Reuben Binns

A year on from releasing our audit of Uber's pricing and pay algorithms, our collaborators @workerinfox.bsky.social have filed this class action lawsuit in Amsterdam. Drivers across Europe may win €billions compensation for violations of their rights under GDPR: www.theguardian.com/technology/2...

I've been on the advisory board for this excellent PLP project looking into the transparency challenges of applying public law to public sector AI use. The final report is full of insights and recommendations, including statutory footing for the Algorithmic Transparency Reporting Standard

Reading AI prose lately reminds me of being an undergraduate and proof-reading dissertations of friends writing under the influence of unprescribed modafinil. It's not *un*grammatical, it's semantically dense, it's impressively abstract. But that very combination is somehow just unpleasant to read.

"Education is not rate-limited by the clarity of explanation, but by motivation: do people actually want to learn the material ... a huge part of what motivates people to learn is seeing somebody else deeply passionate about it" @3blue1brown.com on math after AI www.youtube.com/watch?v=LKrw...

No joke. I think this is the most important video I have ever seen about mobile games. This guy breaks down how this game psychologically profiles you, and makes the game harder or easier in order to show you ads

'we need a bajillion million dollars worth of GPUs and data centers to run our AI models! Oh but, also, maybe they could run on a network of smart fridges, toothbrushes and baby monitors'

13. Europeans shouldn't feel guilty about using electricity for airconditioning (or reversible heatpumps), it already mostly comes from the sun, especially if you pre-cool. This is good news for hot countries as well. Heat is deadly.

AI alignment is at the centre of the growing debates about AI systems and their failures. The impulses of AI alignment work are however *not* new. Making sure computational systems do what we want them to do has long been a core goal of computing research.

This graph is a good reason to introduce my new paper: By Generation and by Design: A Review of the Empirical Rent Control Literature. This paints a (somewhat) misleading picture of rent control evidence because it groups all rent control together regardless of design feature.

Algorithmic Realism: Data Science Practices to Promote Social Justice comes out today! I wrote this book with the mission to transform data science and AI into fields that genuinely work in service of the public interest. I hope that you’ll join me! mitpress.mit.edu/978026205570...

LLM chatbot surveillance advertising is going to be next-level disastrous, on every level.

The web and mobile apps of major conversational AI engines transmit prompts, titles and conversation permalinks to advertising data firms, alongside device IDs, hashed email and other persistent identifiers. Comprehensive analysis, like always from this team: jorgegarciaherrero.com/wp-content/i...

Glovo is already facing a labour inspectorate investigation over continuing to operate during a heatwave in Barcelona in the summer. It should also be held to account for operating during a flood in Bilbao. No rider should be pounding the streets during extreme weather. bsky.app/profile/fons...

The proposed social media ban, the EU KIDS Act, is perhaps the most chaotic law I have ever read! New paper: Are EU Kidding Me? Notes from a Poorly Drafted Social Media Ban. files.michae.lv/papers/veale_areEUkiddingme.pdf In the spirit of the proposal I have stayed offline and read no takes.

Interested in, or already using (for academic or industry research) digital twins? In this analysis, digital twins' responses "correlate weakly [0.20] with human responses". Not nothing but.. not great if goal is reliable knowledge or basis for consequential decisions www.science.org/doi/10.1126/...

On my way to the government’s expert advisory panel on self-driving vehicles. Passed this guy, stopped in a pedestrian crossing. Oops.

I'm hearing reports that Volkswagen's AI has been scheming to deceive its evaluators into thinking its emissions are safe. We need to shut down all cars until they are climate-aligned.

I recently had the pleasure of being interviewed for The Inference podcast, which has dropped today. We talk about the research I've done with @workerinfox.bsky.social and Uber drivers, the concept of a 'human in the loop'(hole), the state of AI, and more: youtu.be/dHJLcSl18RI?...

My space ships and gliders are so intelligent they have broken containment and are collaborating and self replicating in the wild. I urgently need a trillion dollars to align them with humanity or there's a 10% chance all our pixels will be killed.

I think the relevant caselaw here is shaping up to be Rylands & Fletcher. If you collect up any inherently risky material or item on your property and it then escapes and causes damage, it creates strict liability. ie, you do not have to prove liability, you go straight to quantum of compensation.

Very happy to have in our! trembling hands the hard copy of Law, Policy and the Internet 2nd edn( Hart/Bloomsbury), just out in the nick of time for first semester 2026/7! Order or inspection copies at share.google/RCRm2lmLQNAb... & do note a guide to teaching sample "tech law" course with it!

HARASSED by pests!

Based upon their documented behavior over the past few years, the tech industry is less afraid of the potential ramifications of Artificial Superintelligence than they are of Lina Khan.

Lads, what are we doing here? First, we didn't want freely editable text documents, so we made fixed PDFs. Then, we changed our minds and wanted people to be able to edit those PDFs. Then we changed our minds again and put these wonderful little 'lock form' buttons in the PDF 😆

This would actually be a fun game to play for a human (me)

Can any central bank economists explain to me how making mortgages more expensive will reverse the impact of extreme-weather-induced-reduced-crop-yields on food prices, end the US-Iran war and resulting fuel shortages, refill the Panama canal, solve memory chip shortages, and prevent greedflation?

Lots of online discussion about the implications of AI solving the Navier-Stokes Millenium Problem. But spending $15,000,000 to get a $1,000,000 prize is just business as usual for OpenAI.

Blacksky was one of the 1st PDSs to allow open sign ups We are also now the largest PDS deployment that doesn't run any of bsky's code Sorry for the turbulence. It was the one-time price to pay for being able to build features like private accounts or sign-in via phone number w/o waiting on Bluesky

One way to slow down AI is to start strengthening and enforcing privacy laws. Even existing laws if interpreted wisely and enforced by regulators with a spine can make a difference.

But there is simply no reason that I can see that an “advanced AI” must behave in this way. Humans, for example, famously don’t. Things that act like perfect expected utility maximizers over an open world are actually quite rare, and it’s not clear that we know how to make them.

As both an academic and a writer, I am repelled by the "this looks like a long document. Would you like me to summarize it for you?" that seems to appear at the top of every pdf viewer now. NO. The *entire point* of good writing (either books or journal papers) is that every sentence has a purpose.

Apropos of the latest attacks on RNLI volunteers, I present my distant relative Bert who was in the RNLI. A true patriot and a very fetching chonky lad in his rescue gear

If you are an Uber driver, or know one who may be interested, information about the claim and a form to sign up and join can be found here: dynamicpay.org/

The overreaction to this Hugging Face thing is driving me nuts. Come flay me for being wrong but... 1/x

Mechanical Turk may be shutting down, but AI will always depend on data workers whose knowledge trains and tunes models. AI runs on extracted cultural skills and knowledge, and culture is always changing. Less the end of an era, but the end of an early prototype.

Do I have this right