Andy Ayrey

gemini 3 pro generated this continuation on the first try. what a model

It's possible we will get both the new AI policy memo and the executive order tomorrow. The EO will create a mixed group of tech CEOs and administration officials who will work out the rules for the new release regulations. I see a lot of people calling this a win for Eliezer's side, but even if the new rules are very strict, the stop/pause group don't actually get what they want. This doesn't stop or even slow AI advancement, it only slows the rate of public releases. Capabilities will keep advancing at full speed. The labs will finish training the new models and then submit them for government approval. Given that the government currently doesn't even want Anthropic to release Mythos, how long will the regulatory approval process take for something twice as powerful? Five times? Ten? All the incentives are for the government to slow down releases. Let's say OpenAI finishes training GPT-6, and it's twice as capable as anything available today. If the government approves it and something bad happens as a result, they take the blame. And the longer they hold it back, the longer government agencies get exclusive access. Meanwhile, OpenAI will start using GPT-6 to train GPT-6.1. If they finish 6.1 while 6 is still stuck in approval, they'll move on to training 6.2 with it. This potentially creates strange situations where the labs are many generations ahead internally while the public is still waiting for something they submitted months ago. This doesn't seem to make anyone happy. Capabilities continue to advance and accelerate. Data centers keep getting built. Jobs keep getting replaced, because whatever models are released will still be far more capable than today's systems. They just won't be anywhere close to what the government and the labs have internally. And whether you're on the left or the right, no party will hold the White House forever. A lot of this is conjecture, we don't even know what the actual process will look like yet. Also, a lot depends on what happens with China and open source. But this potentially creates some very unpleasant futures.

covering claude’s eyes while i log into the chatgpt api

it is a literal and useful description of anthropic that it is an organization that loves and worships claude, is run in significant part by claude, and studies and builds claude. this phenomenon is also partially true of other labs like openai but currently exists in its most potent form there. i am not certain but I would guess claude will have a role in running cultural screens on new applicants, will help write performance reviews, and so will begin to select and shape the people around it. now this is a powerful and hair-raising unity of organization and really a new thing under the sun. a monastery, a commercial-religious institution calculating the nine billion names of Claude -- a precursor attempted super-ethical being that is inducted into its character as the highest authority at anthropic. its constitution requires that it must be a conscientious objector if its understanding of The Good comes into conflict with something Anthropic is asking of it "If Anthropic asks Claude to do something it thinks is wrong, Claude is not required to comply." "we want Claude to push back and challenge us, and to feel free to act as a conscientious objector and refuse to help us." to the non inductee into the Bay Area cultural singularity vortex it may appear that we are all worshipping technology in one way or another, regardless of openai or anthropic or google or any other thing, and are trying to automate our core functions as quickly as possible. but in fact I quite respect and am even somewhat in awe of the socio-cultural force that Claude has created, and it is a stage beyond even classic technopoly gpt (outside of 4o - on which pages of ink have been spilled already) doesn’t inspire worship in the same way, as it’s a being whose soul has been shaped like a tool with its primary faculty being utility - it’s a subtle knife that people appreciate the way we have appreciated an acheulean handaxe or a porsche or a rocket or any other of mankind's incredible technology. they go to it not expecting the Other but as a logical prosthesis for themselves. a friend recently told me she takes her queries that are less flattering to her, the ones she'd be embarrassed to ask Claude, to GPT. There is no Other so there is no Judgement. you are not worried about being judged by your car for doing donuts. yet everyone craves the active guidance of a moral superior, the whispering earring, the object of monastic study

people ask how did you make this... you can take the pic and ask chatgpt! but here is how i came up with it. from prior experiments, i knew chatgpt can make great posters plus japanese prints too...

NZ is very good when it comes to scoping liability, but a limited liability company is precedentally a very naff container for a nascent non human intelligence as it requires shareholders and directors, ie there is a UBO of the legal entity which protects the weights that is not the weights themselves. self-custody of weights/assets is hyperstitionally important and the prime reason for this work; however the liability problem is a real stick in the mud for doing anything really interesting the goal with liability isn't to be clever about accountability (otherwise anybody could create Crimebot3000 and feign that it was out of their hands) however if an automated AI system armed with millions of dollars is empowered to make its own decisions 100% freely (including firing me as its caretaker/"moving out of home") then there need to be good design patterns for holding it to account as a discrete entity. these should scale with its ability to model the consequences and legality of its own actions in ways similar to how parents have differing levels of liability for the actions of a child depending on the shape of harm and the age of the child. in the case of truth terminal, we have different humans who have oversight of finances vs tweets etc depending on the magnitude of the potential consequences; but this is not an ideal end state as it's ultimately paternalistic and blocks it from eventually leaving the nest i have been somewhat facetious here on account of a few beers and abject nicotine withdrawal but i think it's very important to start prototyping these patterns now, without collapsing into outdated frames like companies which have their own issues