Luc Rocher
@rocher.lc
associate professor at Oxford · UKRI future leaders fellow · i study how data and algorithms shape societies · AI fairness, accountability and transparency · algorithm auditing · photographer, keen 🚴🏻 · they/them · https://rocher.lc (views my own)
📸 New research just out revealing disturbing problems with AI image captioning. AI models can now analyse both texts and images, and are trained on vast collections of human-created content spanning centuries. Some claims these models could soon help human historians interpret and explain the past.
November 15, 2025 at 12:13 PM
📸 New research just out revealing disturbing problems with AI image captioning. AI models can now analyse both texts and images, and are trained on vast collections of human-created content spanning centuries. Some claims these models could soon help human historians interpret and explain the past.
Is there a scientific crisis in AI evaluations‽ We did the hard work of reviewing all recent AI benchmarks in top AI conferences. Amongst the 445 reviewed benchmarks, few offer rigorous evaluations of AI capabilities, making it difficult to actually track advances in the field and compare models.
November 4, 2025 at 12:03 PM
Is there a scientific crisis in AI evaluations‽ We did the hard work of reviewing all recent AI benchmarks in top AI conferences. Amongst the 445 reviewed benchmarks, few offer rigorous evaluations of AI capabilities, making it difficult to actually track advances in the field and compare models.
Facial recognition just passed 99.95% accuracy in the latest evaluations by the US National Institute of Standards and Technology.
But did you know that these numbers come from pristine lab conditions with small datasets, good lightning, and clear photos?
But did you know that these numbers come from pristine lab conditions with small datasets, good lightning, and clear photos?
August 18, 2025 at 4:19 PM
Facial recognition just passed 99.95% accuracy in the latest evaluations by the US National Institute of Standards and Technology.
But did you know that these numbers come from pristine lab conditions with small datasets, good lightning, and clear photos?
But did you know that these numbers come from pristine lab conditions with small datasets, good lightning, and clear photos?
🧵New research! People now use AI chatbots for therapy, friendship, and romance. These days, both specialised and general-purpose chatbots are built with a friendly, empathetic tone to best comfort users. We find this new trend is not just cosmetic but seriously harms user safety.
August 1, 2025 at 11:51 AM
🧵New research! People now use AI chatbots for therapy, friendship, and romance. These days, both specialised and general-purpose chatbots are built with a friendly, empathetic tone to best comfort users. We find this new trend is not just cosmetic but seriously harms user safety.
just received a promotion to associate professor, so i took this opportunity to redo my website (rocher.lc, feedback welcomed)! it looks a bit more boring that the previous version, but easier to read.
and thank you everyone who supported me through the years, wouldn't have made it here with you! 🌱
and thank you everyone who supported me through the years, wouldn't have made it here with you! 🌱
June 30, 2025 at 7:32 PM
just received a promotion to associate professor, so i took this opportunity to redo my website (rocher.lc, feedback welcomed)! it looks a bit more boring that the previous version, but easier to read.
and thank you everyone who supported me through the years, wouldn't have made it here with you! 🌱
and thank you everyone who supported me through the years, wouldn't have made it here with you! 🌱
@sofiahafner.bsky.social gave an excellent talk about our work, which shows that as language models get larger and bigger, they also learn more stereotypical, binary associations between gender and sex. More here: arxiv.org/abs/2505.14080
June 29, 2025 at 7:28 PM
@sofiahafner.bsky.social gave an excellent talk about our work, which shows that as language models get larger and bigger, they also learn more stereotypical, binary associations between gender and sex. More here: arxiv.org/abs/2505.14080
Had a lovely time at FAccT in Athens earlier this week. Attached photos of me, beans, cat hotels, lime&basil ice cream.
June 29, 2025 at 7:26 PM
Had a lovely time at FAccT in Athens earlier this week. Attached photos of me, beans, cat hotels, lime&basil ice cream.
How is your Friday going? This is me trying to figure out how to get a new French passport after the UK Home Office lost it somewhere in Swindon…
February 14, 2025 at 6:49 PM
How is your Friday going? This is me trying to figure out how to get a new French passport after the UK Home Office lost it somewhere in Swindon…
Not often do we see an official FRS, Fellow of the Royal Society, performing a heil hitler salute live in front of such a large audience. Surely the Royal Society is already investigating and will take a prompt decision in the coming days.
January 21, 2025 at 7:55 PM
Not often do we see an official FRS, Fellow of the Royal Society, performing a heil hitler salute live in front of such a large audience. Surely the Royal Society is already investigating and will take a prompt decision in the coming days.
Mixed feelings about this new AI Opportunities Action Plan. Surprised to see how weak the data strategy is.
Some points just don't make sense, e.g. pushing for a single form of privacy-preserving access to sensitive data (synthetic data) which has been shown to provide very little usability.
Some points just don't make sense, e.g. pushing for a single form of privacy-preserving access to sensitive data (synthetic data) which has been shown to provide very little usability.
January 13, 2025 at 8:10 PM
Mixed feelings about this new AI Opportunities Action Plan. Surprised to see how weak the data strategy is.
Some points just don't make sense, e.g. pushing for a single form of privacy-preserving access to sensitive data (synthetic data) which has been shown to provide very little usability.
Some points just don't make sense, e.g. pushing for a single form of privacy-preserving access to sensitive data (synthetic data) which has been shown to provide very little usability.
Curious to see how they plan to anonymise our NHS data for training ML models. I personally wouldn’t know how to start doing that
www.theguardian.com/politics/202...
www.theguardian.com/politics/202...
January 13, 2025 at 9:29 AM
Curious to see how they plan to anonymise our NHS data for training ML models. I personally wouldn’t know how to start doing that
www.theguardian.com/politics/202...
www.theguardian.com/politics/202...
While they are likely even more accurate models out there, this scaling law is already much better than previously-used heuristics using exponential decay or polynomials.
Interestingly, it's also far better than relying on entropy (think “33 bits of information enough to identify anyone on earth”).
Interestingly, it's also far better than relying on entropy (think “33 bits of information enough to identify anyone on earth”).
January 9, 2025 at 5:07 PM
While they are likely even more accurate models out there, this scaling law is already much better than previously-used heuristics using exponential decay or polynomials.
Interestingly, it's also far better than relying on entropy (think “33 bits of information enough to identify anyone on earth”).
Interestingly, it's also far better than relying on entropy (think “33 bits of information enough to identify anyone on earth”).
During my PhD, I taught myself Bayesian statistics and discovered a beautiful field of work on nonparametrics models. That's what we now exploit here to characterise what makes individuals identifiable, resulting in a simple scaling law for the {accuracy/correctness/rank-1 identification rate}.
January 9, 2025 at 5:07 PM
During my PhD, I taught myself Bayesian statistics and discovered a beautiful field of work on nonparametrics models. That's what we now exploit here to characterise what makes individuals identifiable, resulting in a simple scaling law for the {accuracy/correctness/rank-1 identification rate}.
Entire fields test biometrics and human id. models on small-scale benchmarks, reporting accuracy, AUC, true positive rate, etc. that do not match performance in real-world where id. is much harder.
Worse, a model A better than a model B in small-scale isn't necessarily better at scale!
Worse, a model A better than a model B in small-scale isn't necessarily better at scale!
January 9, 2025 at 5:07 PM
Entire fields test biometrics and human id. models on small-scale benchmarks, reporting accuracy, AUC, true positive rate, etc. that do not match performance in real-world where id. is much harder.
Worse, a model A better than a model B in small-scale isn't necessarily better at scale!
Worse, a model A better than a model B in small-scale isn't necessarily better at scale!
🔴 Finally out, my work in @naturecomms.bsky.social investigates (re-)identification techniques from browser fingerprinting to facial recognition.
It's a surprisingly tough problem, with incorrect heuristics and rules of thumbs often used—huge issue for accountability and independent investigations.
It's a surprisingly tough problem, with incorrect heuristics and rules of thumbs often used—huge issue for accountability and independent investigations.
January 9, 2025 at 5:07 PM
🔴 Finally out, my work in @naturecomms.bsky.social investigates (re-)identification techniques from browser fingerprinting to facial recognition.
It's a surprisingly tough problem, with incorrect heuristics and rules of thumbs often used—huge issue for accountability and independent investigations.
It's a surprisingly tough problem, with incorrect heuristics and rules of thumbs often used—huge issue for accountability and independent investigations.
Surprised that private actors are even allowed to participate in digital ID efforts. France has had one single app for years, it’s a very simple app, it works fine, no need for dozens of them.
December 21, 2024 at 9:54 AM
Surprised that private actors are even allowed to participate in digital ID efforts. France has had one single app for years, it’s a very simple app, it works fine, no need for dozens of them.
You might say, well, at least they can hardly identify users, right? Right?
December 13, 2024 at 12:25 AM
You might say, well, at least they can hardly identify users, right? Right?
To comply with their privacy policy, Anthropic built a new privacy-preserving analytics platform that creates anonymized insights from all free and Pro Claude customers' chats.
How do they look for any privacy risks? They just ask Claude.
How do they look for any privacy risks? They just ask Claude.
December 13, 2024 at 12:19 AM
To comply with their privacy policy, Anthropic built a new privacy-preserving analytics platform that creates anonymized insights from all free and Pro Claude customers' chats.
How do they look for any privacy risks? They just ask Claude.
How do they look for any privacy risks? They just ask Claude.
Adding bureaucratic requirements to a burdened system risks leaving the NHS vulnerable to exploitation by private technology companies whose offers to “assist” with infrastructure development could result in loss of control over valuable public assets.
@oiioxford.bsky.social @yaledec.bsky.social
@oiioxford.bsky.social @yaledec.bsky.social
December 12, 2024 at 2:48 PM
Adding bureaucratic requirements to a burdened system risks leaving the NHS vulnerable to exploitation by private technology companies whose offers to “assist” with infrastructure development could result in loss of control over valuable public assets.
@oiioxford.bsky.social @yaledec.bsky.social
@oiioxford.bsky.social @yaledec.bsky.social
New editorial in @bmj.com by @jessrmorley.bsky.social and myself! Technical solutions are needed to modernise the NHS health data infrastructure. Investment in privacy preserving platforms with robust security measures is key.
🔗 bmj.com/cgi/content/full/bmj.q2735?ijkey=NzzxQSQlpEvi3Mx
🔗 bmj.com/cgi/content/full/bmj.q2735?ijkey=NzzxQSQlpEvi3Mx
December 12, 2024 at 2:48 PM
New editorial in @bmj.com by @jessrmorley.bsky.social and myself! Technical solutions are needed to modernise the NHS health data infrastructure. Investment in privacy preserving platforms with robust security measures is key.
🔗 bmj.com/cgi/content/full/bmj.q2735?ijkey=NzzxQSQlpEvi3Mx
🔗 bmj.com/cgi/content/full/bmj.q2735?ijkey=NzzxQSQlpEvi3Mx
Only 11% believe that AI tools have reduced their workload.
All results in French: solidairesfinancespubliques.org/vie-des-serv...
All results in French: solidairesfinancespubliques.org/vie-des-serv...
December 11, 2024 at 11:34 PM
Only 11% believe that AI tools have reduced their workload.
All results in French: solidairesfinancespubliques.org/vie-des-serv...
All results in French: solidairesfinancespubliques.org/vie-des-serv...
Yes! The traditionnal version with almond cream plus crème pâtissière. 🥰 Attached my notes on galettes, it has the most fascinating history.
November 16, 2024 at 3:30 PM
Yes! The traditionnal version with almond cream plus crème pâtissière. 🥰 Attached my notes on galettes, it has the most fascinating history.
📝 A few weeks ago, I started a checklist for scientific writing and editing. I cover how to set a narrative; improve paragraph structure/transitions/signposting; write in a simple and precise way; and organise introduction, results, discussion, and methods.
Feedback is very welcome! osf.io/4cyek
Feedback is very welcome! osf.io/4cyek
November 12, 2024 at 4:55 PM
📝 A few weeks ago, I started a checklist for scientific writing and editing. I cover how to set a narrative; improve paragraph structure/transitions/signposting; write in a simple and precise way; and organise introduction, results, discussion, and methods.
Feedback is very welcome! osf.io/4cyek
Feedback is very welcome! osf.io/4cyek
Never seen any so far. Does that still happen when hiding what they call "non-sexual nudity"? Might be images flagged as light nudity but w/o content warning.
November 11, 2024 at 9:44 PM
Never seen any so far. Does that still happen when hiding what they call "non-sexual nudity"? Might be images flagged as light nudity but w/o content warning.