🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

#NeuroAI meeting: Blake Richards makes great point that we cannot treat math, statistics, and comp sci as optional in neuroscience training. For Neuro, all are just as essential as biology (Though in my experience biologists are better with stats than comp sci graduates are)

The submission deadline for NICE has been extended until December 1 easychair.org/cfp/NICE-2025 We hope to see all of you in Heidelberg next year!!!

Clear numbers to show that LLMs are overkill by orders of magnitude, and as universal function approximators, they really aren't that interesting. At this point it is just Big AI wanting to make $$$$ and block competetion and the rest of the community playing into their hands.

Ironic since the biggest technical problem with AI is also extrapolation.

To go full snark: the brain evolved energy efficiency through resource constraints and survival of the fittest. So my proposal is to give every AI researcher a single GPU and come back in 10 years and see who survives. Or we can just subsidize Nvidia nature.com/articles/d4158…

Seems like something that may cause the pitchforks to come out for 21st century Clippy.

AI industry question: Did Google making its commitment to specialized hardware (TPUs) slow their transition / adoption of Transformers? Whereas companies relying on more generic hardware (GPUs) could adopt more quickly?

We can't force the brain to fit our existing conceptual frameworks. We need something new. But Bergson here points out how difficult this is.

Today is the day!!!!

An accompanying News & Views by Felix Wang and Brad Aimone is also available! nature.com/articles/s4358… 🔓rdcu.be/d4bjR

New preprint! (w/ Brad Aimone) We present a spiking neuromorphic algorithm for solving large, sparse linear systems (Ax=b) like those arising from finite element methods for solving partial differential equations. We also demonstrate the method on Intel's Loihi 2 neuromorphic

Best US cities for random restaurants while traveling: LA, Chicago, Houston Those cities are spread out, but literally anywhere you eat will be great. Most disappointing: DC, Atlanta There is good food, but random places are very hit or miss Most surprising: Salt Lake City

This is a great post. And for that matter, a great illustration about how physics is the most boring of all of the sciences

So much wrong out there today...

A nice paper illustrating how far away modern AI and neuromorphic computing is from the type of learning actually seen in the brain