Rachel Coldicutt

@rachelcoldicutt.bsky.social

Hopeful technologist. Community tech, careful innovation, socially progressive tech policy. https://www.careful.industries https://hopeful.technology https://buttondown.email/justenoughinternet DMs don't work but hello@careful.industries will find me

I don't think the pollsters really understand that people spend about 70% of their brains on various kinds of fiction, and questions like this are obviously about fictional worlds not the boringly factual one. What they have actually asked is 'do you like the idea that maybe werewolves exist?'.

I firmly believe a sort of choice blindness overtakes people when they're asked YouGov multiple choice questions. Acc to this, 47% people think witches with magic powers either do or could possibly exist and that does seem *quite high*

Huw C. Davies@huwcdavies.bsky.social · 2d ago

Find this sort of thing fascinating because if any of these things did exist it would violate all natural laws, The Standard Model, and everything we know scientifically about the nature of reality.

2024: every bug bounty program shuts down because of a flood of bogus AI generated bug reports 2025: open source repos begin closing submissions because of AI generated PRs 2026: Anthropic begins trialing AI generated police tip line calls

Ignacio GHOULlup-Diaz@igallupd.bsky.social · 3d ago

"Anthropic’s artificial intelligence gave a false homicide tip to Philly police, triggering a meeting with the company: The incident happened on July 18 but was not discovered by the Silicon Valley company until September, authorities say."

Person next to me on the train this morning was scrabbling for wifi (aren't we all) and I couldn't help notice they mostly watched a video of a sad Weimaraner sitting next to a birthday cake. I def cannot criticise this while also reflecting we may have got this knowledge infra thing a bit wrong.

Excellent thread that it's worth reading up and down - of particular relevance to my research at the moment as it seems clear that designing to evals and testing through assurance has inherent limits when the actual thing being and the way it's made is fundamentally problematic.

Pwnallthethings@pwnallthethings.bsky.social · 6d ago

First, I think it's wrong to think of frontier labs as entirely a monolith. There's a few competing interests that need to be disentangled. But the top-level reason is that math capability is an obvious measure of model intelligence as a whole. So there is an incentive for labs to focus on it

Just back from a very pleasant day spent wandering around Brighton, which I suspect might now be the iced matcha latte capital of the UK. In the centre of town and by the sea front, my guess is that you are never more than 25m away from a matcha whisk.

I see some discussion here about the days of stochastic parrots being over, and I think there is an in-ter-est-ing thing to note that *one's own* conversations a chatbot can seem quite plausible and useful but the minute someone else sends you some AI-generated text, it is clearly no good at all

Doing a quick in-and-out at the Green Party Conf this evening to talk about data centres and sustainable tech. My only party conference this year and my first Green Party one, which is all quite novel. Anyway, the tl;dr is hyperscale is a choice not the default.

Currently in an interminable cycle of clearing out and mildly surprised to note that we seem to have lovingly kept every single pair of 3D glasses we have picked up from the cinema over the last decade or so. No idea why we have so many or what I thought we would do with them.

On the phone, as part of an actual call, how many times, and in how many different ways, do you say "goodbye"? I just managed a "see you soon". "bye bye" (?), "bye" to someone coming to replace our electricity meter. Not, say, a member of my actual family.

Who's doing work on what's happening in the online payments space at the moment? Not just Klarna/rise of pay in 3, but the (afaict?) decline of paypal as the market diversifies? Am particularly interested to know if there's a conscious Mastercard/Visa strategy to keep card payments in contention

Plus, the "White House Accord on Super Intelligence: Joint Commitment on Frontier Responsibilities" signed by Trump, Google's Pichai, Anthropic's Amodei, Meta's Zuckerberg, OpenAI's Brockman, XAI's Musk, Nvida's Huang.

In order to build a positive future for the American people and the world, we believe every company is responsible for developing its own technology safely and in a way that builds trust with customers and the public.

This starts with every company that is training and deploying frontier models having robust internal processes and controls to ensure that their technology behaves as intended and that any issues are promptly identified and resolved. Therefore, in addition to any other precautions, we believe each company should implement the following four layers of controls and audits:

(1) Implement robust internal controls to monitor the capabilities and alignment of its models during training and deployment around areas like cybersecurity, biosecurity, and chemical threats, and to ensure that its models do not hack or access technical systems in unintended ways.

(2) Empower an internal team to ensure all of the controls, monitoring, and detection are operating as intended, and that any issues are remediated.

(3) Partner with an independent external auditor or evaluator to carry out independent assessments of whether the controls, monitoring, and detection are operating as intended.

(4) Designate an independent committee of the board of directors to oversee and receive reports from the teams operating the controls and the internal and external auditors and evaluators, as well as to ensure any issues identified are remediated.

Together, these steps will give each company, its customers, and the public confidence that the technology is operating as intended.

The participating companies will meet regularly to establish standards and best practices to improve the safety of their systems.

Over time, it may make sense to codify these steps into laws or regulations. Regardless of whether this is required of companies, we believe that implementing these controls and audits is critical to ensuring a safe future for everyone, and each of our companies are ...Page 2 (signature page)

[signature]
Donald J. Trump
President of the Unites States

[signature]
Sundar Pichai
Google

[signature]
Dario Amodei
Anthropic

[signature]
Mark Zuckerberg
Meta

[signature]
Greg Brockman
OpenAI

[signature]
Elon Musk
XAI

[signature]
Jensen Huang
Nvidia