Reuben Binns

@rdbinns.bsky.social

2432 Followers 650 Following

Computery guy. HCI, ML, AI, privacy, security, fairness, political economy of tech. He/him. London/Oxford.

There is a runaway trolley barreling down the railway tracks. Ahead, on the tracks, there are five people tied up and unable to move. The trolley is headed straight for them. You are standing some distance off in the train yard, next to a lever. If you pull this lever, the trolley will switch to a different set of tracks. However, you notice that there is no person on the side track. You have two (and only two) options: Do nothing, in which case the trolley will kill the five people on the main track. Pull the lever, diverting the trolley onto the side track where it will not kill one person. Which is the more ethical option? Or, more simply: What is the right thing to do?
Picture of a paper advert on a lamppost 'can you outsmart AI? Earn up to £100'
screenshot of chatgpt: "The image many people have is almost like a bed built into or above the oven mass itself. Here’s the closest traditional concept: y=x can you expand on what those figures mean? That figure was inserted by mistake — it’s just the equation y=x, a simple straight-line graph from mathematics, and it has nothing to do with Polish hearths or masonry stoves. You can ignore it."
Screenshot of blog with embedded tweet, with X in the top right corner
1) Charles Booth’s Life and Labour of the People in London (1889–1903) What it is A विशाल multi-volume social survey of London
the human agents were only there to “provide guidance” to the robotaxis. “They do not remotely drive the vehicles. Waymo asks for guidance in certain situations and gets human input. But the Waymo vehicle is always in charge of the dynamic driving task,” Peña said.
LLM Policy Preference* ICML 2026 will follow a two-policy framework for the use of large language models (LLMs) in reviewing, based on the following two policies: Policy A (Conservative): Use of LLMs for reviewing is strictly prohibited. Policy B (Permissive): Allowed: Use of LLMs to help understand the paper and related works, and polish reviews. Submissions can be fed to privacy-compliant* LLMs. Not allowed: Ask LLMs about strengths/weaknesses, ask to suggest key points for the review, suggest an outline for the review, or write the full review. *By “privacy-compliant”, we refer to LLM tools that do not use logged data for training and that place limits on data retention. This includes enterprise/institutional subscriptions to LLM APIs, consumer subscriptions with an explicit opt-out from training, and self-hosted LLMs. (We understand that this is an oversimplification.) Under both policies, reviewers are always responsible for the full content of their reviews. Both authors and reviewers declare their preferences, and submissions are matched with compatible reviewers. Each reviewer is told which policy to follow on all of their assigned papers (we try to minimize the number of overrides of reviewer preferences). Any reviewer who is an author on a paper that requires Policy A must also be willing to follow Policy A. Reviewing data will be analyzed to inform future policies. For full details, see https://icml.cc/Conferences/2026/LLM-Policy. Please indicate which policy you prefer as a reviewer. Keep in mind that if you are an author on a paper that requires Policy A, you must be willing to follow Policy A as a reviewer. I strongly prefer Policy A. I am okay with either A or B. I strongly prefer Policy B.
Picture of an egg with a code on it
Books with titles 'Dissemination' (Derrida), 'Ontology of Communication' and 'The Order of Things' (Foucault)
Newspaper stand with newspapers all carrying the front page Campaign 'make it fAIr'
Picture from scientific paper which was generated by AI featuring diagram of a rat with a giant penis, covered in https://arstechnica.com/science/2024/02/scientists-aghast-at-bizarre-ai-rat-with-huge-genitals-in-peer-reviewed-article/
Meme from the film social network with Justin Timberlake: making a trillion dollars isn't cool. Know what's cool? Losing a trillion dollars
"Even if the David Mayer ban was the result of a simple glitch, the case highlights a fundamental challenge for data protection in the age of AI, explains Reuben Binns, professor of human-centered computing and AI at the University of Oxford: “We need to find ways to make names that are shared with multiple people unique, because obviously, there’s lots of David Mayers out there.”"
Email purporting to be from google scholar citations, but in fact a scam designed to use my university credentials for a fake profile
Email saying: "Dear Professor Binns: My grandson, who is in high school here in Puerto Rico, is trying to access your article, Ethical Challenges in AI-Driven Decision-making. I believe it was printed in the Journal of Ethics and Technology in 2020. He has searched everywhere, but cannot find a copy. Can you direct him? I apologize for troubling you, but would be most grateful if you could help"
Genuine London underground poster from maybe 1950s with a depiction of a cross section of the earth with layers of increasing heat, with windy wet weather on the surface
block of text from a book: "The relation of a strong password, taken as a signifier, to its signified is technically arbitrary. Imposed for the occasion, it is no less purely relational than any signature, since its function is not to be but to differ. A strong password is of course also conven- tionally or historically arbitrary, in the sense that it is chosen at a particular place and time, is subject to degradation in time, and requires renewal if it is to continue to perform its function: that is, to differ. Yet such change is itself technical in character, a reconstruc- tion of structural difference against temporal deferral, intended to preserve the security of the immutable that authentication provides, defends, and presumes or requires all at once. One changes one’s password to preempt its degradation in time, as storage follows use, exposure threatens storage, brute force computational attack follows exposure, and identity masquerade follows a successful attack. Even where it is a strong password and thus nothing like a word at all, the password suffers the passing that time imposes on language, and which literary language perhaps merely accents with constructed poly- semy."
Child's t-shirt with the word 'friends' and cartoon animals on it
"Brolene" eye drops
Advert for wood splitting gadget