artlu's Bear Blog

A Hopeless Rant

Two Ways to Grapple with AI Ethics

it's probably too late to pause

Chinese AI researchers worry less about advancing P(doom), inasmuch as their assessment of P(Americans will use AI to impoverish non-Americans) remains more salient

Chinese will not pause, even if Americans agree to pause

Since current AI capabilities allow for recursive self-improvement, Chinese will surpass any paused/throttled American lab

Therefore, Americans will not pause


wat do?

Which AI model iz 4 conscientious ppl?

If we listen to every bit of the tl, then OAI is evil, Anthropic is evil, Microsoft is evil, Facebook is Evil, Amazon is evil, Google is Evil, Apple is shite, Mistral is a cat, open weights models are CCP, etc.

Picking a model provider is kinda like, what kind of straw do we use, given concerns about global warming: plastic, paper, or steel?

Or, as in Taiwan, what kind of restaurant chopsticks do you use:

They also have concerns about local vs American pork, local vs imported rice, farm-raised vs ocean-caught fish, etc.

Most Brits and Europeans are shocked how little Americans recycle, and how imperfect is our trash sorting. Most Americans are shocked at European room temperature soda, and you-have-to-pay-for-water-and-public-restrooms. Don't get started on how the rest of the world sees queues/lines in the Middle East, or open defecation across the sub-continent1.

I think the choice of AI model is kind of on the same effectiveness level of using an AeroPress2. It lets you snark about Nespresso drinkers killing turtles and dolphins, and feel better than Chemex pour-over and French press users ("they don't know that...")



  1. TIL Phase 2 of Swachh Bharat Mission is scheduled to end in 2027

  2. with expensive metal filters rather than paper, natch