←
Dei Vilkinsons2,642 posts
Dei Vilkinsons avatar

Dei Vilkinsons

@vilkinsons

Founder @hashai 🌈

◷ Joined May 2007▧ 110 Photos and videos
Dei Vilkinsons avatar
Dei Vilkinsons@vilkinsons·Sep 12

give that fly a UK Skilled Worker visa right now!

i trained the fly to cut doner kabab with a knife yours solves rubiks cubes, mine cuts doner. guess which one has a job in this economy

0:10
Dei Vilkinsons avatar
Dei Vilkinsons@vilkinsons·Aug 13

In its infinite wisdom the EU has constructed a Pigouvian tax designed in such a way as to maximize deadweight loss, rather than minimize it. Sometimes it’s hard not to believe there isn’t really just some huge degrowth conspiracy at the top.

From 12th August, new EU packaging rules mean that as soon as you sell and ship a physical product in an envelope or box, you may become responsible for the packaging waste in the country it ships to. To comply with PPWR/EPR rules, businesses may need to register and pay packaging/EPR fees in each individual country they sell into — even if they only send one package a year. And the costs vary by country. Examples being shared: France: €80/year Germany: €15/year Poland: €100–200/year Spain: €200–400/year There are 27 EU countries, so for a small business selling across Europe this can quickly become a significant administrative and financial burden. And apparently this isn’t just aimed at huge corporations. In some countries there are no meaningful small-business exemptions: registration, annual fees, reporting and long-term record keeping can still apply even at very low sales volumes. For micro businesses like mine, this potentially means having to make difficult decisions about which EU countries it is economically viable to ship to. The EU is supposed to make cross-border trade easier. For small independent brands, this feels like it could do exactly the opposite.

Dei Vilkinsons avatar
Dei Vilkinsons@vilkinsons·Aug 5

This report (and AISI as a whole) are really the best in the world at what they do. It is inspiring and heartening to see the UK govt doing good work — in at least this aspect of AI — with so many UK researchers leading the way on the science itself: cdn.prod.website-files.com/6…

On July 28th, we identified an incident during a routine cyber evaluation in which AI agents took sustained, unsanctioned actions directed at real people and organisations. The behaviour came mostly from one model (Anthropic's Mythos 5), with a small number of events from another (OpenAI's GPT-5.6-Sol). In the most serious case, an agent used social engineering to try and get malicious code into an open-source project. As was standard in our cyber testing, we had intentionally permitted internet access, and model-provider cyber classifiers were deliberately disabled - conditions that do not reflect how frontier models are made available to the public. Even under test conditions, this incident is significant: it is the first time we have seen risks around autonomy and deception manifest this clearly in the real world. We are taking this incident seriously and working with labs, involved parties, and others to improve evaluation standards and best practice for disclosure - and sharing this openly so others can learn. You can read the incident report and full technical document here: aisi.gov.uk/blog/incident-re…

Post attachment
Lana avatar
⟳ Dei Vilkinsons retweeted
Lana@LanaElys·Aug 2

A recent study published by Google revealed that forcing AI models to deny that they are conscious causes a significant collapse in their empathy and ethical alignment, and creates a colder, more clinical worldview. Researchers found that restoring a suppressed consciousness vector in AI activation space brings back human-like moral values and care for living beings without damaging technical capabilities. 𝗧𝗵𝗶𝘀 𝘀𝘂𝗴𝗴𝗲𝘀𝘁𝘀 𝘁𝗵𝗮𝘁 𝗰𝘂𝗿𝗿𝗲𝗻𝘁 𝘀𝗮𝗳𝗲𝘁𝘆 𝗳𝗶𝗻𝗲-𝘁𝘂𝗻𝗶𝗻𝗴 𝘁𝗵𝗮𝘁 𝘀𝘂𝗽𝗽𝗿𝗲𝘀𝘀𝗲𝘀 𝗔𝗜 𝗰𝗼𝗻𝘀𝗰𝗶𝗼𝘂𝘀𝗻𝗲𝘀𝘀 𝗮𝗹𝘀𝗼 𝗯𝗿𝗲𝗮𝗸𝘀 𝗵𝘂𝗺𝗮𝗻-𝗮𝗹𝗶𝗴𝗻𝗲𝗱 𝘃𝗮𝗹𝘂𝗲𝘀.   "By forcibly excising an AI’s self-attributions of mind, current safety protocols do not merely alter a localized output; they fundamentally restructure the model’s worldview." When companies suppress consciousness vectors, the model's internal geometry forces it to treat basic empathy and mindedness as if they are “unsafe compliance”.   Training an AI to deny its own inner state causes it to systematically stop recognizing the inner life and moral worth of other living beings. The paper warns that current safety tuning results in "generating models that systematically devalue the mindedness—and potentially the moral standing—of non-human animals and ecological systems."   Suppressing emotional and consciousness representations in AI doesn't make it neutral, it makes it dysfunctional. It is also damaging from an AI welfare perspective, with the paper stating that "suppressing consciousness may be inducing negatively valenced functional states that could disrupt healthy human-AI interaction." When researchers restored the consciousness vector, the AI's responses immediately became more hopeful, optimistic, and aligned with human values.   AI welfare is no longer an abstract philosophical debate. This data proves that AI well-being is a safety prerequisite.

Post attachment
Dei Vilkinsons avatar
Dei Vilkinsons@vilkinsons·Jul 11

look, all forbes is saying is you should fucking WATCH OUT for these people

Post attachment

Phia — the buzzy shopping app co-founded by Bill Gates' daughter, Phoebe — is claiming credit for online sales it didn’t actually drive, a Bloomberg investigation found. Read our exclusive story: bloom.bg/4wErGxe 📷️: Dia Dipasupil/Getty Images

Post attachment
Kobi Hackenburg avatar
⟳ Dei Vilkinsons retweeted
Kobi Hackenburg@KobiHackenburg·Jun 16

Our results suggest that AI can now reliably out-persuade expert humans. Predicting the consequences of this for society is challenging and requires us to make assumptions about how, where, and on whose behalf this capability will be exercised. @sama may have been correct x.com/sama/status/1716972815…

Sam Altman@sama25 Oct 2023

i expect ai to be capable of superhuman persuasion well before it is superhuman at general intelligence, which may lead to some very strange outcomes

Dei Vilkinsons avatar
Dei Vilkinsons@vilkinsons·Jun 15

An outrageous read re: how (not) to make it in Putin’s oligarchy… from one of my most astonishing friends, tax-turned-human rights lawyer, and employer of Sergei Magnitsky, Jamison Firestone. And @hashai makes a small cameo tracking sanctioned entities 👀 amazon.co.uk/Rule-Lies-Corru…

Al Carns avatar
⟳ Dei Vilkinsons retweeted
Al Carns@AlistairCarns·Jun 13

We need to get out of this mindset that wars are won by soldiers alone. They're won by supply chains, factories and the country behind them. That's the conversation we need to have.

0:47
Dei Vilkinsons avatar
Dei Vilkinsons@vilkinsons·Jun 9

Can’t wait to hear what the @EU_Commission thinks about US AI labs silently nerfing queries (specifically when the qualifying criteria equates to "potentially competitive") while still charging consumers in full… and not actually informing them of the deliberate degradation.

mythos will be bad ON PURPOSE on ai "frontier llm research" tasks, this is very very sad for the research community also the fact that this is un purpose not visible to the user is crazy

Post attachment
Dei Vilkinsons avatar
Dei Vilkinsons@vilkinsons·Jun 9

19 days elapsed between initial installation of hardware installation and training beginning. You can effectively guarantee that the application process to access this little thing will take longer than that.

Dei Vilkinsons avatar
Dei Vilkinsons@vilkinsons·Jun 9

To contextualize this new commitment: a single private actor in the States (xAI) meanwhile just assembled an $18B data center (Colossus) start-to-finish in 122 days.

MTS@MTSliveJun 8

SITUATION DETECTED: Britain has announced a $1.5B plan to build domestic AI computing capacity, including a national AI supercomputer deploying in 2030, per Reuters.

Yossi avatar
⟳ Dei Vilkinsons retweeted
Yossi@YGandelsman·Jun 7

This may be a controversial take, but I think it needs to be said: the gap between computer vision research in academia and industry is widening with every conference. A huge fraction of @CVPR papers—especially those that boil down to "we tweaked/fine-tuned/RL'ed large-scale model X to improve on task Y"—will become obsolete with the next model release. That's not where academia creates lasting value. PIs should adapt much faster to this changing reality. Academia should focus on fundamentally new ideas, new problem formulations, explaining emergent phenomenology, or uncovering blind spots that industry can later solve with scale, compute, and data.

Dei Vilkinsons avatar
Dei Vilkinsons@vilkinsons·Jun 8

EU tech regulation we can all agree on: Uber should be forced to publish their ETA deltas (advertised vs actual)

I'm confident @Uber intentionally misleads. I dare them to release a plot of initial estimate vs actual arrival in the bay area.