Sign in

Thomas Dietterich

@tdietterich.bsky.social
8.3K followers 573 following 1.7K posts

ArXiv Editor in Chief and Chair of the Computer Science Section. Safe and robust AI/ML, computational sustainability. Former President AAAI and IMLS. Distinguished Professor Emeritus, Oregon State University. web.engr.oregonstate.edu/~tgd

PostsRepliesMedia
Thomas Dietterich @tdietterich.bsky.social · 01/10/2026
Corrected link: blog.arxiv.org/2026/10/01/u...
blog.arxiv.org
Fair Moderation, Equitable Access, and AI: arXiv’s Updated Rate Limit Policy
arXiv, and the scientific community at large, are facing a watershed moment. Scholarly publishing is currently changing at a rapid pace, and we are seeing a…
072
Thomas Dietterich @tdietterich.bsky.social · 01/10/2026
It seems that all of the organizations training AI models are training for broad, general intelligence. This includes knowledge about dangerous topics such as bioterrorism and cybersecurity. I don't think such models should be available to the general public.
160
Thomas Dietterich @tdietterich.bsky.social · 01/10/2026
Attention arXiv authors. Because of the exponential increase in submissions, arXiv is limiting the number of papers that a person can submit each month to two. blog.arxiv.org/2025/10/31/a...
blog.arxiv.org
Attention Authors: Updated Practice for Review Articles and Position Papers in arXiv CS Category
arXiv’s computer science (CS) category has updated its moderation practice with respect to review (or survey) articles and position papers. Before being…
511131
Reposted by Thomas Dietterich
steinly0.bsky.social @steinly0.bsky.social · 01/10/2026
"If this goes on..." arxiv.org/stats/monthl...
arxiv.org
Monthly Submissions
021
Reposted by Thomas Dietterich
Jane Rosenzweig @janerosenzweig.bsky.social · 29/09/2026
Something that has troubled me over and over in these AI years is how we seem to need to redraw every line to accommodate AI. It is possible for writing to have value for thinking and for oral exams and discussions to have value. We don't need to remake the world to eliminate writing/5
38015
Reposted by Thomas Dietterich
Clément Canonne @ccanonne.github.io · 29/09/2026
The report from the AI + TCS Working Group held at the @simonsinstitute.bsky.social is out, including "twelve concrete near-term recommended actions." Currently reading it: while I have some reservations, but there is some good in it! Notably items 7, 8, and 11. simons.berkeley.edu/ai-tcs-worki...
7. Best works and technical talks in interviews.

- Emphasize quality over quantity. Evaluate candidates on no more than three  works of their choosing, with this fact explicit in the instructions for candidates and their letter writers.

- Test technical depth. Evaluate the depth of the candidates’ understanding of the aforementioned works more than was done normally until now.

8. Additional postdocs. Establish additional postdoctoral positions for the TCS community, a temporary measure that will be phased out after a few years.

9. Workshops for identifying new theory directions. First, organize a workshop (or several) with the goal of identifying a list of modern topics related to AI; and second, run a series of workshops on each of these topics to identify fundamental directions and establish foundations and agendas for future work.

10. “TCS in the Age of AI” series at the Simons Institute. Establish a recurring and widely accessible “TCS in the Age of AI” series at the Simons Institute.

11. Opportunities for training and evaluation without AI assistance. To ensure development of necessary skills for a successful career in academia, a PhD in TCS should partially include training and evaluation components where the use of AI is forbidden.
43112
Reposted by Thomas Dietterich
Thomas Dietterich @tdietterich.bsky.social · 28/09/2026
There are definitely multiple uses of "agent" in AI. The engineering view: A system designed to take actions in order to achieve goals. (Russell & Norvig).
151
Thomas Dietterich @tdietterich.bsky.social · 27/09/2026
Excellent guest post analyzing the (somewhat sparse) data we have on recursive self improvement
1111
Thomas Dietterich @tdietterich.bsky.social · 27/09/2026
Excellent essay by @philkoopman.bsky.social on how to regulate automated vehicle companies to be careful and competent. avsafety.substack.com/p/careful-an...
avsafety.substack.com
Careful & Competent Computer Drivers
Why aren’t we seeing data about reckless driving by AVs?
051
Reposted by Thomas Dietterich
Katie Martin @katie0martin.ft.com · 25/09/2026
I have blamed The agents That hacked Your website And which You probably Reasonably thought I controlled Forgive me I am rich And unaccustomed To taking responsibility
5776462129
Reposted by Thomas Dietterich
David Ho @davidho.bsky.social · 24/09/2026
This gatekeeping question is inevitable: "Does the project involve any work that could be interpreted as scientific research?"
856427
Thomas Dietterich @tdietterich.bsky.social · 23/09/2026
The number one most frustrating thing about windows 11 is making it go to sleep and stay asleep. Maybe it is just a problem with Lenovo laptops. But a slight jiggle in my backpack, the my laptop comes out of sleep and starts heating itself up. This is extremely stupid design.
361
Thomas Dietterich @tdietterich.bsky.social · 23/09/2026
Excellent analysis by Denis Stetkov open.substack.com/pub/techtren...
open.substack.com
How Four AI Labs Breached Nine Real Companies
One vendor's misconfiguration reached four AI labs. Nine intrusions, weak passwords and leaked credentials, and a boundary that existed only in the prompt.
021
Thomas Dietterich @tdietterich.bsky.social · 20/09/2026
Have people noticed that the word "evolution" is suddenly popular in machine learning papers that have nothing to do with modeling evolution or evolutionary algorithms? It seems to be used simply as a synonym for "training". I assume some LLM has started using it...
1122
Thomas Dietterich @tdietterich.bsky.social · 20/09/2026
A lesson from decades of work on automation is that autonomous systems require continual monitoring and human oversight. System designers never(*) succeed in anticipating all of the ways that a system can fail.
13917
Reposted by Thomas Dietterich
Yoshitomo Matsubara @yoshitomo-matsubara.net · 17/09/2026
Here is the followup bsky.app/profile/yosh...
051
Reposted by Thomas Dietterich
Gautam Kamath @gautamkamath.com · 16/09/2026
Nihar Shah did a heroic experiment for TMLR: he spent 20-25 hours over two weeks interviewing authors of seemingly low-quality submissions about their own papers. He confirmed what we all suspected: people submitting these papers have *no idea* what is going on in them.
5293117
Thomas Dietterich @tdietterich.bsky.social · 15/09/2026
It's not exponential; it's a step function!
2434
Thomas Dietterich @tdietterich.bsky.social · 13/09/2026
We are seeing a new trend in submissions to @arxiv.bsky.social (and presumably to conferences and journals): Authors submitting papers whose contents they likely do not understand. 1/
88524
Thomas Dietterich @tdietterich.bsky.social · 28/08/2026
Just discovered this very interesting blog post by @davidbau.bsky.social. davidbau.com/archives/202... It is an interesting formulation of the idea that it is up to humans to decide what world is worth building (and to take responsibility for the results)
davidbau.com
davidbau.com The Art of Wanting
0122
Reposted by Thomas Dietterich
Reece @thiccreese.bsky.social · 28/08/2026
My latest for @wired.com: I tried exercising my data access rights over 100 times and almost lost my sanity during the process. www.wired.com/story/i-dema...
wired.com
I Asked 100 Companies for My Data. I Got Deletion Notices Instead
California residents have a legal right to access the data that companies collect about them. Actually exercising that right is a burdensome nightmare.
524692
Thomas Dietterich @tdietterich.bsky.social · 25/08/2026
@bsky.app why do you always think I’m in Texas whenever I change my IP address? You are always wrong
110
Thomas Dietterich @tdietterich.bsky.social · 24/08/2026
"readout" seems to be the latest LLM word of the day, judging from the arXiv papers I'm moderating today
170
Thomas Dietterich @tdietterich.bsky.social · 14/08/2026
The agriculture water issue is definitely real. Water rights and pricing are crazy
130
Reposted by Thomas Dietterich
Birds of the World | The Cornell Lab @birdsoftheworld.bsky.social · 10/08/2026
The Cornell Lab invites applications for Edward W. Rose Postdoctoral Fellowships. These competitive awards support innovative, independent research by early-career scholars of exceptional promise. Two to four awards are made in each cycle. academicjobsonline.org/ajo/jobs/32334
academicjobsonline.org
Cornell University, Lab of Ornithology
Job #AJO32334, WDR-00059934 Rose Postdoctoral Associate, Cornell University, Ithaca NY, Cornell Lab of Ornithology, Lab of Ornithology, Cornell University, Ithaca, New York, US
14243
Reposted by Thomas Dietterich
Deyan Ginev @dginev.bsky.social · 12/08/2026
Today we add fresh arXiv HTML for 98.8% of articles with LaTeX sources, available immediately. Full summer reset. Rust rewrites rock!
0203
Reposted by Thomas Dietterich
Thomas Dietterich @tdietterich.bsky.social · 09/08/2026
I suspect that personal agents will not see a big uptake precisely because it is a form of programming that requires anticipating everything that could go wrong. And with agents, "everything" covers a lot.
251
Thomas Dietterich @tdietterich.bsky.social · 30/07/2026
I'm delighted that Dr. Penelope Lewis is joining arXiv as our CEO!
0173
Reposted by Thomas Dietterich
steinly0.bsky.social @steinly0.bsky.social · 28/07/2026
blog.arxiv.org/2026/07/28/r...
blog.arxiv.org
Remembering Professor Ralph Wijers
arXiv remembers our friend and colleague Professor Ralph Wijers
023
Thomas Dietterich @tdietterich.bsky.social · 06/07/2026
Trump's corruption damages everything it touches. It is hard to believe that FIFA's World Cup could become more corrupt, but Trump always finds a way
0110
Thomas Dietterich @tdietterich.bsky.social · 06/07/2026
Great post by Daphne Koller on linkedin: www.linkedin.com/posts/daphne...
linkedin.com
#ai | Daphne Koller | 19 comments
Today is not only the 250th anniversary of the US, it's also my own anniversary as an immigrant: on July 4, 1989, I arrived with two suitcases and little else to start a PhD in Stanford University's C...
050
Thomas Dietterich @tdietterich.bsky.social · 04/07/2026
One of @noahpinion.blogsky.venki.dev's best essays. open.substack.com/pub/noahpini...
open.substack.com
The American age was the human age
Reflections on 250 years.
021
Reposted by Thomas Dietterich
Pavel @spavel.bsky.social · 04/07/2026
Two businessmen meet. One says, I've got a truckload of vodka. The other says, I'll give you 300 dollars for it. They shake on the deal. Then they part ways — the first to try and find a truckload of vodka, the second to try and find $300. This is a joke about AI.
323837
Reposted by Thomas Dietterich
Mark Gongloff @markgongloff.bsky.social · 01/07/2026
On energy, Trump and the Republicans have dragged the US back to the 20th century while the rest of the world creeps into the future www.energyinst.org/exploring-en...
A chat of the change in carbon intensity of total energy supply (%). North America was up by 0.7%. Every other region was down. Asia Pacific was down by a full 1%. South and Central America by 0.8%, Africa by 0.7%, the whole world by 0.5%, roughly matching Europe. CIS was down about 0.1%. These are measured as CO2 emissions per EJ. Source: The Energy Institute
0147
Reposted by Thomas Dietterich
Pablo Samuel Castro @pcastr.bsky.social · 01/07/2026
I'm excited to share our latest work “Representation Learning Enables Scalable Multitask Deep Reinforcement Learning”. In this work, we revisit a fundamental question in reinforcement learning: What if representation learning was the key ingredient behind scalable RL? 1/🧵
15213
Reposted by Thomas Dietterich
petersuber @petersuber.fediscience.org.ap.brid.gy · 01/07/2026
Today #arXiv officially leaves Cornell to becomes an independent nonprofit org. Details: blog.arxiv.org/2026/06/30/arxivs-ne…
blog.arxiv.org
**On July 1, 2026, arXiv will spin out from Cornell University, its home for the past 25 years, to become an independent nonprofit organization.** With this next phase in arXiv’s journey quickly approaching, you can read more about arXiv’s history and the decision to spin out from Cornell in this recent article in the Cornell Chronicle. arXiv has been discussing the opportunity to become an independent organization for a long time, and we’re excited for the new flexibility and opportunities that this change will bring. This step is an important part of arXiv’s story that will have a positive impact as we continue to grow and has required great effort from staff to transition smoothly. Thanks to this, in the coming months, arXiv authors, readers, and the community that relies on us will not notice much change at all. arXiv was built for scientists, by scientists, and while our home may be changing, our mission, vision, and values never will. arXiv will remain committed to being free to read and to submit to, and we are dedicated to supporting scientists across the globe with equitable access to new ideas and discoveries. As we spin out from Cornell University, arXiv staff, volunteers, and our supporters are working hard to ensure there is no interruption to the important service arXiv provides. arXiv is also committed to transparency. We are a community resource, and we want to make sure that our community is aware of any changes to arXiv that may affect them. We also want to make sure the community has a better understanding of how changes support arXiv’s long-term goals as an essential scholarly publishing infrastructure. arXiv has created a page dedicated to frequently asked questions and information about the arXiv’s establishment as an independent organization. This FAQ page will be updated promptly as new information is available. On this dedicated FAQ page, you can find out more about the new arXiv’s leadership, governance, and special projects. While the FAQ page will be the best place to find the most up-to-date information on arXiv’s spinout, major updates will also be announced through arXiv’s blog and social media as they occur. In addition to the FAQ page, arXiv will be releasing a series of blogs to cover recent arXiv milestones, new organizational updates, and upcoming changes, including: * Introduction to arXiv’s new Engineering Director and dev team projects * arXiv’s 3 million submission milestone * Recent changes to arXiv policy practice around AI articles These blogs will be up in the coming weeks, and as always, comments from our community are welcome.
1157
Reposted by Thomas Dietterich
Aaron Roth @aaroth.bsky.social · 01/07/2026
write a paper that makes the key idea crystal clear. We are writing for people, and the writing should make it very easy for the reader to learn something. In a world like this, the bottleneck is not producing the proof but in digesting it, and we should review with that in mind.
1151
Reposted by Thomas Dietterich
Thomas Dietterich @tdietterich.bsky.social · 17/06/2026
Perhaps we need a "how to think" boot camp. An entire year of classes in which no computational tools are used. It would be like wilderness training. Build skills using paper, pencil, and a set of books. Build confidence in your own learning and reasoning abilities. 1/
2172
Reposted by Thomas Dietterich
Pavel @spavel.bsky.social · 10/09/2025
Solving any design problem (any wicked problem) requires exposing yourself to the vulnerability of knowing nothing. You will never know less about the problem than when you first start solving it. Design problems are solved the same way that we learn to do design: by experiment. 5/
1272
Reposted by Thomas Dietterich
ORCID @orcid.org · 10/06/2026
These are all great reasons for researchers to get started with an ORCID iD. 💫 Others include name disambiguation, name flexibility, and reduced administrative burden. Check out our page for researchers for more: info.orcid.org/researchers/ #ResearchSky
info.orcid.org
ORCID for Researchers
23218
Reposted by Thomas Dietterich
The Washington Post @washingtonpost.com · 08/06/2026
“My mayor Muslim. My bagel’s Jewish. My Christian Dior. Knicks in four!” MD Hossain’s ode to his beloved basketball team in a viral 15-second clip is being embraced by people looking for joy in a fractious national moment.
washingtonpost.com
He grabbed a mic to riff about the Knicks. His rhyme became an anthem of unity.
A fan blurted out an exuberant rhyme for his beloved basketball team. It has transcended sports and become a joyful anthem to unity.
210712
Reposted by Thomas Dietterich
Sara Beery @sarameghanbeery.bsky.social · 08/06/2026
There is a little less than a week left to apply for the 5th CV4Ecology Short Course!! This program is a one-of-a-kind opportunity for ecologists to DIRECTLY build robust, applied AI expertise that is immediately usable and useful in their own research, for their own projects, on their own data.
cv4ecology.csail.mit.edu
CV4Ecology Workshop
273
Thomas Dietterich @tdietterich.bsky.social · 08/06/2026
Excellent essay by @mazmhussain.bsky.social (and Anand Gopal). open.substack.com/pub/mazmhuss...
open.substack.com
Conversation with Anand Gopal: Author of "Days of Love and Rage"
Anand Gopal is someone whose work I had been following for years prior to the publication of his gripping new history of the Syrian revolution, “Days of Love and Rage” which was released this year.
030
Reposted by Thomas Dietterich
Murray Shanahan @mpshanahan.bsky.social · 07/06/2026
Ted Chiang's role-playing argument is essentially my role-playng argument (as he kindly acknowledges), which in turn was much inspired by the "Simulators" essay by Janus (who probably wouldn't endorse his conclusion). www.nature.com/articles/s41...
nature.com
Role play with large language models - Nature
By casting large-language-model-based dialogue-agent behaviour in terms of role play, it is possible to describe dialogue-agent behaviour such as (apparent) deception and (apparent) sel...
2103
Thomas Dietterich @tdietterich.bsky.social · 06/06/2026
Excellent and thought-provoking essay by Ted Chiang. [gift link] www.theatlantic.com/philosophy/2...
theatlantic.com
No, Artificial Intelligence Is Not Conscious
Taken to its logical conclusion, this line of thinking is absurd—and damning.
12312
Reposted by Thomas Dietterich
Robin Z @firepile.bsky.social · 04/06/2026
Possibly the best way to phish academics would be to just send them email that looks like a request to review an article and embed your malware in the "decline to review" link
824941
Thomas Dietterich @tdietterich.bsky.social · 29/05/2026
Interesting post by Jessica Hullman. In a future where autonomous research is possible at scale, one can imagine companies building their own bodies of scientific knowledge and not sharing it. Unlike human scientists, AI scientists are not motivated to share their work.
4283
Thomas Dietterich @tdietterich.bsky.social · 24/05/2026
Excellent post on Waymo's handling of safety incidents. open.substack.com/pub/philkoop...
open.substack.com
Robotaxi Safety Problems Are The New Normal
Standing down to manage risk is smart. But what needs to happen next?
0112
Thomas Dietterich @tdietterich.bsky.social · 23/05/2026
At @arxiv.bsky.social, we are receiving a new type of paper that I call an "I did this experiment" paper. These papers typically report some experiment with an LLM or LLM "agentic" workflow. They are the kind of experiments an "insider" engineer would run to optimize a system. 1/
95611
Reposted by Thomas Dietterich
Dalmeet Singh Chawla @dalmeet.bsky.social · 19/05/2026
#arXiv is banning researchers for one year whose manuscripts contain hallucinated references. My latest for @nature.com: www.nature.com/articles/d41... @briannosek.bsky.social, @tdietterich.bsky.social, @richardsever.bsky.social, @dermotlynott.bsky.social, @philipncohen.com, #WCRI2026
nature.com
Researchers who use hallucinated references to face arXiv ban
The preprint server is the latest to impose stiff penalties on authors who contribute to AI ‘slop’ — but not everyone is convinced it’s the right approach.
15029