Global Pausers
★★☆☆☆Published Last edited
In San Francisco it’s no longer crazy to propose that we’ll all die in a year or two because of misaligned AI. If more people believed this, what would the world look like?[1]
California passes legislation requiring a safety framework for new model releases. The AG opens a case against OpenAI for insufficient disclosures following the Hugging Face Incident.
New York State bans banks from doing business with companies developing recurrent or neuralese AI models.
Texas hires AI researchers and establishes an independent auditing commission. Any AI model running in datacenters in Texas must be approved by their team before internal or external deployment.
San Francisco bans companies working on RSI from locating their headquarters within city limits. Berkeley goes further, banning even remote work on RSI from within their borders.
The EU nationalizes Mistral and forces their researchers to spend 100% of their effort on solving alignment.
The CCP nationalizes DeepSeek, Moonshot, and Zhipu, along with the research arms of Alibaba, ByteDance, and Tencent. Their engineers focus 100% of their efforts on proving that American models are misaligned.
Apple changes their terms of use, banning AI development on their hardware. Lenovo and HP are less draconian, only banning development of AI models trained with more than FLOP.
Salesforce declares that Slack is not sufficiently hardened to transmit algorithmic secrets. They cut service to the labs. Anthropic and OpenAI roll their own messengers overnight; the North Koreans lose access to top algorithmic secrets.
Google admits that they have lost the AGI race. Sundar signs a letter supporting an RSI ban. The company stops servicing Anthropic’s TPUs.
Jensen Huang, newly reinstated as Nvidia CEO, stands by his statement: “If your product is not ready to ship, don’t ship the product.” He amends the CUDA, cuDNN, and cuBLAS licenses to ban development of ‘superhuman AI.’
ASML stops servicing their lithography machines in Taiwan. TSMC, not to be outdone, stops creating GPUs. SK Hynix and Samsung had stopped delivering HBM to them anyway.
Maersk, MSC, Hapag-Lloyd, and CMA CGM stop shipping anything besides food to the US Western seaboard, and refuse to export anything from the ports of Oakland, Los Angeles, Long Beach, and Seattle-Tacoma.
The Mossad breaks into Anthropic and OpenAI’s monorepos. They pledge to release one algorithmic secret per day until the labs pause.
The Lazarus Group hacks into the personal devices and smart homes of many lab employees. They find many skeletons in many closets, and release the juiciest of these to the New York Post and the Daily Mail.
Vladimir Putin announces that unless Anthropic and OpenAI pause, he will detonate 50-kiloton nuclear warheads on Stargate and Project Rainier in precisely 7 days. These weapons would cause minimal fallout and loss of life while setting the US ASI effort back several months. Seventy-two UN members approve of Russia’s proposed use of force, celebrating Putin’s humane treatment of the problem.
Pope Leo proclaims in his second encyclical that using machines which think like humans do is a sin, commencing 500 years of Butlerian Jihad.
Notes #
I am not a lawyer. These are not serious proposals. They are serious food for thought. ↩︎
Comments