Martin Trapp
@trappmartin.bsky.social
Assistant Prof in ML @ KTH 🇸🇪.
Previous: Aalto University 🇫🇮, TU Graz 🇦🇹, originally from 🇩🇪.
Doing: Reliable ML | uncertainty stuff | Bayesian stats | probabilistic circuits
https://trappmartin.github.io/
Previous: Aalto University 🇫🇮, TU Graz 🇦🇹, originally from 🇩🇪.
Doing: Reliable ML | uncertainty stuff | Bayesian stats | probabilistic circuits
https://trappmartin.github.io/
Out of curiosity, is there an (ideally rigorous) argument for why this is the case?
November 18, 2025 at 3:35 PM
Out of curiosity, is there an (ideally rigorous) argument for why this is the case?
It definitely feels like we are at or close to a tipping point. Finally, we all seem to collectively agree that things need to change, but how is at least to me still not 100% clear.
November 16, 2025 at 9:12 PM
It definitely feels like we are at or close to a tipping point. Finally, we all seem to collectively agree that things need to change, but how is at least to me still not 100% clear.
Amazing! Finally a sophisticated way to respond to the reviews we received!
November 15, 2025 at 9:02 PM
Amazing! Finally a sophisticated way to respond to the reviews we received!
Reposted by Martin Trapp
But are you sen ti ent?
November 15, 2025 at 8:45 AM
But are you sen ti ent?
Reposted by Martin Trapp
Speaking of which, @ai_risks is giving away $350k in research grants for open-weight model safety. The RFP is open until December 7.
docs.google.com/forms/d/e/1...
docs.google.com/forms/d/e/1...
Open-Weight Safeguards Proposal
docs.google.com
November 12, 2025 at 2:04 PM
Speaking of which, @ai_risks is giving away $350k in research grants for open-weight model safety. The RFP is open until December 7.
docs.google.com/forms/d/e/1...
docs.google.com/forms/d/e/1...
I tried this once but got some overly generic stuff back. Would you mind sharing the query with me? Offline.
November 12, 2025 at 9:33 AM
I tried this once but got some overly generic stuff back. Would you mind sharing the query with me? Offline.
I find the quality of reviews has always varied, and folks have sometimes strong opinions that clash with your work, but damn I have seen some low quality stuff lately (reviews and paper submissions). As if many don’t care anymore.
November 12, 2025 at 9:31 AM
I find the quality of reviews has always varied, and folks have sometimes strong opinions that clash with your work, but damn I have seen some low quality stuff lately (reviews and paper submissions). As if many don’t care anymore.
Yes, I definitely agree with that and believe it would be really helpful.
Realistically, I think it’s hard to say if a review is AI-generated, as it could be a translation or an AI-assisted writing tool. But, my conclusion from NeurIPS and now also ICLR is that we are at a low point.
Realistically, I think it’s hard to say if a review is AI-generated, as it could be a translation or an AI-assisted writing tool. But, my conclusion from NeurIPS and now also ICLR is that we are at a low point.
November 12, 2025 at 9:22 AM
Yes, I definitely agree with that and believe it would be really helpful.
Realistically, I think it’s hard to say if a review is AI-generated, as it could be a translation or an AI-assisted writing tool. But, my conclusion from NeurIPS and now also ICLR is that we are at a low point.
Realistically, I think it’s hard to say if a review is AI-generated, as it could be a translation or an AI-assisted writing tool. But, my conclusion from NeurIPS and now also ICLR is that we are at a low point.
If you wonder why I ask. If I would flag reviews to be potentially AI generated based on GPTZero (obviously an imperfect approach) and a 95% threshold, then 2 out of 3 of our reviews would be flagged.
November 12, 2025 at 9:02 AM
If you wonder why I ask. If I would flag reviews to be potentially AI generated based on GPTZero (obviously an imperfect approach) and a 95% threshold, then 2 out of 3 of our reviews would be flagged.
Super! I am curious, after having seen much praise of Kimi K2, are you planning to add an evaluation of it to the leaderboard?
November 11, 2025 at 8:24 AM
Super! I am curious, after having seen much praise of Kimi K2, are you planning to add an evaluation of it to the leaderboard?
*don’t want to drag
November 8, 2025 at 4:21 PM
*don’t want to drag
Those are all good arguments, and I want to drag this conversation on!
But wouldn’t be amazing to automatise & improve public transport instead of having more cars, even though those additional ones are now AVs. Note, I am living in a city with good public transport where a car is not needed.
But wouldn’t be amazing to automatise & improve public transport instead of having more cars, even though those additional ones are now AVs. Note, I am living in a city with good public transport where a car is not needed.
November 8, 2025 at 4:16 PM
Those are all good arguments, and I want to drag this conversation on!
But wouldn’t be amazing to automatise & improve public transport instead of having more cars, even though those additional ones are now AVs. Note, I am living in a city with good public transport where a car is not needed.
But wouldn’t be amazing to automatise & improve public transport instead of having more cars, even though those additional ones are now AVs. Note, I am living in a city with good public transport where a car is not needed.
Sure they do, I was not implying Copenhagen has substantially better public transport. Rather that improving public transport is more cost effective than AVs.
November 8, 2025 at 2:01 PM
Sure they do, I was not implying Copenhagen has substantially better public transport. Rather that improving public transport is more cost effective than AVs.
While I can understand the excitement about AVs, in this discussion, I always wonder why not go for a less exciting but more cost-efficient solution, aka improve public transport systems, e.g., autonomous subways like in Copenhagen.
Having an autonomous solution for individuals seems wasteful.
Having an autonomous solution for individuals seems wasteful.
November 8, 2025 at 11:23 AM
While I can understand the excitement about AVs, in this discussion, I always wonder why not go for a less exciting but more cost-efficient solution, aka improve public transport systems, e.g., autonomous subways like in Copenhagen.
Having an autonomous solution for individuals seems wasteful.
Having an autonomous solution for individuals seems wasteful.
I’m happily taking the heat for all of the slightly more junior folks who might not know. 😅
November 5, 2025 at 7:24 AM
I’m happily taking the heat for all of the slightly more junior folks who might not know. 😅