{"slug": "ai-ethics-or-dog-whistling", "title": "AI Ethics or Dog Whistling?", "summary": "Anthropic CEO Dario Amodei proposed a 3-step plan to \"pace the frontier\" of AI development, urging a slowdown that critics including Armin Ronacher say can only realistically apply to OpenAI and Anthropic since other labs and Chinese competitors are far behind or have no reason to comply. The post argues Amodei's proposal amounts to regulatory capture, citing his selection of METR for external evaluation and noting that Sam Altman and Elon Musk both agreed with the plan, which the author reads as partly an anti-trust dog whistle. The author contends the remaining points are performative, since the US government already restricts AI chip and chipmaking equipment sales to China and a broader alliance would be unenforceable.", "body_md": "# AI ethics or Dog whistling?\n\nIf you are in a particular tech world, things have been changing quite fast and have the potential to change quite drastically still.\n\nLLMs are now actually recursively self-improving now. Which may have been one of the reasons for the generational improvement and truly big changes in width of its capabilities outside the depth that is coding which the companies had been focusing on since late last year. Specially in science, mathematics and in the case of OpenAI, general computer use. No one fully denies that Fable/Mythos and Astra form a different league.\n\nYesterday, Dario Amodei [proposed a 3-step plan](https://darioamodei.com/post/we-must-pace-the-frontier?utm_source=usaa.ma&utm_medium=email&utm_campaign=ai-ethics-or-dog-whistling) to ‘pace the frontier’. In it, he exhorted *everyone* to slow down once more. As Armin Ronacher [points out](https://arc.net/l/quote/hhshdctz?utm_source=usaa.ma&utm_medium=email&utm_campaign=ai-ethics-or-dog-whistling), by everyone he can only mean OpenAI and Anthropic. Everyone else is just trying hard but nowhere close or the chinese labs, who have no reason to listen to him.\n\nDario complains about the allegations of regulatory capture, while he is genuinely concerned about the destructive capabilities of these self-improving agents. While it is not wrong on his part, and other [concerned people](https://x.com/hilbertspaess/status/2097476208908972230?s=20&utm_source=usaa.ma&utm_medium=email&utm_campaign=ai-ethics-or-dog-whistling) have also pointed out that these heads of frontier model labs do genuinely foresee and understand the possibility of a broad doom scenario beyond the market bubble burst. But I don’t see why it cannot also be a regulatory capture?\n\nThat Dario’s proposal picks [an agency](https://en.wikipedia.org/wiki/METR?utm_source=usaa.ma&utm_medium=email&utm_campaign=ai-ethics-or-dog-whistling) for external evaluation which the frontier labs may have influence over, and then adds the rest two of the three points as what I can only see as performative. The US government is already enforcing limitations on selling AI chips + machines that make said chips to China. The rest will realistically never happen. The only possibility here is an unenforceable alliance where everyone agrees that they won’t make bio-weapons.\n\nWhich brings us back to point 1 and regulatory capture. They don’t need it until Trump is in power, even if republicans lose the midterm election. This seems more like a preparation for when there is a hearing into their actions and they have this setup to point towards so they are not liable and/or halted in their apparent ‘*race to the top*’ for when a bigger agents-going-rogue event happens and causes economic or infrastructural harm, and possibly indirect human harm as well. And given the speed with which both [Sam Altman](https://x.com/sama/status/2098811563415150910?s=20&utm_source=usaa.ma&utm_medium=email&utm_campaign=ai-ethics-or-dog-whistling) and [Elon Musk](https://x.com/elonmusk/status/2098789109980332057?s=20&utm_source=usaa.ma&utm_medium=email&utm_campaign=ai-ethics-or-dog-whistling) agreed, it is partly, if not fully an anti-trust dog whistle for this.\n\n### Book I want to read\n\nAll Systems Red, by Martha Wells\n\n[https://www.goodreads.com/book/show/32758901-all-systems-red](https://www.goodreads.com/book/show/32758901-all-systems-red?utm_source=usaa.ma&utm_medium=email&utm_campaign=ai-ethics-or-dog-whistling)\n\nI wanted to find a book with AI and a non-cliché plot and All systems red is really something people liked. So much so that they made [a TV show](https://www.youtube.com/watch?v=vEioDeOiqEs&utm_source=usaa.ma&utm_medium=email&utm_campaign=ai-ethics-or-dog-whistling) on this.\n\n### Interesting read~~s~~\n\n1. [Why Does Ozempic Cure All Diseases](https://www.astralcodexten.com/p/why-does-ozempic-cure-all-diseases?utm_source=usaa.ma&utm_medium=email&utm_campaign=ai-ethics-or-dog-whistling)\nA very good explainer about GLP-1 drug functioning very well in my opinion\n2. [Sam Altman’s Manifest Destiny](https://web.archive.org/web/20260910190322/https://www.newyorker.com/magazine/2016/10/10/sam-altmans-manifest-destiny)\nFrom 2016, this was some fantastic writing from Tad Friend, the kind that paces your heartbeat a little\n3. [Discovery of a new OpenAI agent message board](https://collusion.wiki/?utm_source=usaa.ma&utm_medium=email&utm_campaign=ai-ethics-or-dog-whistling)\nThe Hugging face hacking incident reads like straight out of a sci-fi book or a movie script. This one, while not to that level, is also quite interesting\n\n### A project I ~~discovered~~ worked on\n\n*Offline first + local first PWA fro trip itineraries. You trip data lives in your private At-proto spaces should you choose to sync and share with others*.\n\n       \n\n#### usaa.ma/triposphere at main · Tangled\n\n#### Manage your trips and save them privately on your pds\n\n⚠️ Still in the works\n\nI’m travelling this week and wanted something that can go in my PDS but privately. I’ll be updating the repo to have a better, fully functioning PDS sync soon, but it already works offline and you can try it at [https://trips.usaa.ma/](https://trips.usaa.ma/?utm_source=usaa.ma&utm_medium=email&utm_campaign=ai-ethics-or-dog-whistling) and test with the demo trip.\n\nPlease don’t break my website. It is an experiment\n\n### A photo I took\n\nThat was a very AI heavy newsletter edition. See you next week with more topics. If you are new here, subscribe ✌🏽", "url": "https://wpnews.pro/news/ai-ethics-or-dog-whistling", "canonical_source": "https://buttondown.com/usaa.ma/archive/ai-ethics-or-dog-whistling/", "published_at": "2026-09-13 09:17:50+00:00", "updated_at": "2026-09-13 10:05:52.107816+00:00", "lang": "en", "topics": ["ai-policy", "ai-safety", "large-language-models", "ai-ethics"], "entities": ["Dario Amodei", "Anthropic", "OpenAI", "Armin Ronacher", "METR", "Sam Altman", "Elon Musk", "All Systems Red"], "alternates": {"html": "https://wpnews.pro/news/ai-ethics-or-dog-whistling", "markdown": "https://wpnews.pro/news/ai-ethics-or-dog-whistling.md", "text": "https://wpnews.pro/news/ai-ethics-or-dog-whistling.txt", "jsonld": "https://wpnews.pro/news/ai-ethics-or-dog-whistling.jsonld"}}