Andy Crouch

Extremely good interview between two very sharp people (@nwilliams030 & @ahc) on technology, the family, co-living, and much more! thirdoikos.com/p/life-in-the-…

@natenorberg I truly meant "unforeseeable" . . . who knows. I just expect that there will be a dramatic event that shifts the policy landscape. At least one fiction writer conceived of the Butlerian jihad!

@ahc Oh, so you're thinking much stronger than a total ban?

I expect (vs predict) that there will be a 9/11-level event (causing grievous harm to >thousands of people) that will change the direction of AI policy in ways that are genuinely unforeseeable right now. Biosecurity is a strong candidate for the vector, though there are others.

@natenorberg Very hard to say. But if GPUs became as difficult to obtain as, say, uranium centrifuges . . .

@ahc I’m curious if you think there would be AI policy that would be effective for these kinds of dangers, even with strong public support

@michaelhilton Yes, that's fair, but I suspect it won't be just a handful of brave astronauts harmed, and in that sense much more like 9/11.

@ahc Simon Willison (lennysnewsletter.com/p/an-ai-state-…) is predicting a "challenger like event" which I think is probably more likely, in that it will be people lulled into a false sense of security vs an overt action like 9/11

@DavidFish7 oof

I get an out-of-the blue email from someone wanting to connect. Seems like a nice guy but man, it sure reads like AI at least co-wrote it. In any case, these days I have to decline most such requests (to my frustration). I write back, personally, saying so as kindly as I can.

@dwchurchwell I use it a lot for research broadly, especially getting reasonable depth on new-to-me concepts (recently e.g., entrainment, literacy acquisition theory, neurophysiology of addiction) — just not for generating any words I would use verbatim in another context.

@ahc If not for writing/research...what other ways do you utilize AI?

@RobbieSap90 "asymmetric leverage" . . . VERY suspicious . . . :)

you're absolutely right! (honestly though I have begun to wonder how much I'm being trained on AI outputs these days)

@RobbieSap90 did an ai write this 😂

I think all of this is heading in the direction of asymmetric leverage for IRL networks and warm referrals, which is not without tradeoffs, but I think could be a healthy social incentive for people to live embodied life together again. (and now I ask myself if I sound like a bot)

@michaelhilton And people are also starting to (unconsciously) write like LLMs!!

@ahc I thought I was reading AI generated text until I remembered I was grading a hand written exam I saw a student take. When you see so much AI text everywhere, you start to see it everywhere, including places where it is not

I don't even use spell check, and never use AI in personal correspondence, or for that matter any writing that will see the light of day in any form (I use it for lots of other things). But soon enough we are just going to assume that everything is a bot. Sigh.

He accidentally replies to me instead of his business partner, speculating that my email may be an "AI-generated template email response." This dynamic is going to be the death of all remaining trust on the Internet.

The ISO 216 standard for paper sizes is so beautiful. That is all.

@DavidDecosimo @lukeburgis Oh, I agree. Though I should mention one other thing — I have never in my life imagined SV tech folks begging to be regulated the way that I have heard multiple times on recent visits (not just to Anthropic). It's wild. Not universal of course.

@ahc @lukeburgis Importantly, this is not ‘being cynical.’ Rather, it is attending to a basic structural reality: power minus accountability equals domination. They are free to heed what you said or completely ignore it, to invite who they choose or no one at all, etc. That’s the problem.

Just spent two days with senior staff at Anthropic and a group of Christians deeply committed to finding redemptive paths for AI. Encouraging on many fronts. Above all I think the guests were unanimously blown away by the humility and moral seriousness of Anthropic's leadership.

It is SO fascinating that one of the most popular podcasts ever, ACQUIRED, completely changed their focus, from their original title to stories of enduring, often multigenerational brands. Impact (Force / Time) is way overrated. Influence (force ^ time) is so much better.

Huh. Writing that last sentence I first typed, "if I cried," which actually was the heart of the matter in both cases.

And the most important grade I ever got was a C from my beloved 9th grade English teacher (RIP Ellen Nardiello), for an essay that I to this day think was not that bad, but that she knew was far beneath what I could have written if I tried.

"Some of my most important formative educational experiences involved some teacher or authority figure losing patience with me."—John Warner YES. The most significant piano lesson in my life ended early because my teacher was so frustrated with my heartless playing of Beethoven.

@DavidDecosimo That doesn't in any way make them beyond criticism and you are entitled to be as cynical as you like about that impression, but I do think something different is going on at that company compared to the other frontier labs.

@DavidDecosimo I also want to add this: I was truly surprised at the humility and openness to critique of the specific Anthropic team members we interacted with. I've never experienced anything like it in 30+ years of interacting with people of their level of acumen and social power.