✨ Meet our speakers! ✨ Among our speakers this year at #SNS2025 we have Simone Ebert (@simoneebert.bsky.social) & Jan Lause (@janlause.bsky.social), from Hertie AI Institute, University of Tübingen Read the abstract here 💬 👇 meg.medizin.uni-tuebingen.de/sns_2025/abs... #neuroskyscience

Die Zeit lässt ja aktuell die Frage diskutieren, ob links Schuld ist an AfD und der aktuellen CDU/CSU. Statt Selbstverantwortung der Wähler*innen oder gar (mediale) Selbstkritik der vermeintlich erfolgreichen Erinnerungskultur gehts also um Schuldzuweisung. Man kennts ja nicht anders. 1/

It's still possible to register - join us in tuebingen for this very exciting lineup! #neuroskyence #compneurosky

🚨Come work with us in charming Tübingen and become a member of our growing Freigeist group! We offer a 2-year postdoc position on the communicative plasticity in wild and captive bonobos, fully funded by @volkswagenstiftung.bsky.social. Applications open until March 15th. bit.ly/4gVa5Je

I just defended my PhD on single-cell normalization and visualization methods in neuroscience! 🧑‍🎓 🖥️ 🧬 🧠 Thanks a lot to @hippopedoid.bsky.social, @philipp.hertie.ai and everyone in the lab for supporting me ❤️ Now looking forward to my postdoc on what Amacrine cells are computing in the retina 👀

Great Friday read about cell type evolution in the brain! 🧠 🧬 🖥️ #neuroskyence

Hello, Bluesky! ☀️ We fund frontier research in Europe—bold ideas, unexpected discoveries and science that shapes the future. So it’s only fitting we’ve landed here. Sorry for being late. Follow us for updates on ERC funding, research policy, and our grantees' discoveries.

Can LLMs replace human teaching assistants in grading homework assignments? This semester, we ran a study in our Bioinformatics course with 100+ enrolled master's students, whose assignments were graded by 6 LLMs. 1/

“efforts to demystify AI may inadvertently reduce its appeal, indicating that maintaining an aura of magic around AI could be beneficial for adoption.”

Hi, Vox reporter here: If you've had a study section, NIH travel, or other NIH-related meeting cancelled (or not), raise your hand! Reply or DM me, and repost for visibility if you're able. #academicsky @drugmonkey.bsky.social

Worth also pointing out that there are many "tests so easy no AI system can pass them". Moravec's paradox remains. E.g., arxiv.org/abs/2404.12390

Revealed: France, Greece & others lobbied to weaken European AI Act Governments secured loopholes in new law, enabling AI-powered surveillance in border zones and public spaces for 'national security', internal docs obtained by Investigate Europe reveal: www.investigate-europe.eu/posts/france...

Biggest German press agency @dpa.com has joined the sky.

Zuckerberg macht ernst: auf Instagram werden Posts zu den Hashtags #savedemocracy #fascist #project2025 #antifascist #bidenharris nicht mehr angezeigt, da sie „sensible Inhalte“ darstellen würde. Wer Faschisten unterstützt, für den sind das halt „sensible Inhalte“.

Hello, world! I’m moving here from the X sphere, hoping to find a nicer environment 🙂

Institute of Ophthalmic Research in Tübingen is happy to welcome @katrinfranke.bsky.social ! She joined us in January 2025 and wil establish a new "Neuronal Computation in Vision" Group. Congratulations !! @unituebingen.bsky.social @tueneurocampus.bsky.social 👉 further information bit.ly/4fZR1c3

This is a hilarious way to say “we haven’t figured out how to make AI do anything useful”

Super excited to be making this a permanent thing. It has been an exciting ride so far!!! Thanks so much to @plantevolution.bsky.social for sharing it with me for the last year and great news that you are staying with @elife.bsky.social, going forward.

The DFG, the major German science funding agency, just left Twitter and is now on Bluesky! 🦋 #eXit Hello 👋 @dfgpublic.bsky.social

Meta is deleting links to Pixelfed, a decentralized, open source Instagram competitor that has had skyrocketing signups in recent days: www.404media.co/meta-is-bloc...

What if… spontaneous neural activity 🧠 reflects the baseline rumblings of a brainwide dynamical system initialized for learning? We find that the rumblings have macroscopic properties like those emerging from linear symmetric, critical systems 🧵 #neuroscience #neuroAI www.biorxiv.org/content/10.1...

Check our latest in which we leverage shape metrics to compare neural geometry across regions, sessions or subjects and how their differences predict behavior. w/ Nejatbakhsh, Duong, @sarah-harvey.bsky.social, Brincat, @siegellab.bsky.social, @earlkmiller.bsky.social & @itsneuronal.bsky.social

A list of the many German universities that are leaving X/twitter today!

Curious how place cells adapt to uncertainty in reward location? We found that predictable changes boost “reward place cells” & warp hippocampal maps. Check out our new preprint: with Feng Xuan, Jack Mellor, Peter Dayan and Dan Dombeck: www.biorxiv.org/content/10.1...

More confusion about AGI: OpenAI now officially defines it as ‘whenever we can make loads of $$ from it’! Quote: “"AGI" will have been achieved [..] when one [..] AI [..] generates [>]$100 billion in profits” So..Altmans claim translates to “Believe us, we will make profit, invest now please”?!

Who is up for a short thread on the intriguing parallels between OpenAI and Enron, as told by one of Enron's investment bankers -- that is to say, me? Here goes...

I’m just remembering when companies sued kids for millions of dollars for sharing a few songs online. Now big business is saying they can’t stay in business without stealing everyone’s copyright.

Apple now scans iPhone photos for visual landmarks by default, trying to make your photos searchable. Even though it seems to be done in a privacy-preserving way (great article in QT 👇)— wild that this is opt-out, and not opt-in! (If you want to disable: Apps>Photos>Enhanced Visual Search) #MLSky

Here's a question I have NEVER managed to get a straight answer to: if I paste an internal company strategy document into ChatGPT or similar to ask for a summary, what are the chances that a competitor in six months time can ask the model about my strategy and get it to leak internal secrets?

Some of my thoughts on OpenAI's o3 and the ARC-AGI benchmark aiguide.substack.com/p/did-openai...

When i learned latex, inline equations were denoted by dollar signs, e.g. $ ... $. But these days everyone is using \( ... \). What's up with that?

\( … \) looks like the little dude is cheering for your silly math letters and what could be better than that? ☺️

@garymarcus.bsky.social on why OpenAIs is too optimistic about o3 Biggest issue imo: They did not use a consumer o3, but trained an o3 specifically on the puzzle type they were testing! Normal users could not do that with their tasks, so this is far from real-life performance. 🧪 🤖📈 #MLSky #scicomm

Regardless of what you think of Gary Marcus overall AI stance, these kind of level-headed analyses are very important. o3 is probably quite impressive, but these openAI announcements are marketing for VCs and have nothing to do with science. This will slow down progress rather than helping it.

openAIs latest o3 model might cost up to $5000 per query—and still only solves <90% of certain logic puzzles (examples attached). I mean — amazing that this works at all, but I would happily do it for $100/puzzle ;-) more interesting details in @natolambert.bsky.socials post ⬇️ 🧪 🤖📈 #MLSky

Thank you so much!! You can find the starter pack here: go.bsky.app/JKv3BWU

Can we build neural networks whose structure and computational abilities match a real brain? We are not quite there yet, but recent work by @lappalainenjk.bsky.social et al. shows a strategy for getting closer to this goal. Read more on our blog: www.machinelearningforscience.de/en/improving...

"The Illusion-Illusion: Vision Language Models See Illusions Where There are None" osf.io/preprints/ps... (this is a more systematic examination of the thing I was looking at a few days ago)

Looks like we finally made it to Bluesky! @nature.com 🥳

1/4) I was thinking about this paper this morning, and there's one thing that I'm stuck on here... Would be curious to hear your thoughts @mameister4.bsky.social. Isn't the "outer brain", as you put it, producing very high-throughput sensorimotor behaviours constantly?

Paper claims human cognition operates very slow at around 10bit/s - not without making fun of Elon’s plans to directly talk to AI with Neuralink brain electrodes: “Instead of the bundle of Neuralink electrodes, Musk could just use a telephone [to be symbiotic with AI]“ 🤖 🧠 🧪

IT'S OFFICIAL! From today, @plos.org and its journals are no longer posting on Twitter/X. If you want to carry on hearing about great science in our journal, follow @plosbiology.org on Bluesky here: bsky.app/profile/plos...

Make sure all your early vision friends are on this starter pack!

PSA for authors posting to bioRxiv, who don't want to miss if someone comments on their preprint or writes a review: Click "follow" on your papers! Otherwise, you won't be notified. ✉️🚨

New preprint! A wonderful collaboration of our @FellerMarla lab with Matthew Po, and @shekharlab. Here, we dive into the impact of spontaneous activity on the transcriptome of retinal ganglion cells (RGCs), the sole output neurons of the retina. www.biorxiv.org/content/10.1...

Bsky is being flooded with German speaking AI powered bots. A thread on why this is important, ultimately making all social media is unusable. #deutschland #trolls #AI #election 1/ What happened? All these responses to a @zeit.de post have something in common (spoiler: they are not real):

What would you think might a scared fruit fly hide from curious scientists in its legs? A) Some tasty fruit? 🍇 B) Its soul?? 👻 C) An additional HEART??? 💓 This is so cool!