Newsletter

AI is never human-like and other things

March 14, 2025

Hello,

Here’s everything since my last little missive to you:

New Articles:

AI is never human-like

Excerpt: People treat lots of stuff like they treat humans. AI is one of them. We talk about how human-like they are. How long until their ‘intelligence’ is like our intelligence. How long until they start doing human things, like murdering their competitors. Things like this. But AI isn’t even approaching human-like. In two very fundamental ways. And until those things change, they’ll continue to be completely incomprehensible to us.

Main idea: AI has human-like output, but a very different environment and different values for than environment, and until all three align, they will never actually be human-like.

New Marginalia:

A possibly good AI app? A Pocket/Reader/RSS-type app for saving and reading articles of interest has a new feature, co-reader.

Co-Reader anticipates questions as you read, offering instant answers and deeper context without interrupting your flow.

The magic lies in its effortlessness. Tap a paragraph to see questions. Tap a question to see answers. No typing. No switching apps. Just knowledge on tap.

I’m almost certainly not going to try it, but this is the kind of direction I like to see.

Link

Yale Review article asks ‘Was an accessory the secret to evolution?’ The value of the handbag was interesting. I myself have found myself adopting the cross-body bag, which is guess is the newest iteration of the man-bag. But the more general commentary of the utility of the feminisation of tasks is probably more interesting.

Link

People With Parents With Money. Interviewing the secret beneficiaries of inter-generational wealth in New York. It’s worth a skim, if only to get surprised by the sheer variety of forms it takes.

Link

On asking the right questions. This guy approaches the concept from the perspective of troubleshooting. What makes good troubleshooting? I approach very similar questions from a different angle. I’ve talked about the ‘language problem’, ‘naming problem’, and most relevant, the ‘question problem’ before (see also this). This is where whatever you’re trying to get after might be quite straightforward, but you’re suffering from either:

  1. The language you’re using is stopping people from understanding. You know this one is true because someone else will suggest the same thing with different words, and everyone will love their idea, but not yours.
  2. The fact that you’ve named something makes you feel like you’ve explained it. We have a ‘mind’, and maybe you think you understand the mind by naming it, but actually no one really knows how this thing works (or even if it exists).
  3. Simply asking the wrong question. You want to get people ‘to the next level’, but actually there aren’t levels, and so seeking the next ‘level’ ends up being a wild goose chase (this is something that actually happened to me).

Anyway. Since the language problem is the thing that bothers me the most, here is an article that uses different language to explain similar stuff. The language of troubleshooting.

Link

How The Mormons Conquered America. I talk a lot about how cults seem like some kind of human default. Not destructive cults, but groups based on shared values rather than oriented around shared experiences or interests. Fast growing groups like the Mormons should be interesting to us then, for what they represent for this little facet of the human experience. According to this article, fundamentally:

Mormonism is built on “a narrative structure rather than a philosophical belief system,” explains Kathleen Flake, professor of Mormon Studies at the University of Virginia. “That allows people to find meaning in it through narratives.” It allows them to see God’s impact on their own lives.

Mormonism, then, has adapted in part by adopting a structure that allows for local flexibility and requires local involvement, which encourages members to be more involved. This allows the religion to mesh with many different cultures.

Link

I hope you found something interesting.

You can find links to all my previous missives here.

Warm regards,

Dorian | btrmt.