No

Very annoyed to learn I was just granted a patent for the dumbest project I ever worked on

We can make them do it if they're connected to the rest of a brain!

But yeah, the chips have something like 200k neurons, each of which is far more complex than the units in an artificial neural network.

Yeah, Bubbies brand is a pretty good version of this

The less cool thing is how many people choose to present themselves as jerks.

The cool thing about discussions on the internet is you can always take a step back and reflect on how you’re coming across. Without the pressure of needing an immediate response, you can always make a decision about how you want to present yourself.

Memory as the Brain's Slime Trail Your brain is a scratchpad for your future selves

Full article on this: cognitivewonderland.substack.com/p/memory-as-...

Which raises a weird question: maybe internal memory is stigmergy too. Past cognitive processes leave traces in the brain. Current processes pick them up. When small kids are around, that internal "medium", working memory, becomes less reliable because of constant interruptions

We use stigmergy too! The form on the kitchen table so you remember to fill it out. The half-written draft you can actually edit. Packing kids' lunches by just looking at the counter to see what step you're on. The environment is doing cognitive work.

Single agents use it too! Slime molds leave a slime trail so they don't re-explore ground they've already covered. The world becomes their memory. A thermostat is doing a version of this too: reading the room it's currently changing.

Stigmergy is indirect coordination through marks left on a medium. The action leaves a trace, and the trace triggers the next action. No central planner. No direct communication. Just agents responding to what's in front of them.

Memory as the Brain's Slime Trail Your brain is a scratchpad for your future selves

My new favorite concept: stigmergy!
It's how ants build complex nests without any central plan. One ant drops a dirt pellet, and that pellet becomes a signal for other ants to drop pellets there too. The environment itself becomes the blueprint cognitivewonderland.substack.com/p/memory-as-...

One of my kids will pretty much only eat pickles

On the one hand, he's having trouble gaining weight and it's a constant concern for us and his pediatrician

On the other hand, the kid has great taste

“Living well is the best revenge,” I repeat to myself in the mirror, psyching myself up for yet another idyllic day fuelled by the thought of my eudaimonia crushing my foes.

It sure is a lucky coincidence that I ended up with not just the best kid, but the two best kids in the world

Sucks for the rest of you though, ngl

My longer write-up about this: cognitivewonderland.substack.com/p/science-co...

Overhyped science actively undermines wonder.
"Brain cells play Doom!" gets a quick "whoa" and then you move on. It doesn't give you a new way of seeing the world. That's what good science communication can do, and it's why the hype machine is worth pushing back against.

But I think debunkers do something beyond correcting the record. They're part of an accountability ecosystem. When scientists and journalists know their claims will be scrutinized, they have reason to be more careful. Accountability is a counterweight to those incentives pushing toward exaggeration.

Debunking exaggerations or outright misconceptions works. Stating the misconception, explaining what's wrong, giving the correct picture is effective across a wide range of topics.

There are incentive structures that lead to this hyping. Scientists exaggerate in abstracts and press releases to attract more attention for funding. Journalists need to churn things out quick and get clicks.

"Human brain cells on a chip learned to play Doom in a week"

Recently I got mad about a science news story that was not only clickbait, but the article itself was fluff and left readers less informed than if they didn't read it.

It made me question what the point of science news is anyways, if it's treated as cheap entertainment with no factual basis.

As a neuroscientist, I recommend disregarding any advice you read that begins with “As a neuroscientist”

‘Digital sphinx’ raises questions about connectome models The sphinx, with a worm’s brain and a fly’s body, illustrates the potential pitfalls of using deep-learning techniques to model biological processes.

But using these techniques to model biological processes also has pitfalls, as illustrated by the model Brunton and her team created.

Campus, who was not involved in the work.

To circumvent these shortcomings, researchers are starting to use deep reinforcement learning to relate the connectomes to behavior.

The problem with connectome models is that they do not capture the biophysical properties of neurons or the pools of neurotransmitters that modulate neural communication, and they exist without a body, says Srinivas Turaga, group leader at the Howard Hughes Medical Institute’s Janelia Research

It enables them to ask, “If I just created a randomly connected connectome, could it also do the same behaviors?” he says.

Brunton’s work points to an important control for other researchers looking to combine deep learning and connectomics to simulate fly behavior, says Benjamin Cowley, assistant professor at Cold Spring Harbor Laboratory, who was not involved in the preprint.

Working with her team, Brunton strung a biophysical model of a Drosophila body to a simulation of the Caenorhabditis elegans connectome and trained that “digital sphinx,” a term coined in the preprint, to walk using deep reinforcement learning—all with a “brain” that wasn’t a fly brain at all.

“We need to be really careful in interpreting this kind of work,” says Bing Wen Brunton, professor of biology at the University of Washington, who posted the new preprint on bioRxiv last week in response to Eon Systems’ announcement.

It has little to do with the architecture of the network used.