
Reuben Binns @[email protected]
9.2K posts

Reuben Binns @[email protected]
@RDBinns
computery guy @CompSciOxford, @KelloggOx. HCI; ML; privacy; security; political econ of tech. [email protected] #THFC he/him [email protected]









💯🎯 Also NB. during covid, I had to tell the Cabinet Secretary to tell the ICO to effectively publicly suspend GDPR (which they did) under threat of getting the PM to say publicly 'ignore GDPR' because Whitehall left to itself literally prioritised 'legal advice on EU GDPR compliance' over lives and suffering and people all over the system said 'we can't do X unless you get officials to shut up about GDPR' - e.g there would have been *no shielding program* because Cabinet Office thought it was effectively legally impossible to connect the databases. As you shd expect this issue has been buried by Inquiry because doesn't fit narratives Whitehall want discussed. Similar insanity happens daily. MPs ignore it all



NEW: Amidst a frenzy around public AI investment in Europe, our new report takes us back to the drawing board. What kind of (digital) future does Europe want? What role can (should) AI play? Who should have a say? Read expert perspectives: ainowinstitute.org/redirecting-eu… 1/

NEW: The UK government is set to consult on a scheme that would allow AI companies to scrape content from publishers and artists unless they “opt out”, in a move that will rile the creative industry. W/ @DanielThomasLDN ft.com/content/26bc3d…



'LLMs can't really reason, they just pattern match' Wittgenstein: 'ah, but how do we know they aren't following their own private reasoning rules which are the same as ours, except when the reasoning problem has superficial differences to their training data?'

"A new paper from Apple draws attention to a persistent problem in LLMs: their reliance on pattern matching rather than genuine logical reasoning. As a result, adding irrelevant information—can lead to vastly different answers from the models." macrumors.com/2024/10/14/app…










