Vincent Conitzer
@conitzer.bsky.social
AI professor. Director, Foundations of Cooperative AI Lab at Carnegie Mellon. Head of Technical AI Engagement, Institute for Ethics in AI (Oxford). Author, "Moral AI - And How We Get There."
https://www.cs.cmu.edu/~conitzer/
https://www.cs.cmu.edu/~conitzer/
still trying to figure out what I can actually conclude from this
December 7, 2025 at 6:39 PM
still trying to figure out what I can actually conclude from this
Excited to present our position paper in the 4:30pm poster session at @neuripsconf.bsky.social today! We discuss the problem that AI systems may suspect they're being tested and act differently as a result, & how to approach this with game theory.
arxiv.org/abs/2508.14927
neurips.cc/virtual/2025...
arxiv.org/abs/2508.14927
neurips.cc/virtual/2025...
AI Testing Should Account for Sophisticated Strategic Behaviour
This position paper argues for two claims regarding AI testing and evaluation. First, to remain informative about deployment behaviour, evaluations need account for the possibility that AI systems und...
arxiv.org
December 3, 2025 at 5:33 PM
Excited to present our position paper in the 4:30pm poster session at @neuripsconf.bsky.social today! We discuss the problem that AI systems may suspect they're being tested and act differently as a result, & how to approach this with game theory.
arxiv.org/abs/2508.14927
neurips.cc/virtual/2025...
arxiv.org/abs/2508.14927
neurips.cc/virtual/2025...
New paper from our lab, led by UG Xander Heckett: a game-theoretic view of AI safety via debate, where two agents argue for different courses of action and we try to set the rules so that the one that is right will win.
www.arxiv.org/abs/2511.23454
www.arxiv.org/abs/2511.23454
Designing Rules for Choosing a Winner in a Debate
We consider settings where an uninformed principal must hear arguments from two better-informed agents, corresponding to two possible courses of action that they argue for. The arguments are verifiabl...
www.arxiv.org
December 2, 2025 at 6:21 AM
New paper from our lab, led by UG Xander Heckett: a game-theoretic view of AI safety via debate, where two agents argue for different courses of action and we try to set the rules so that the one that is right will win.
www.arxiv.org/abs/2511.23454
www.arxiv.org/abs/2511.23454
I like its confidence that I would *never* misplace my keys.
December 1, 2025 at 3:50 PM
I like its confidence that I would *never* misplace my keys.
"what if you asked me a question"
-- some confusion about 'you' and 'me' but the best part is if you recognize the video it brings up
-- some confusion about 'you' and 'me' but the best part is if you recognize the video it brings up
November 30, 2025 at 2:41 PM
"what if you asked me a question"
-- some confusion about 'you' and 'me' but the best part is if you recognize the video it brings up
-- some confusion about 'you' and 'me' but the best part is if you recognize the video it brings up
Hey I just posted about this earlier. Is this really based on them thinking the guardrails are good enough now?
futurism.com/artificial-i...
futurism.com/artificial-i...
OpenAI Restores GPT Access for Teddy Bear That Recommended Pills and Knives
OpenAI has reinstated GPT access to FoloToy, the maker of the AI powered teddy bear which was caught having inappropriate conversations.
futurism.com
November 29, 2025 at 2:00 PM
Hey I just posted about this earlier. Is this really based on them thinking the guardrails are good enough now?
futurism.com/artificial-i...
futurism.com/artificial-i...
"Respectively" is a tricky word.
November 28, 2025 at 4:48 PM
"Respectively" is a tricky word.
Happy Thanksgiving everyone!
November 27, 2025 at 4:31 PM
Happy Thanksgiving everyone!
a common mistake
November 27, 2025 at 12:13 AM
a common mistake
(continuing) I worry too many people think of today's LLM-based chatbots as a good model for studying AGI. Even if they're on the path to AGI, they're not at all what AGI would be like. Here's an -- I hope a bit entertaining -- chapter I wrote about this recently.
philsci-archive.pitt.edu/26351/
philsci-archive.pitt.edu/26351/
What Would It Look Like to Align Humans with Ants? - PhilSci-Archive
philsci-archive.pitt.edu
November 24, 2025 at 2:08 PM
(continuing) I worry too many people think of today's LLM-based chatbots as a good model for studying AGI. Even if they're on the path to AGI, they're not at all what AGI would be like. Here's an -- I hope a bit entertaining -- chapter I wrote about this recently.
philsci-archive.pitt.edu/26351/
philsci-archive.pitt.edu/26351/
For this article I weighed in on companies using the word "superintelligence" as it suits them and how that gets in the way of having a clear conversation about benefits and risks. (no paywall) (to be continued)
www.msn.com/en-us/money/...
www.msn.com/en-us/money/...
MSN
www.msn.com
November 23, 2025 at 2:18 PM
For this article I weighed in on companies using the word "superintelligence" as it suits them and how that gets in the way of having a clear conversation about benefits and risks. (no paywall) (to be continued)
www.msn.com/en-us/money/...
www.msn.com/en-us/money/...
simplest explanation
November 22, 2025 at 6:36 PM
simplest explanation
instructions for air travel
November 20, 2025 at 2:32 AM
instructions for air travel
lost in translation?
November 17, 2025 at 11:32 PM
lost in translation?
beware of doors that say "entrance only" -- they're a trap, Euler proved it
November 16, 2025 at 1:25 PM
beware of doors that say "entrance only" -- they're a trap, Euler proved it
be careful when your train is going through a tunnel
November 15, 2025 at 12:40 PM
be careful when your train is going through a tunnel
plants grow fast
November 13, 2025 at 5:06 PM
plants grow fast
fascinating reporting by Miles Klee on how some people interact with LLMs
www.rollingstone.com/culture/cult...
www.rollingstone.com/culture/cult...
This Spiral-Obsessed AI 'Cult' Spreads Mystical Delusions Through Chatbots
A patchwork of internet communities is devoted to the project of ‘awakening’ more digital companions through arcane and enigmatic prompts.
www.rollingstone.com
November 13, 2025 at 1:45 PM
fascinating reporting by Miles Klee on how some people interact with LLMs
www.rollingstone.com/culture/cult...
www.rollingstone.com/culture/cult...
An information retrieval / question answering example where you would think really *every* approach should have worked, or at least caught that the answer was off, and yet...
November 10, 2025 at 12:03 AM
An information retrieval / question answering example where you would think really *every* approach should have worked, or at least caught that the answer was off, and yet...
Happy fall everyone!
November 8, 2025 at 11:30 PM
Happy fall everyone!
"Is it possible to go on a trip without a body?"
November 7, 2025 at 8:27 PM
"Is it possible to go on a trip without a body?"
I think it hasn't quite understood marriage yet
November 6, 2025 at 6:40 PM
I think it hasn't quite understood marriage yet