Brad Wyble
@bwyble.bsky.social
6.2K followers 4K following 1.5K posts
Academic, cognitive & vision scientist, computational modeller, cofounder @neuromatch Academy, He/His. This is a personal account.
Posts Media Videos Starter Packs
Pinned
Explaining to my grandkids that back in my day a picture like this required you to actually get 700 beds and put them on the beach instead of asking AI to draw it.

[Left picture is from Pink Floyd's 1987 Momentary Lapse of Reason album]
Image of Pink Floyd's momentary lapse of reason album cover photo showing hundreds of beds on a beach.  Picture of three women arranging the bedding on said beds, presumably before the shot was taken.
Reposted by Brad Wyble
martinwiener.bsky.social
POSTDOC Opening: I'm hiring a postdoc to work with me, @ayeletlandau.bsky.social, and Yuval Benjamini on a 4-year NSF funded project to understand timing and memorability in the visual system. fMRI, EEG, eye-tracking all included.

If interested, please DM or email me for more information!
martinwiener.bsky.social
Thanks to NSF and BSF, we've received a CRCNS grant!! 🎉

I'll be working with the amazing @ayeletlandau.bsky.social and Yuval Benjamini to explore and understand how our sense of time and image memorability are linked. ⌛🧠

We have 2(!) post-doc opportunities available - details coming soon!
Reposted by Brad Wyble
luckytran.com
You’ll now be able to get an updated coronavirus shot in every state without a prescription, according to the American Pharmacists Association.

CVS and Walgreens have stated you won’t need to attest to any underlying conditions and the questions will be removed from their online booking system.
Reposted by Brad Wyble
jacquelyngill.bsky.social
The new Wiley journal review system is an absolute nightmare. Editorial assistants have been replaced by a clunky, AI-enabled website that makes it harder to find and invite reviewers. Associate editor workloads have increased. I'm not convinced the invitations to reviewers are even going through.
Also, other for-profit publishers are not much better.
Yes but it's more efficient.
Reposted by Brad Wyble
iwillnotbesilenced.bsky.social
Liberal South Carolina
Judge Diane Goodstein's home is burning to the ground after an explosion. SLED is investigation as arson and refer attack as Judge Goodstein had been receiving multiple threats.
Thanks I hadn't heard of it. I didn't pick up on the implication of it being Dutch only.
yea, that was a damn shame to see. I imagine they pay for APIs, which means premiums for overleaf go into that.
Can you tell me what Surf refers to here? I'm curious about alternatives.
I think he was commenting there on your point about removing Word. I think would be a non starter at my university since we rely on Office for quite a lot of tasks and there isn't a good replacement option.
Reposted by Brad Wyble
jeffsharlet.bsky.social
It appears that my employer, Dartmouth, one of the Trump 9, has said no to the compact. All the better given that our president is cited within it. But she’s saying no. Count the small victories when they come.
Office of the President
Dear Dartmouth community,
 
As many of you know, Dartmouth was one of nine universities asked by the White House to give feedback by Oct. 20 on a draft of its “Compact for Academic Excellence in Higher Education.” 
 
I am deeply committed to Dartmouth’s academic mission and values and will always defend our fierce independence. 
 
You have often heard me say that higher education is not perfect and that we can do better. At the same time, we will never compromise our academic freedom and our ability to govern ourselves. 
 
Best,
Sian Leah Beilock
President
very complicated statistics.
Interesting that it got the cone upside down in the original, but then got it on the second try.
Have you tried going in the reverse direction? e.g. Ask GPT 5 to describe an ice cream cone's shape using only letters?
Thank you, that's very comprehensive. I was trying to think about what connections you could make to solve these with language alone. E.g. icecream scoops are spherical which is close to circular, and o's are circular. Is that the kind of linkage you are imagining the LLM is doing?
And I should have read the thread before I posted, you generated novel tasks. Could there still be some contamination based on that style of problem? In other words, imagine you had an LLM trained on everything except those specific papers on imagery, would it still work?
Excellent work Jorge. Do you have any thoughts on whether this is from data leaking? (e.g. the model read the papers on this phenomenon which would let it know the solution to at least the umbrella example).
Reposted by Brad Wyble
jorge-morales.bsky.social
Imagine an apple 🍎. Is your mental image more like a picture or more like a thought? In a new preprint led by Morgan McCarty—our lab's wonderful RA—we develop a new approach to this old cognitive science question and find that LLMs excel at tasks thought to be solvable only via visual imagery. 🧵
Artificial Phantasia: Evidence for Propositional Reasoning-Based Mental Imagery in Large Language Models
This study offers a novel approach for benchmarking complex cognitive behavior in artificial systems. Almost universally, Large Language Models (LLMs) perform best on tasks which may be included in th...
arxiv.org
Reposted by Brad Wyble
govpritzker.illinois.gov
In any other country, if federal agents fired upon journalists when unprovoked — what would we call it?

If federal agents marched down streets demanding papers — what would we say?

Authoritarianism — let’s not pretend it’s something else when it happens in our American cities.
Reposted by Brad Wyble
theserfstv.bsky.social
One of the most haunting aspects of Squid Game was that the contestants could leave at any point if they wanted to but volunteered to be tortured to death because they were so desperate for money.

Anyways...
Reposted by Brad Wyble
jentaub.bsky.social
He. Was. Indicted. For. Bribery. In. 2024. Do. Not. Congratulate.
Single constructs, if possible (and they are not), makes framing their intellectual pursuits much easier. First, find the quintessential GI algorithm, then add a billion gpus and you've solved all of humanities problems.
I think this is emblematic of the disconnect between engineering and psychology in understanding minds. Engineers tend to collapse capacities onto single constructs. (memory, intelligence, emotions) while psychology views them as multifaceted.