Peter Bloem
@pbloem.sigmoid.social.ap.brid.gy
Assistant prof. at the Learning and Reasoning group, Vrije Universiteit Amsterdam (searchable).
[bridged from https://sigmoid.social/@pbloem on the fediverse by https://fed.brid.gy/ ]
[bridged from https://sigmoid.social/@pbloem on the fediverse by https://fed.brid.gy/ ]
@leonoverweel I totally agree (this is one of my main uses of ChatGPT at the moment).
I don't disagree with this feature. My only worry is that they'll remove the original Scholar when they feel that not enough people are using the new version.
I don't disagree with this feature. My only worry is that they'll remove the original Scholar when they feel that not enough people are using the new version.
November 23, 2025 at 9:54 AM
@leonoverweel I totally agree (this is one of my main uses of ChatGPT at the moment).
I don't disagree with this feature. My only worry is that they'll remove the original Scholar when they feel that not enough people are using the new version.
I don't disagree with this feature. My only worry is that they'll remove the original Scholar when they feel that not enough people are using the new version.
We were so busy worrying about whether Google would kill Scholar, we didn't stop to think they might use it as another vector to push their AI.
November 22, 2025 at 9:34 AM
We were so busy worrying about whether Google would kill Scholar, we didn't stop to think they might use it as another vector to push their AI.
@futurebird The bees at Buckingham palace got some sad news quite recently.
https://www.independent.co.uk/life-style/royal-family/royal-beekeeper-bees-queen-death-b2164345.html
https://www.independent.co.uk/life-style/royal-family/royal-beekeeper-bees-queen-death-b2164345.html
Royal beekeeper informs Buckingham Palace bees that the Queen has died
The palace has five beehives, each containing around 20,000 bees
www.the-independent.com
November 8, 2025 at 11:45 AM
@futurebird The bees at Buckingham palace got some sad news quite recently.
https://www.independent.co.uk/life-style/royal-family/royal-beekeeper-bees-queen-death-b2164345.html
https://www.independent.co.uk/life-style/royal-family/royal-beekeeper-bees-queen-death-b2164345.html
As an epilogue, here is the proof of the main theorem with my annotations.
As proofs go it's pretty simple, mostly building on set theory and some juggling of inequalities.
The key structure is given above the heading: start with the statement of the […]
[Original post on sigmoid.social]
As proofs go it's pretty simple, mostly building on set theory and some juggling of inequalities.
The key structure is given above the heading: start with the statement of the […]
[Original post on sigmoid.social]
October 28, 2025 at 5:32 PM
As an epilogue, here is the proof of the main theorem with my annotations.
As proofs go it's pretty simple, mostly building on set theory and some juggling of inequalities.
The key structure is given above the heading: start with the statement of the […]
[Original post on sigmoid.social]
As proofs go it's pretty simple, mostly building on set theory and some juggling of inequalities.
The key structure is given above the heading: start with the statement of the […]
[Original post on sigmoid.social]
This bring me to the second question we had about the paper. Does it suggest hallucinations are a fundamental part of the way we do AI? No. It suggests non-hallucinating models are perfectly possible (if we give up calibration) and it offers a very cheap and simple hack to help us better find […]
Original post on sigmoid.social
sigmoid.social
October 28, 2025 at 5:24 PM
This bring me to the second question we had about the paper. Does it suggest hallucinations are a fundamental part of the way we do AI? No. It suggests non-hallucinating models are perfectly possible (if we give up calibration) and it offers a very cheap and simple hack to help us better find […]
The authors suggest a global confidence level (I think), but you could also change it from question to question to see how well the model adapts.
The key point is that this is a very simple adaptation to existing benchmarks. It doesn't cost much to simply run the same benchmark again with this […]
The key point is that this is a very simple adaptation to existing benchmarks. It doesn't cost much to simply run the same benchmark again with this […]
Original post on sigmoid.social
sigmoid.social
October 28, 2025 at 5:22 PM
The authors suggest a global confidence level (I think), but you could also change it from question to question to see how well the model adapts.
The key point is that this is a very simple adaptation to existing benchmarks. It doesn't cost much to simply run the same benchmark again with this […]
The key point is that this is a very simple adaptation to existing benchmarks. It doesn't cost much to simply run the same benchmark again with this […]