Jaan Aru
@jaanaru.bsky.social
3.4K followers 1.1K following 110 posts
Raising two kids, co-PI-lot of the Natural and Artificial Intelligence Lab & trying to do some funky research on the intersection of neuroscience, psychology and AI. Estonian nail.cs.ut.ee https://scholar.google.de/citations?user=FvFOzS8AAAAJ&hl=en&oi=a
Posts Media Videos Starter Packs
jaanaru.bsky.social
There is a lot of talk about LLMs revolutionizing education. But once you get into it, things are not that trivial.

Here, we summarize the questions that emerge when trying to implement AI Tutors nationwide.

Out now in @cp-trendscognsci.bsky.social

#edusky

authors.elsevier.com/a/1ltdo4sIRv...
Reposted by Jaan Aru
adykstra.bsky.social
Our open-access article - Testing circuit-level theories of consciousness in humans - together with Yunkai Zhu, Carolina Fernandez Pujol, @dvwz.bsky.social, @jonescompneurolab.bsky.social, @tmarvan.bsky.social, and @danclab.bsky.social, was just published @cp-trendscognsci.bsky.social.
jaanaru.bsky.social
Thank you very much for remembering and noting my work! :)
jaanaru.bsky.social
Could we understand vision as a type of problem-solving? In this new paper, we develop a computational model that iteratively refines the hypothesis about the visual input with evolutionary search.

www.biorxiv.org/content/10.1...

work led by @tarunkhajuria.bsky.social

#visionscience #neuroAI
The difficult constellation image is solved by generating candidate solutions with a GAN and refined using a genetic search conditioned on best fitting of the solution outline to the dots on the constellation image.
jaanaru.bsky.social
Thank you! this doesn’t mean that psychedelics don’t have any effects mediated by membrane receptors in L5p neurons — they probably do, and several. But regarding the reduction of excitatory currents and the emergence of late eEPSCs the intracellular receptors might be crucial
jaanaru.bsky.social
Thank you! I'm happy if our review is useful.
jaanaru.bsky.social
We systematically reviewed 23 in vitro and 26 in vivo electrophysiological studies on psychedelic compounds, with an emphasis on layer 5 pyramidal neurons.

Our results challenge the simplified view that psychedelics uniformly increase cortical excitability

www.biorxiv.org/content/10.1...
jaanaru.bsky.social
Thanks! Indeed, there are many further possibilities. Let's keep on exploring them!
jaanaru.bsky.social
🙏 I would have stopped working on this a long time ago if there weren't people like you
jaanaru.bsky.social
I'm making some further points in the chapter but probably only three of my best friends are reading this post down here in the thread (and I might lose one of them if I continue) so I just stop here.

🙏 if you got this far.

Sometimes I really hate working on consciousness
jaanaru.bsky.social
This doesn't mean that one cannot capture consciousness computationally at all but we might be really far from understanding these specific computations. The potential space of biological computations is completely unknown to us.

The stupid figure comes from www.sciencedirect.com/science/arti...
jaanaru.bsky.social
As I claim in the chapter, matter matters for the mind. I think the physical brainy stuff matters for the computations. You cannot simply abstract all these details away. The physical properties significantly constrain what types of computations can be run on them. Part of computations is physical.
jaanaru.bsky.social
Third, the most problematic assumption is that we are closing in on the computations of consciousness

The Butlin paper really broke my heart because the key assumption of this paper is that our theories are close to figuring it out. I'm not so sure. We might be really far

arxiv.org/abs/2308.08708
Consciousness in Artificial Intelligence: Insights from the Science of Consciousness
Whether current or near-term AI systems could be conscious is a topic of scientific interest and increasing public concern. This report argues for, and exemplifies, a rigorous and empirically grounded...
arxiv.org
jaanaru.bsky.social
There is an (implicit) assumption that probably consciousness corresponds to some type of pattern of spiking (the details might be different according to GNWT, RPT and other theories)

But this might be just a historical artifact not a real thing. Spikes might not be the currency of consciousness.
jaanaru.bsky.social
Second, I think our "progress" with respect to understanding the neural correlates of consciousness has been relatively pathetic. Our own theory sucks but it is clearly the best recent theory coming from neurobiology. We must do better! We can do better 💪
www.cell.com/trends/cogni...
Cellular Mechanisms of Conscious Processing
Recent breakthroughs in neurobiology indicate that the time is ripe to understand how cellular-level mechanisms are related to conscious experience. Here, we highlight the biophysical properties of py...
www.cell.com
jaanaru.bsky.social
First, I acknowledge that consciousness research has made progress but to me that progress has come with constraints: successful theories constrain further progress. What if these theories are not close to figuring out consciousness? Our field was intellectually more fluid 20 years ago. #consci
jaanaru.bsky.social
If you're interested in consciousness, please check this accepted chapter where I try to confront two major assumptions

1) the NCC is probably some type of pattern of spiking
2) we're close to figuring out the computations underlying consciousness

arxiv.org/abs/2506.21485
#consci #assc28
🧵👇
The assumptions that restrain us from understanding consciousness
The science of consciousness has been successful over the last decades. Yet, it seems that some of the key questions remain unanswered. Perhaps, as a science of consciousness, we cannot move forward u...
arxiv.org
jaanaru.bsky.social
oh no, these damn humans! 🙂In the first experiments, we explicitly mentioned not to use any external help, but in this second set, which you now played, we didn't explicitly say this, so it is indeed possible that some humans did it. From debriefing we know that many still modeled it in their mind.
jaanaru.bsky.social
Of course, OpenAI updates their models. Our o3 experiments were done in January, so that would be my first guess. We're happy to check this in a systematic fashion in 10 days.