Lightnews — Scholar-powered news

Reposted by hakwan lau

Ann Kennedy @antihebbiann.bsky.social · 5h

Finally out: our recent work with Nick Betley is a view into how the brain reshapes its behavior in the face of competing survival needs- and also a potential angle on treatment targets for enduring pain.

A brief rundown...

www.nature.com/articles/s41...

A parabrachial hub for need-state control of enduring pain - Nature

Activity in a set of parabranchial neurons in the mouse brain is increased during chronic pain, predicts coping behaviour, and can be modulated by circuits activated by survival threats.

www.nature.com

7 21 55

Reposted by hakwan lau

Alex Shackman @ajshackman.bsky.social · 21h

www.sciencedirect.com/science/arti...

Sensory reformatting for a working visual memory

A core function of visual working memory (WM) is to sustain mental representations of recent visual inputs, thereby bridging moments of experience. Th…

www.sciencedirect.com

4 12

Reposted by hakwan lau

Tom Wallis @tsawallis.bsky.social · 13h

Check out this lovely new paper by the inimitable @emilya-izzeddin.bsky.social ! We examine how humans use natural image statistics to judge similarity.

Emily A-Izzeddin @emilya-izzeddin.bsky.social · 1d

The final bit of work from my PhD just got published at JOV! We looked at similarity judgements made for naturalistic image patches, and whether these are predicted by simple image statistics… (spoiler: yep!)

Link to paper: doi.org/10.1167/jov....

1/11

Low-level features predict perceived similarity for naturalistic images | JOV | ARVO Journals

doi.org

2 13

hakwan lau @hakwan.bsky.social · 9h

i dunno. *maybe*?

bsky.app/profile/hakw...

hakwan lau @hakwan.bsky.social · 14h

maybe :-)

i've always thought, in science, we make sure things are almost certainly right, & then we say maybe.

but this is too old fashioned these days. we should maybe just come up with some possibilities & promote them hard (!!)

or maybe not.

1

hakwan lau @hakwan.bsky.social · 12h

but chemistry computes. chemical synapses *compute* differently depending on their biochemical properties. for subcomputational functional stuff, you're left with the likes of 'non-representational orgasmic feelings', whatever they are

2 1

hakwan lau @hakwan.bsky.social · 14h

i wonder, if the field hasn't become so 'diverse' as of late, re: rigor vs speculations, to speak of 'realistic possibilities' like political manifestos.... what would this paper be like

but i do appreciate the very careful qualifications throughout

we can go back to boring experiments now

15/bye

1

hakwan lau @hakwan.bsky.social · 14h

i predict: some may try to shut down synapses, & find that, as *theoretically predicted*, it changes consciousness!!

sadly it may fool more ppl than u think

but anyway, shoutout to my old friend @neddo.bsky.social for waking me up from my slumber. i don't get to read C papers often these days

14/

1 1

hakwan lau @hakwan.bsky.social · 14h

maybe :-)

i've always thought, in science, we make sure things are almost certainly right, & then we say maybe.

but this is too old fashioned these days. we should maybe just come up with some possibilities & promote them hard (!!)

or maybe not.

1

hakwan lau @hakwan.bsky.social · 15h

so, in sum, it rather confirms my two suspicions, that 1) biopsychism may be an interesting possibility (so is religion, to some), but is pretty much a scientific deadend, at least for now, & 2) the field is overall going exactly where i worry it is going -

bsky.app/profile/hakw...

13/

hakwan lau @hakwan.bsky.social · Apr 30

i have uploaded this preprint a while back, but hadn't promoted it directly here. in this piece i explain why i can no longer recommend trainees to participate in my former home field.

The End of Conscioussness - osf.io/preprints/ps...

but i've learned a lot. thank you for everything.

🧠📈

OSF

osf.io

1 1

hakwan lau @hakwan.bsky.social · 15h

& just to be clear, synaptic circuits vs non-chemical signaling surely give distinct computational properties. so all these seems to be just alluding to some vague possibility that there may be computationally identical circuits realized by distinct bio mechanisms that happen to also support C

12/

1

hakwan lau @hakwan.bsky.social · 15h

or maybe anesthetized brains too? or maybe not? how do we ever tell?

so i liked it that the piece is a lot sharper than other biopsychist papers i've read

but it feels a lot less useful for generating experiments, compared to Ned's other work

& w/o experiments we all know where we're headed

11/

1 4

hakwan lau @hakwan.bsky.social · 15h

how do we then rule out that e.g. binocular rivalry isn't constant phenomenology w/ fluctuating access? the synapses are always there, maybe it is only cognitive access that fluctuates?

so blindsight maybe phenomally conscious too, just that the computational part is gone?

10/

1 2

hakwan lau @hakwan.bsky.social · 15h

specifically they are meant to *maybe* support the kind of functional but not fundamentally computational mechanisms for C.

but how exactly?? like in other attempts at arguing for biopsychism, i fail to find a plausible mechanistic explanation

but alright. problem is....

9/

1 2

hakwan lau @hakwan.bsky.social · 15h

& many of us disagree. but it's ok.

i'm more confused by the new stuff on synapses & sponges etc

so the TL:DR seems to be: we have synapses, some early animals do, & some don't. it is not *totally* clear what are the functional advantages of having synapses, therefore, *maybe* they support C?

8/

2 1

hakwan lau @hakwan.bsky.social · 15h

this part isn't clear. it feels like we're back to something like those 'orgasmic feelings' w/o representation content again. sorry for the example but just trying to follow the exact literature....

so these are supposed to functional yet not computational, alright

fitelson.org/prosem/block...

7/

fitelson.org

1 1

hakwan lau @hakwan.bsky.social · 15h

but ok maybe the last part isn't so important. maybe there are subcomputational biological mechanisms somehow, which are functionally identified. so they may even have behavioral consequences (?), but are NOT computation somehow.

but what are they exactly? how do they *work* in the brain?

6/

1 1

hakwan lau @hakwan.bsky.social · 15h

i agree. i don't think i ever gonna go down a level. it was Ned's counter argument though.

1

hakwan lau @hakwan.bsky.social · 15h

i mean, at the very bottom we have physical laws, not mechanisms relevant for bio explanations anymore. & Ned's example of gravity not being fundamentally computational is exactly there

& bio isn't just macro-physics. evo history matters, & it needs its own vocab

www.jstor.org/stable/30378...

5/

The Autonomy of Biology: The Position of Biology Among the Sciences on JSTOR

Ernst Mayr, The Autonomy of Biology: The Position of Biology Among the Sciences, The Quarterly Review of Biology, Vol. 71, No. 1 (Mar., 1996), pp. 97-106

www.jstor.org

2 4

hakwan lau @hakwan.bsky.social · 15h

but bio mechanisms are often considered fundmantally computational, espcially in evolutionary contexts.

he says ok, maybe, but we can always go down to a lower level of description where things are not fundamentally computational

i'm not sure. if we go low enough, are they still relevant?

4/

2 3

hakwan lau @hakwan.bsky.social · 16h

but Ned is, as ever, a cut above. for a start he makes clear he is NOT talking about a-functional substrates, i.e. bio stuff that can't be identified with purely functional descriptions. that would be as unscientific as denial of evolution, to my mind.

he is talking about biological mechanisms

3/

1 4

hakwan lau @hakwan.bsky.social · 16h

similar articles have generally been disappointing to me. they say: yeah maybe, just maybe, something like this bio thing or that, not sure which, maybe not totally irrelevant, *somehow*

i mean, whatever man. as i said elsewhere, it just feels like magicalist thinking

bsky.app/profile/hakw...

2/

hakwan lau @hakwan.bsky.social · Aug 17

i agree w/ ‪@sussillodavid.bsky.social‬ that this so-called biopsychism just feels like magicalist thinking. besides the claims that 'maybe' some bio sauce is needed, i just don't hear any concrete arguments as to why. sure maybe we need to take into account survival contexts to understand the ...

1 4

hakwan lau @hakwan.bsky.social · 16h

this paper takes me by surprise a bit. of coz, we all know Ned's been thinking along these lines for decades: philpapers.org/rec/BLOBVC
but is he really gonna seriously publish a new paper on this now, given all the AI hype & debates re: how unscientific some popular views on C are these days?

1/

1 6 22

Reposted by hakwan lau

Alicia Izquierdo @aliciaizquierdo.bsky.social · 1d

🎇 Excited to finally share JL Romero Sosa’s publication! Results are from single-cell imaging in different subregions of rat frontal cortex during ✨de novo learning. Spoiler: everything is not everywhere all at once www.nature.com/articles/s41...

Neural coding of choice and outcome are modulated by uncertainty in orbitofrontal but not secondary motor cortex - Nature Communications

Neural mechanisms underlying flexible learning and decision-making are not fully understood. Using single-cell calcium imaging, authors here found that neurons in orbitofrontal and secondary motor cortex exhibit complementary roles in reward learning, with neurons in the former exerting a sustained role in conditions of uncertainty.

www.nature.com

3 35 81

Reposted by hakwan lau

Dan Levenstein @dlevenstein.bsky.social · 1d

The Wu Tsai Institute at Yale is hiring another faculty member in neurocomputation. Come work with us in a growing community at the interface of neuroscience and AI!

More info below 👇

Wu Tsai Institute | Yale University @wutsaiyale.bsky.social · 6d

📣 WTI is hiring faculty positions! Are you interested in advancing our understanding of the brain + how it gives rise to cognition?

Two calls are open:

Open-rank search, Neurocomputation, deadline: 12.1.25
Senior search, Neurodevelopment, rolling review

🔗 wti.yale.edu/opportunities

#KnowTogether

1 9 18

hakwan lau @hakwan.bsky.social · 1d

arxiv.org/abs/2509.251...

"compositional ability generalizes [and]... transfers to a different target task... RL fundamentally changes the reasoning behaviors of the [LLMs]. In contrast, next-token training with the same data yields none of these findings."

🧠🤖 🧠📈

From $f(x)$ and $g(x)$ to $f(g(x))$: LLMs Learn New Skills in RL by Composing Old Ones

Does RL teach LLMs genuinely new skills, or does it merely activate existing ones? This question lies at the core of ongoing debates about the role of RL in LLM post-training. On one side, strong empi...

arxiv.org

1 16