computingnature, to random

Check out our lab’s recent work on visual learning and representations at the workshops:

computingnature, to Neuroscience

"A combinatorial code of action and context for motor memory." Jae-Hyun Kim, Kayvon Daie, Nuo Li. Really nice talk, context-specific codes in continual learning, and no representational drift observed in ALM https://www.youtube.com/live/Kq5f36OwFKU?feature=shared

tdverstynen, to random
@tdverstynen@neuromatch.social avatar

The more that I think about it, the more I feel that Blaise Aguera y Arcas's presence at was both toxic and counter productive to the mission and goals of the meeting and community.

Get out your sewing kits, a vent thread is inbound!

1/4

tdverstynen,
@tdverstynen@neuromatch.social avatar

First, let's summarize Blaise's talk itself. I posted my opinions on it yesterday during the talk.

https://neuromatch.social/@tdverstynen/112020275221817874

Here's a quick summary of his main points:

  • AGI is exemplified by the abilities of LLMs like ChatGPT, therefore AGI is here and LLMs are an imperfect example of it. People who cannot accept this fact are ignorant of the issues.

  • Prediction is the core microfunction that makes intelligence work.

  • Life is intelligence.

  • Genetic algorithms work. (No seriously, he presented a genetic algorithm model as if it was the first time anyone had looked at the emergence of intelligent structure, with absolutely no acknowledgement of the prior work that goes back almost a half century now)

The talk was full of very bold, but unsupported, statements and a complete lack of acknowledgement of prior work (except for the occasional obligatory references to Schrodinger, Van Neumann, and Turing). Despite a bold talk title "What is intelligence?" we didn't really learn the answer to that question (unless you accept his premise that LLMs are AGI and thus intelligence)

2/4

tdverstynen,
@tdverstynen@neuromatch.social avatar

Now let's move on to his behavior during the panel discussion later in the evening.

Here's a sampling of his positions.

  • Neuroscientists who do not accept modern AI in its current SOTA form (really only the models being pushed by Silicon Valley) are clearly ignorant about the issue of intelligence.

  • Academic research cannot afford the $100+ million to train the next gen large models (so it just shouldn't try to work in this space).

  • AI doesn't really need neuroscience, but neuroscientists need AI (a position also supported by a few other panelists).

  • Aging is something to be cured (an extremely ableist position on aging)

  • When asked what we will be debating in 20 years, his response was AI personhood and rights. This was said in a room filled with women and multiple trans/non-binary scientists from America whose rights are currently being taken away with no clear resolution in sight.

3/4

tdverstynen,
@tdverstynen@neuromatch.social avatar

Blaise's presence stood in stark contrast to the other speakers and panelists at , who have so far given beautiful and thoughtful talks, considered questions deeply, and acknowledged the limitations of their work or the field at large.

Blaise's presence amplified all those toxic traits that are stereotypical of Slicon Valley and anti-thetical to the mission of societies like COSYNE.

  • Uncritical acceptance of SOTA technology without any consideration of the limitations or criticisms that have been raised.

  • A lack of acknowledgement of almost any prior work, particularly that done by anyone who isn't a white male.

  • A lack of consideration of relevant ethical issues.

  • Advocacy of positions that ignore or dismiss the humanity of others (e.g., the ableist stance on aging, putting considerations of AI "rights" over current civil rights issues affecting humans).

Now, I love COSYNE and its community dearly. It's a conference I always look forward to attending when I can and much of the work that I cite and study comes from the COSYNE community. I think having a Silicon Valley techbro come and lecture us about the glory of modern AI, without any clear awareness of the context of issues, was a notable stain on this otherwise wonderful conference. I hope we can avoid such unnecessary and toxic behavior in the future.

4/4

tdverstynen, to random
@tdverstynen@neuromatch.social avatar

Watching the panel on Neuro-AI and amused at the lengths that biologists and computer scientists will go to avoid using the term “cognitive science” or even acknowledge its existence.

tdverstynen, to ai
@tdverstynen@neuromatch.social avatar

Apparently, according to Blaise Aguera y Arcas at , AGI is defined as the abilities that LLMs do, thus LLMs have AGI and it has arrived.

Maybe the folks working on should study circular inference a little bit more.

aheadofthenerve, to random
@aheadofthenerve@neuromatch.social avatar

can someone at @CosyneMeeting tell me what "Mind Matching (Neuromatch)" is gonna be so I know whether to rush over? TY!

cian, to random
@cian@mstdn.science avatar

Synapses fluctuate in size over days-weeks (example data in pic below). Why?

We used math and computational modelling to ask if it could be explained by stochastic gene expression.

Turns out yes! And interestingly it induces

  1. strong temporal correlations in protein fluctuations (days timescale)
  2. Strong spatial correlations across synapses, due to dendritic mRNA count fluctuations.

If you're at #cosyne2024 check out postdoc Oleg Senkevich's poster 2-151 on this on Friday.

Picture of Oleg Senkevich
Poster abstract
image/png

tdverstynen, to random
@tdverstynen@neuromatch.social avatar

Folks at interested in basal ganglia pathways, come check out our poster during Session 1 tomorrow.

#1-067 GPe arkypallidal neurons can mediate inhibitory control by disrupting competition in the striatum

ekmiller, to random

Going to ? Stop by these fine posters for cool results.

1-081 Estimating flexible across-area communication with neurally-constrained RNN

1-111
Subspace transformations underlie decision-driven working memory prioritization

3-034
Anesthesia fragments cortical activity within a hemisphere, but synchronizes it across hemispheres

dlevenstein, to Neuroscience

Presenting this at for what should be the last time as a manuscript in prep 📝

Sequential predictive learning accounts for hippocampal representation and replay. With @alxecome, Roy Eyono, @adrien , and @tyrell_turing

Come say hi at the Friday poster session!

image/jpeg
image/jpeg

aheadofthenerve, to random
@aheadofthenerve@neuromatch.social avatar

Any other wet lab folks packing for
@CosyneMeeting ?

aheadofthenerve, to random
@aheadofthenerve@neuromatch.social avatar

so the workshop hotel for @CosyneMeeting is full-- where is everyone else staying? helppp

aheadofthenerve, to random
@aheadofthenerve@neuromatch.social avatar

Just registered for and , in Portugal and Italy respectively! Who else will be there?

CosyneMeeting, to random

workshops have been announced!!🤓🧠🤩

Check them out 👇👇👇

https://www.cosyne.org/workshops-program-2024

CosyneMeeting, to random

Vibing at our all hands✌️🧠

CosyneMeeting, to random

Check out our stellar Organizing & Executive Committee members working hard to bring you the best 20th anniversary Cosyne 👇👇👇
https://www.cosyne.org/cosyne-committees

Over the next few weeks we'll be highlighting them by learning what they are most looking forward to at 🧠🤓🇵🇹

CosyneMeeting,

Aside from seeing his friends & colleagues and taking stock of the field's progress at Co-GC Blake Richards @tyrell_turing is looking forward to:

1⃣ research into the 🧠​'s predictive learning (which he feels is on an upswing)

2⃣ the latest cortical comp. DNN models

CosyneMeeting,

At Co-PC Bing Brunton is👀 for systems neuroethology, especially using biomechanics to model the 🧠& behavior!

Fun fact: Cosyne was Bing's first academic conference presentation as a grad student 🤓 making us ☘️ to have her expert guidance for the 20th anniv!

CosyneMeeting,

Meet our wonderful DEIA Co-Chair @lukesjulson who is eyeing:

1⃣meeting students & postdocs
2⃣discussing science
3⃣eating fresh-baked pão com chouriço after a late night of dancing at 💃🕺
4⃣work bridging the gap between spiking data & dynamical systems on 👀

CosyneMeeting,

The superb @memming is 👀 his "absolute favorite conference" 🤩 Every year there are new ideas, experiments, tools, theories, and friends😊

He first presented as a graduate student in 2007 & has had 31 abstracts so far!!

P.S. he's pumped for Metamersion @neuro_cf 😉

CosyneMeeting,

At the wonderful Marcelo Mattar is most looking forward to friends, beach weather & amazing science 😎

Namely:
1⃣RNNs to model episodic memory & decision making
2⃣Clever experiments designed to validate those models
3⃣Connections between transformers & the 🧠

  • All
  • Subscribed
  • Moderated
  • Favorites
  • JUstTest
  • mdbf
  • ngwrru68w68
  • modclub
  • magazineikmin
  • thenastyranch
  • rosin
  • khanakhh
  • InstantRegret
  • Youngstown
  • slotface
  • Durango
  • kavyap
  • DreamBathrooms
  • megavids
  • GTA5RPClips
  • tacticalgear
  • normalnudes
  • tester
  • osvaldo12
  • everett
  • cubers
  • ethstaker
  • anitta
  • provamag3
  • Leos
  • cisconetworking
  • lostlight
  • All magazines