LightNews
Thomas Dietterich
@tdietterich.bsky.social
7.5K followers 490 following 900 posts
Safe and robust AI/ML, computational sustainability. Former President AAAI and IMLS. Distinguished Professor Emeritus, Oregon State University. https://web.engr.oregonstate.edu/~tgd/
web.engr.oregonstate.edu
Posts Media Videos Starter Packs
Thomas Dietterich @tdietterich.bsky.social · 15h
Politics in Britain is increasingly just a derivative of US politics. How embarrassing!
Thomas Dietterich @tdietterich.bsky.social · 1d
I keep hoping for an “authenticity backlash”. It can’t come too soon
hedgielib.bsky.social
Abigail Goben @hedgielib.bsky.social · 1d
This sounds awful

www.theguardian.com/lifeandstyle...
‘I realised I’d been ChatGPT-ed into bed’: how ‘Chatfishing’ made finding love on dating apps even weirder
Where once people were duped by soft-focus photos and borrowed chat-up lines, now they have to watch out for computer-generated charm. But it’s one thing to use a witty phrase – another thing entirely...
www.theguardian.com
1 7
Thomas Dietterich @tdietterich.bsky.social · 2d
"Don't trust -- verify!"
Thomas Dietterich @tdietterich.bsky.social · 4d
Nice work! Thank you
1 1
Reposted by Thomas Dietterich
thedailyshow.com
The Daily Show @thedailyshow.com · 4d
The following is REAL footage from Portland, 2025. Viewer discretion is advised.
1.9K 23K 56K
Thomas Dietterich @tdietterich.bsky.social · 4d
Windows 11 is also the most inefficient Windows OS I've used. I was forced by my IT folks to put it on a Lenovo T480S, and it became unusable. Thinking fond thoughts about linux. What distro are people using these days?
4 6
Reposted by Thomas Dietterich
histoftech.bsky.social
NightMar on Hicks Street, Part 4: The Rabbit’s Revenge @histoftech.bsky.social · 4d
tbh I think Microsoft is going to blink right before October 14, and if they don’t… well, that’s a very, very bad decision on their part.
16 42 130
Thomas Dietterich @tdietterich.bsky.social · 5d
Interesting story!
marketplace.org
Marketplace @marketplace.org · 5d
As GLP-1 medications become more common, restaurants and brands are trying to adapt to the public’s changing diet preferences. That means more protein and smaller portions, but some of those preferences aren’t necessarily good for business.
As Ozempic changes how Americans eat, the food industry is trying to adapt
Restaurants and snack companies are moving towards high-protein foods and smaller portions.
mktplc.org
4
Thomas Dietterich @tdietterich.bsky.social · 6d
If all of our cameras digitally signed their images and our web browsers verified those signatures and prominently displayed the verification (or lack thereof), perhaps that would go part way to addressing the decay of authenticity?
2
Thomas Dietterich @tdietterich.bsky.social · 7d
And digital signatures?
1 9
Thomas Dietterich @tdietterich.bsky.social · 12d
Highly misleading headline @washingtonpost.com
1
Thomas Dietterich @tdietterich.bsky.social · 16d
IIRC Thomas Griffiths and his students have made some successful predictions of this kind.
1
Thomas Dietterich @tdietterich.bsky.social · 16d
It doesn’t have to match in every way, but a useful simulation can predict how human cognition will behave under a wide range of circumstances.
1 1
Thomas Dietterich @tdietterich.bsky.social · 16d
As an AI researcher, I study how to build useful systems. But I would never claim that they simulate human cognition. The systems mimic some aspects of human intelligent behavior, nothing more.
1
Thomas Dietterich @tdietterich.bsky.social · 16d
But you used the word “simulate”, and I want to take that seriously. Statistical ML does not match human generalization in my (extensive) experience.
1 1
Thomas Dietterich @tdietterich.bsky.social · 16d
Do you know of work comparing GPT memory behavior to human memory behavior? I have not been tracking this literature
1
Thomas Dietterich @tdietterich.bsky.social · 16d
A simulation only matches the simulated system in some ways and not in others. What aspects of cognition do you claim are matched by GPTs?
2 2
Thomas Dietterich @tdietterich.bsky.social · 16d
It would be great to invent some more precise terms for all of the various technologies that are referred to as AI. Ideas?
2 2
Thomas Dietterich @tdietterich.bsky.social · 16d
These AI are being used for much more than chatbots. They are enabling better retrieval and summarization of the scientific literature. They are improving language translation. And they certainly help less experienced programmers
1 3
Thomas Dietterich @tdietterich.bsky.social · 19d
Thanks! Always eager to learn more.
1
Thomas Dietterich @tdietterich.bsky.social · 19d
This is not true if you are crossing languages. LLMs have unquestionably improved the technical writing of Chinese authors
Thomas Dietterich @tdietterich.bsky.social · 19d
Based on the photo, I thought this was from @theonion.com
Thomas Dietterich @tdietterich.bsky.social · 19d
Can you recommend an alternative that also seeks to be objective?
1
Thomas Dietterich @tdietterich.bsky.social · 19d
Copyright law doesn't protect what people seem to want it to protect. This talk by Pamela Samuelson @pamelasamuelson.bsky.social was very enlightening.
icml.cc/virtual/2025...
ICML Invited Talk Generative AI's Collision with Copyright LawICML 2025
icml.cc
1 3 4
Thomas Dietterich @tdietterich.bsky.social · 19d
Great article on autism.
www.nytimes.com/2025/09/24/o...
Opinion | Autism Has Always Existed. We Haven’t Always Called It Autism.
www.nytimes.com
15 66