What's this? Two months in a row? Two posts, two months in a row?
Kind of a spur-of-the-moment thing. What spurred it is, I dumped this whole blog into the Claude LLM and asked it to summarize the endeavor - who is this person? why are they doing this? and what exactly is "this"? This blog, I mean. 21 years of this.
Anyways, as we all know, the results of such an exercise is that the LLM gives a concise and readable overview, tries to make some kind of synthetic point (I did ask it to synthesize something), recites some pablum about what an interesting character is this Andrew (I did not tell it that, o ho, I am Andrew!).
Anyways - the spur-of-the-moment topic for this post is AI, and what's it all about, and what am I doing with it here in May of 2026?
Answers in reverse order:
3. What am I doing with it:
Shamed to admit that I use Claude/chatGPT/deepseek pretty regularly. In fact, I did a similar exercise with Claude itself the other day (before the AMHAZ exercise): I had it dump all our conversations, and then feed them back to it to ask it for an overview - what do I do with Claude? I can't quite remember, but it said similar kinds of things to what it said about the blog, and they were generally accurate: I probe it about statistical analyses (been working on a paper), get feedback on connections between my resume and a job ad, or ask about throughlines over my music listening habits.. sometimes a query about some unfamiliar research literature, sometimes an exchange about how some weird idea might play out. chatGPT has faded a bit, while I have come to use deepseek (the popular Chinese AI) to ask about China-related stuff. It's amazing how much more depthful its responses are than the "American" AIs, telling you just how little depth there is in the reams of text that the AAIs are trained on.
Anyways, digression. Claude pointed out what I already knew, which is that I'm fascinated by structure and mechanism, and I'm usually interested in ways to discover some underlying structure, or express it or extrapolate it or etc. I do regularly use it for those kinds of things - "here's a bunch of stuff - don't just summarize, but suppose there's are some underlying rules, what might they be?" I recognize that its answers aren't based in deep analysis but they are often interesting pointers into how to think about complicated things.
So yeah, that's what I do with it. It's also totally replaced stackexchange etc for looking up computer help, coding tips - I still write my own code, I live by the Feynman dictum about knowing and creating, but I do use it the way I used to scour old forum pages for discussion of how to resolve some problem (I realize that the LLM has already done the scouring).
2. What is it all about:
Brave new world? It's certainly about making some things easier and some things harder.
Easier: writing cover letters! Analyzing this stack of hours of patient interviews! Figuring out how to do a Bayesian analysis of a dataset, an analysis I neither want nor know how to do, but need to at least check in with, just to see... instead of spending weeks learning a bunch of stuff I don't want to know, I get a quick tutorial and some demo code and I can do my thing and move on. Nice, time saved!
Harder: clearly not a great time to be moving out of academia and into "data science". Laugh out loud. Also, it's very very weird not knowing whether any text you read is actually created by a human being, or is the output of a text mill. I'm still solidly on the side of Team Human. I want to read text by a human, hear music by a human. I worry that a generation of AI music is coming, AI art, AI games. I worry for the sake of my kids.
Brave new world!
1. AI
Is it, though? Is it really?
I swear to you that every word of this gripping installment of AMHAZ journal has been written by me, a living, breathing, farm-raised (not really) human being. An LLM would never say such a thing.
No comments:
Post a Comment