Twitter can sometimes feel like a zoo, and this week, two behavior-modeling studies had our feeds wild with excitement.
Up first was a new preprint shared by Benjamin Cowley, a computational neuroscientist at Princeton University, in which he and his colleagues used a deep neural network, or a ‘deep net,’ to model the visual system of a fruit fly. “Deep nets are great at predicting visual neurons. Yet, they are unable to tell us which artificial neuron directly corresponds to a biological neuron… until now!” Cowley wrote.
His results showed “that visual projection neurons at the interface between the eye and brain form a distributed population code that collectively sculpts social behavior.”
Excited to share a new manuscript!
Deep nets are great at predicting visual neurons. Yet, they are unable to tell us which artificial neuron directly corresponds to a biological neuron… until now!https://t.co/nQJ4f7K29c
(yes, that is indeed a fictive female fly, good guess!) pic.twitter.com/PSkexLJldZ
— Benjamin Cowley (@BenjoCowley) July 21, 2022
“Here, biological and artificial knockouts predict neural activity and distributed function in complex brain areas. I am dazzled,” tweeted Cori Bargmann, a neuroscientist at Rockefeller University in New York City.
Classical genetics uses knockouts to infer functional relationships. Here, biological and artificial knockouts predict neural activity and distributed function in complex brain areas. I am dazzled. https://t.co/r1HVxSkxKK
— Cori Bargmann (@betenoire1) July 22, 2022
Dan O’Shea, a neuroscientist at Stanford University in California, was also excited about the paper, highlighting that “it underscores the tractability of the fly visual system.”
This is such an impressive experimental and computational paper, and it underscores the tractability of the fly visual system. So cool! Congratulations! https://t.co/uXt3xdV9Xd
— Dan O’Shea (@djoshea) July 22, 2022
There were many other tweets we don’t have room to share, but they had one thing in common: big excitement about this work and its implications for the future of neuroscience.
Other fly-modeling work had Twitter buzzing, too — and this one also included mice. The team in question announced the release of a new dataset “from real-world behavioral neuroscience experiments.”
The dataset “consists of mouse (9 mil frames) and fly (4 mil frames) social interactions for studying behavioral representation learning!” tweeted Jennifer Sun, a graduate student at the California Institute of Technology in Pasadena.
We are excited to release the dataset from the 2022 MABe Challenge! ????????
Our dataset consists of mouse (9 mil frames) and fly (4 mil frames) social interactions for studying behavioral representation learning!
Paper: https://t.co/QV1KynfVkR
Challenge: https://t.co/deeqxcf61L pic.twitter.com/cnq4fpcZNI— Jennifer J. Sun (@JenJSun) July 22, 2022
Ann Kennedy, assistant professor of neuroscience at Northwestern University in Chicago, Illinois, who contributed to the dataset and was featured in a Spectrum profile article this week, shared her excitement in a quote tweet.
Super excited to share the dataset + evaluators from this year’s Multi-Agent Behavior Challenge on unsupervised + self-supervised representations of behavior!
We also share the winning submissions: three teams used transformers (Perceiver, GPT, and BERT), one used Pointnet. https://t.co/5S6EpDjBAh
— Ann Kennedy (@Antihebbiann) July 22, 2022
This may sound like a far-out concept, but back-and-forths on Twitter can be peaceful, polite and productive. For proof, we offer up one such conversation sparked by Twitter user @drdebah, assistant professor of psychology at the University of New Orleans in Louisiana, after she tweeted out information about participating in a survey on cannabis use in autism.
“Marijuana may or may not be helpful to autistic young people, although autistic adults who use it frequently have positive things to say. If nothing else, the logical thing to do would be to study it more.” << YES! See flyer below to participate in #cannabis #research in #autism pic.twitter.com/JmGautT7Nj
— the tiniest doomwook (@drdebah) April 22, 2022
“Really interesting research area, but very much not a fan of the way some of the ‘demographic’ questions about autism are set up,” Twitter user @science_enby replied.
Really interesting research area, but very much not a fan of the way some of the “demographic” questions about autism are set up
— Sam ????????️???? (@science_enby) July 25, 2022
They went into greater detail in a string of tweets, bringing up such issues as the ranges available in response to a question asking when an autism diagnosis was received, tweeting, “I’d argue there’s a lot more difference between getting diagnosed at 3 vs 17 then there is between 75 and 85.”
@science_enby also pointed out issues with the responses available for the survey question asking respondents about support level, tweeting, “It’s really weird to see different autistic traits ascribed to different ‘support levels.’”
A. The issue with “functioning levels” is that they’re not static and linear, not just the words
B. Support needs are about support needs. It’s really weird to see different autistic traits ascribed to different “support levels” (cont.) pic.twitter.com/boY7nZmiA7— Sam ????????️???? (@science_enby) July 25, 2022
“I hear & receive all this feedback!” replied @drdebah, who responded with explanations but also promises to fix certain issues and clarify others in any write-ups about the study.
I hear & receive all this feedback! TY! For IRB purposes any1 under 18 had to have parents/guardian responding for them. Hence the under or over 18 division. With more adults responding we used the typical survey divisions for ages. But tots agree with your point. Will def fix.
— the tiniest doomwook (@drdebah) July 25, 2022
Agreed. Functioning tots changes with time. Since the survey is like a ecological momentary assessment, any info on the current functioning abilities was of interest. Will definitely mention this point in any write ups tho for clarity.
— the tiniest doomwook (@drdebah) July 25, 2022
And tots agree again. Was trying to meld the two fields/views/literature on functioning abilities and support levels so it could be understood by parents/guardians and autistic people. Will make sure to better clarify in the future.
— the tiniest doomwook (@drdebah) July 25, 2022
“Thanks for the responses! I definitely appreciate the difficulty of turning complex experiences into analyzable variables,” tweeted @science_enby in a final response.
Thanks for the responses! I definitely appreciate the difficulty of turning complex experiences into analyzable variables.
— Sam ????????️???? (@science_enby) July 26, 2022
Finally, the Society for Neuroscience announced the appointment of Sabine Kastner as the next editor-in-chief of the Journal of Neuroscience.
SfN is pleased to announce the appointment of Sabine Kastner as the next Editor-in-Chief of #JNeurosci, the most-cited journal in the field of #neuroscience
Read more: https://t.co/dY0DfVK0Oq
???? photo by @Princeton pic.twitter.com/BZ7SojsoZr
— Society for Neuroscience (SfN) (@SfNtweets) July 21, 2022
Princeton Neuroscience Institute called the appointment “fantastic news for Sabine and the field!” in a quote tweet.
Fantastic news for Sabine and the field! Congratulations, Sabine! https://t.co/esWK2Kfwyg
— Princeton Neuroscience (@PrincetonNeuro) July 21, 2022
In reply, Nicole C. Rust, associate professor of psychology at the University of Pennsylvania, called Kastner an “excellent choice.”
Excellent choice @SfNtweets.
— Nicole C Rust, PhD (@NicoleCRust) July 21, 2022
That’s it for this week’s Community Newsletter! If you have any suggestions for interesting social posts you saw in the autism research sphere, feel free to send an email to [email protected].
Follow us on Facebook, Twitter (@Spectrum), Instagram and LinkedIn.
Subscribe to get the best of Spectrum straight to your inbox.
Cite this article: https://doi.org/10.53053/RXHV2963
By joining the discussion, you agree to our privacy policy.