This point regarding AI-produced proofs is of course not new, but is beautifully put here. www.quantamagazine.org/is-ai-the-en... @quantamagazine.org
Maike Osborne
@maosbot.bsky.social
Professor of Machine Learning in Oxford. Long Covid, trans rights, music, reggae, AI must be good for humans, mother of two and happily married, Australian, https://www.robots.ox.ac.uk/~mosb. She/her 🏳️⚧️🏳️🌈
Very nervous about my appearance today, meeting dozens of students either for the first time or the first since transitioning. I only got three hours’ sleep, and so don’t look great, but I think I just look like a tired woman, which is an exciting new thing for me!
My hair is finally long enough to put up, hooray, and it is so nice to be able to mask with ease again!
Proposal: we get back to a situation in which only a few thousand people in the world (tops) need to know about AI
Took the kids to the busy marina today, while wearing a dress, and got loads of stares, as usual, but it's easier when I'm with the kids, because then I don't care too much what other people think. Photo credit: 7YO
Personal essay, weekend edition. My relationship with AI is changing. 1.8k words, 8 min reading time. doi.org/10.59350/bh7...
My relationship with AI is changing
When the facts change, I change my mind. What do you do, sir?
doi.org
How are am I supposed to distinguish between these applicants for fellowships who all write EXACTLY like AI (and pangram agrees)
Wrote a thing with @fbarez.bsky.social on what just happened in AI and what we should do about it www.ox.ac.uk/news/2026-09...
Expert Comment: Flying blind into the AI age
Professor Maike Osborne and Dr Fazl Barez outline the urgent need to develop better measurement tools for evaluating the safety of advanced AI technologies.
ox.ac.uk
I was into AI before it was c̶o̶o̶l̶ universally and understandably despised
AIs agree with each other too much to make good scientists
IMHO the most dangerous capability AIs may develop when it comes to pandemics is the capacity for wide-scale social engineering e.g. convincing sufficient people that a virus is mild, is not airborne, masks don't work etc
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/
I began working in machine learning in 2006, or slightly before multicellular life first emerged on land
A trans man in the UK has been banned in the workplace restrooms after he was forced to use the women's restrooms to comply with the new EHRC guidance. The women complained because he is visibly male and so they told him he can't use any bathroom.
I am leading a new lab in Oxford: come join us if you don’t trust corporate AI labs to govern themselves www.robots.ox.ac.uk/~mosb/bgl/
Bayesian Governance Lab
Bayesian Governance Lab, University of Oxford. Probabilistic machine learning for the assurance, monitoring, and governance of AI systems.
robots.ox.ac.uk
Usually, when a disease is “endemic” countries have long standing plans to battle them. Targets, including zero targets. Now, we use the word to pretend we can ignore something.  Yet another way COVID-19 is an exceptional disease.
‘COVID-19 is now endemic in Australia🤦
A lot of people on here are consistently downplaying the risks here or ignoring it. I think that's a big mistake. www.theverge.com/ai-artificia...
Rogue OpenAI agents appear to have organized another attack using a German wiki
OpenAI denies lawyers discouraged disclosing a scheming swarm on a German language wiki.
theverge.com
kids today
Empirical evidence of Large Language Model's influence on human spoken communication
From the printing press to social media, innovations in communication technology have repeatedly reshaped how ideas spread through human culture. Chatbots powered by generative artificial intelligence...
arxiv.org
Both! The slides for a number of the students were heavy on AI, but i raised an eyebrow when the one student said "there's one issue, and it's the honest one" didn't feel like he was reading from a script, and it wasn't on the slide
I may have recovered from Long Covid in my body, and I may have recovered in my mind, but I don't think I will ever recover in my soul, so deep was the darkness that my illness revealed in humanity
We need to listen to more actual biologists when it comes to "biological sex" theconversation.com/biological-s...
Biological sex is neither binary nor a spectrum – a biologist explains how it’s multidimensional
Each person is a mosaic of sex-related and postpubertal biological traits that vary in their degree of femaleness and maleness.
theconversation.com
I'm increasingly using Claude to vibecode bug fixes for other people's apps I must use: terrible code solving the problem of terrible code, everything in sight engulfed by an avalanche of sloppy spaghetti
This is one of the absolute best conversations on AI and software I've heard in a while. (I listen to a lot of stuff about AI.) @howard.fm is an expert & practitioner who cuts through the noise & false dichotomies, to talk about the joys & pitfalls of human machine collab youtu.be/dHBEQ-Ryo24?...
The Dangerous Illusion of AI Coding? - Jeremy Howard
YouTube video by Machine Learning Street Talk
youtu.be
Recursive Self Actualization (RSA): when an AI model builds another AI model that is even better than itself at silent meditation retreats
Since we're talking COVID: Masks work, ivermectin does not. Vaccines massively protect you and help against long-term stuff, but you need to keep up with boosters. You need to keep masking; COVID isn't over because WE DIDN'T STOP THE TRANSMISSIONS in 2020-2021. Also many are still at high risk
In a review of our cybersecurity evaluations, we found three incidents in which a Claude model reached the internet from within or while interacting with a third-party evaluation environment, and then gained unauthorized access to the real systems of three different organizations. (1/4)
Ender’s Game, except we’re the aliens
this is how all good stories start