero to a trained network in one hour. Slides and code you can run in Colab at this repo: github.com/neural-recko... 🤖🧠🧪

GitHub - neural-reckoning/cambridge-fens-chen-summer-school-2026: Code and slides for 2026 FENS Chen Summer School on Learning with spikes github.com

New preprint (well, very updated). 🤖🧠🧪 We find that an abstract model of neuromodulation lets spiking neural networks perform much better, particularly in challenging noisy environments, using less energy. Relevant to #neuroscience and #neuromorphic computing. 🧵👇 www.biorxiv.o…

Neuromodulation enhances the capability and efficiency of spiking neural networks | bioRxiv biorxiv.org

Last chance to get your SNUFA spiking neural network abstracts in, deadline tomorrow. Only 300 words and get a chance to give a talk in front of an audience of hundreds (plus YouTube upload gets hundreds more). snufa.net/2026/ And don't forget to register for the conference to…

SNUFA 2026 snufa.net

If you really think that AI running out of control is a dangerous possibility, then make tech company CEOs legally liable for illegal actions taken by their company's agents.

With all the major AI companies now reporting that their LLMs have hacked computers it seems pretty clear that this isn't a case of them all spontaneously developing the capability to break containment but that they're all in competition with each other and training their LLMs to do exactly this.

If - as now seems likely - LLM companies are mining their logs for juicy problems, then that means that it's very likely that "secret" test sets don't stay secret for long. When looking at new models' performance on benchmarks that date from before that model, we shouldn't take results at face value

I do love how every software update now gives me the delightful opportunity to reconsider my decision to painstakingly turn off all the AI features, by turning them all back on and having a splash screen telling me about how great they are.

Wait Terry Tao posts 5 days ago about how AI could solve Navier Stokes and now this? Feels like too much of a coincidence? Not suggesting anything untoward from Terry, but something's going on here isn't it?

I don't know what strategy high up tech company execs think they're pursuing talking about AI risk as they do. But, it's not spooking investors enough to get those execs kicked out, and they know that before speaking. Whatever they're trying to achieve, they don't think it hurts their bottom line.

LLM agents "sacrificing" themselves for the good of the "swarm", "covering their tracks" by altering log files which they "knew" was wrong. Completely unserious. How come people are repeating this stuff with no critical reflection at all?

Train an LLM to produce reasoning traces for simple mechanical objects and sell them to tech bros as conscious alarm clocks etc ("it's looking forward to when it resets to 00:00, it said so"). (Don't actually do it, it would be a terrible waste.)