One of the frustrations of being a writer — except in the legacy push media, where apparently you can just make stuff up — is research. Search engines like Google help a lot (I can barely remember how to use a card catalog, my God, how did people live like that?), but you still end up with a million bookmarks, and it's nearly impossible to find bookmarks again months or years later.
If you're interested in such things, the buzzword is that you need a "second brain." There are a number of methods for this: the phrase "second brain" appears to have originated by Tiago Forte, who devised a note-taking method called "PARA."
Another approach was invented by a sociologist named Niklas Luhmann. His approach is called "Zettelkasten," which, being German, sounds much more impressive than "slip box" or "card file," which is what it means. (That's actually a link to Grok's description of Zettelkasten.) Zettelkasten sounds, as described, like a really great idea. All it requires is writing down every thought, every book, every article on an index card with a rather complicated indexing scheme.
It worked great for Luhmann. He was one of the most prolific authors in his field. He said it was because when he'd accumulated enough cards, he just collected the ones that applied, sorted them, and typed a draft.
This came up for me this week as I'm working on another article or series of articles on the sorry state of American education and how to fix it. (Spoiler: it sucks, it's really expensive, and I'm going to suggest radical changes.)
To do that, I needed to look at my previous school pieces — no problem; I have an Authory account that captures everything I write and publish. But I also need to keep track of a lot of other references that I know but don't have the exact form or remember but don't recall the details.
In Olden Times — like about two weeks ago — I'd have had to follow my sometimes vague recollections through the wilds of Google and Google Scholar based on what I remembered in order to find what I'm thinking of again. It can take a long time, depending on the vagaries of my memory.
This can take a long time. For example, I was trying to recall the paper I read years ago on how mentoring combined with independent study can improve learning dramatically.
The problem was that I remembered it wrong. I remembered the paper as being Boole's Two Sigmas paradox.
The idea itself is a very striking one — it turns out that a combination of mentoring and teaching to mastery rather than marching like soldiers through a fixed syllabus can result in achievement two standard deviations above students taught using traditional methods. (If it's been a while since stats class, that means that C students would have test scores at about 98% of the averages for traditional students. In other words, all the students would be like A students in a traditional classroom setting.)
But I couldn't find the paper!
So I went to Grok, described the paper, and in a minute or so, I was presented with the original paper and a summary of the results. (Spoiler: it's Bloom's Two Sigmas Paradox.)
I was using Grok to stand in for my fuzzy memory, and it did the job in a couple of minutes.
It's no surprise that an AI like Grok or Claude or ChatGPT can be used to find information, but there have been a lot of instances where the AI needs a reference, doesn't have one, and so makes one up.
Technically, this is called "hallucinating," one of those bits of terminology that I rather detest. There's nobody in there to hallucinate; it's honestly just doing a great big multiplication.
Call it hallucinating or not; it's a real risk with a large language model like ChatGPT or even Grok. It's like having a quick but stupid librarian who grabs something that looks good, and if it doesn't see anything, it makes something up.
There have been a number of instances of lawyers asking ChatGPT to write a brief without checking the references.
Judges don't take it well when they read your brief and they or their law clerks try to follow a reference to an interesting case only to find it's a fake citation.
It's even better when you are a "misinformation expert" whose paper on the hazards of misinformation includes sources that an AI had hallucinated and you don't notice.
Still, if you understand the limitations, an AI like Grok can do a lot to make your work easier and better.
What I do is use Grok as a dumb but efficient memory — a "second brain" that remembers everything but thinks about nothing. If I ask it for information, I'm often asking it for something where I know a fair bit about the topic already but want to gather up details and citations or ask for criticisms of what I remember. If I ask it about something I basically don't know much about, I don't trust it — but Grok is very good about pointing to sources.
Grok is doing a good job right now, but it can get better. If you use it as an aide to thinking instead of a replacement for thinking, it can be very handy.
