Michelle Greene

YESSSSSSSSSSSS

Or even sheets. JFC.

A major AI training data set contains millions of examples of personal data Personally identifiable information has been found in DataComp CommonPool, one of the largest open-source data sets used to train image generation models.

NEW FROM ME: new research has found millions of ex's of personal info, including credit cards, passports, résumés, birth certificates etc in 1 of the largest web-scraped datasets used to train image generation AI models.

It's a major privacy violation.

www.technologyreview.com/2025/07/18/1...

👀

But for how many people? I bet I still win per capita.

🚨 We’re hiring! The Computational Cognitive Neuroscience Lab at Virginia Tech is looking for a postdoc to join our team studying the neural + computational mechanisms of structure learning and flexible cognition: ccnvt.github.io#positions

Wat?

“Our chatbot went full Hitler a few days ago so what better time to launch its weird new girlfriend experience functionality? Who wouldn’t want to tell all their most shameful and degrading sex fantasies to a hitlerbot?”

Paper with Bruce Hansen.
🔗 doi.org/10.1167/jov....
🧪 All code/data: osf.io/wcthf/
Thanks to all who helped make this possible 🙏
(7/7) #neuroskyence #VisionScience #PsychSciSky

Across behavior and brain, scene affordances matter most.
They shape similarity judgments, cause more confusion during categorization, and reduce the clarity of neural representations.
🧠 Our visual system is tuned not just to what a scene looks like, but what you can do there.
(6/7)

(Experiment 3: EEG decoding)
We recorded EEG and decoded scene categories over time. The images were identical, but the distractors differed.
❗️But affordance-similar distractors led to weaker category signals after 150ms
→ Even your neural code gets fuzzier when affordances overlap.
(5/7)

(Experiment 2: Categorization)
People categorized scenes amid distractors matched on affordances or surfaces.
🎯 More false alarms when distractors shared affordances.
→ Functional similarity interfered more than surface-based similarity. (4/7)

Left panel shows triplet similarity structure. Right panel is result: affordance outliers were more frequently chosen (at the expense of surfaces).

Experiment 1 used @martinhebart.bsky.social's odd-one-out triplet task, but with a twist: each pair within a triplet was selected to be an outlier in one of three feature spaces:
✅ Affordances
✅ Surfaces
✅ Materials
People picked the affordance outlier as being the most different most often. (3/7)

We know that scene perception is fast. And my previous work has argued for affordances playing a role. But they are correlated with other features, like texture and objects.
We tested whether affordances dominate perception even when other features are equally informative.
(2/7)

🚨 New paper in Journal of Vision!
We show that scene affordances—what you can do in a space—shape how we perceive and categorize scenes. This shapes your similarity preferences, predicts your categorization false alarms, and even alters neural representations. 🧵👇
🔗 doi.org/10.1167/jov....
(1/7)

You’re kind but I learned from @timnitgebru.bsky.social , @emilymbender.bsky.social @abeba.bsky.social and others!

Doc Maretens, fishnets, and a rainbow skirt

Heading to the NYC Dyke March! Who else is going?

I'm pretty sure no police face recognition system can handle this

May all who have despotic leaders be safe (Rabbi Marisa James quote)

Just when I thought I was all out of fucks! Thanks, @benwolfevision.bsky.social!

I'm originally from Kent, Ohio.
Nothing good will come from pitting the National Guard against civilians.
#4DeadinOhio

Wessel Cognitive Neurology Lab | The University of Iowa Thanks for visiting the Cognitive Neurology Lab at the University of Iowa. The lab is directed by Dr. Jan R. Wessel. We are affiliated with the Department of Psychological and Brain Sciences, the Depa...

I will be looking to hire 2-3 post-docs over the course of the next few months to work on questions related to cognitive control in humans, using EEG, TMS, DBS, sEEG, fMRI, or related methods.

If you know anybody, please tell them to email me.

Formal ad to follow. Lab website: wessellab.org

MASS MOBILIZATION FROM LA TO THE BAY: ICE OUT! MONDAY 6/9 @ 6PM 24th & MISSION BART PLAZA, SF

TONIGHT:
#ICEOUT Protest in San Francisco

We recommend viewing the online version because the multimedia features render best. I have the PDF to send if you email me.

Feedback for: Scientific Computing for Cognitive Neuroscience Thanks for taking a look at our book! Please provide any feedback you might have as we revise this living document. All questions are optional.

If you have feedback, please let us know here. We're excited to revise and expand in the coming years! docs.google.com/forms/d/e/1F... 3/3

My students wrote this book with editorial input from me. Check it out here:
mrgreene09.github.io/Scientific_C... 2/3

I'm excited to announce that my lab's open textbook on Scientific Computing for Cognitive Neuroscience (v1.0) has just gone live! Our goal is to help mend the gap between the computational skills needed by cognitive neuroscience, and typical curricula that don't yet include it. 1/3

🚨PSA for anyone using GPT in human subjects research🚨

A judge has ruled that OpenAI must retain ChatGPT and API data indefinitely, with few exceptions.

This might have implications for your IRB compliance and the security of your participants' data. 1/

Glad you're on the other side of it!