See, however, https://allpoetry.com/poem/8481013-Wordsworth-by-James-Kenneth-Stephen

Using AI is like using any other powerful tool—be it a table saw, a lathe, or a motorcycle. If you don't know how to use it properly, you can easily hurt yourself or someone else.

It is also important to understand its technical limits. Every AI model has a limited context window, a limited number of responses. After a certain number of exchanges—whether you are on a free plan or a paid subscription—the system can start to drift, hallucinate, or forget the original guardrails of the conversation. Kinda of like what happens here on the forum...lol

That said, there is absolutely no shame in using AI as a tool to help organize your thoughts, structure research, or polish your prose. Writers, academics, and musicians have always used tools to help articulate their ideas clearly. The key is knowing how to manage the tool, rather than letting it do the thinking for you. Anyone using an LLM has to know, in their own mind and heart, if the response is truly in their own voice.

I certainly don't and won't craft text with AI here (except perhaps for a parody of text, with the source medium clearly attributed; I've not done that yet that I can remember).

But I have no compunction to use it as an amplified search engine for *possible* documentation and sources, and if I find something that *might* prove helpful, I would share it but attribute the source mechanism (I think I've done that once so far, but I could be wrong).

I am not of the AI-is-demonic school that seems to have captured the imaginations of some (not here so much as more broadly). Myriad instances of fallen human nature - thus far - appear to fully explain the evils that it can midwife, and I am of the traditional school of Catholic praxis to never be glib and casual about things that should only be spoken of with greatest discretion.

One of my proverbs about AI: There is no I - nor Me/We/Us - in AI. That means it cannot authentically: love, hate, trust, distrust, hope, or despair*. It bears the impress (by act and omission) of those who (1) determine that and how it ought to be made, (2) its designers, (3) the source material chosen (and not chosen) by said designers, and (4) its prompters (aka "users" - a terrible word except for the ... addicted).

AI does not create authentically original thought. It's a tool, and people seized with cupidity and avarice will rush to profit from it, especially off of the credulous - thus working both ends of fallen human nature. That's true of so much of human artifacts. Collectively, they act as evidence of the Fall.

* By contrast, certain created beings remaining uncategorized here can authentically hate, distrust, and despair, and have perpetual allergies to created beings who can authentically love, trust, and hope.

I must vehemently disagree with the view that it is an acceptable tool for really anything having to do with research in the humanities. Speaking as a teacher, I can only say that it has had a devastating effect on the writing that students submit for music history classes. It's enough to make me want to quit every time I come across it. It has leveled out the style of student writing in a way I can only describe as off-putting, while also multiplying the amount of pure nonsense that gets turned in. In response to pretty much any actually interesting question about music history or theory, you are likely to get pure rubbish in response, at least as of 2026.

Honestly, even as a search tool, research tool, or organizational tool, I am deeply unimpressed by its capabilities. Perhaps this will improve. But I sort of doubt it. When people started relying on Google/Wikipedia, that was bad enough. But this signals the end of higher ed as we know it. I think there is no future in music pedagogy that is not an AI-free future.

Of course, this is a minority opinion!

@Charles_Weaver

What if a researcher wanted to find every instance of the littera significativa 's' in a particular digitized manuscript, as a basis for further research? Is this only a worthwhile research building block if he does it personally by hand?

Yes, Matthew, search engines have been broken for a long while.

Yep. I construct reasoned searches with a given set of factual assumptions (that evolve in the course of the process) in a way that mere Boolean-lite search engines now typically frustrate. It's like an intern substitute. One way to start doing it is to invite it to interview you, as it were, earlier rather than later in the process, but after the initial framework. I correct its biases (and there can be lots of them) as needed.

How I miss my relationship with my living, breathing best executive secretary (who had been a bank officer and then ran her own insurace agency earlier in her life) when I was an associate at a law firm nearly four decades ago.... As my father taught me, "secretary" was a profession, and it was my responsibility to keep her practised in her skills that she had labored to acquire. We were a great team; partners were envious.

Looking for a specific text match in digitised images is a good example of the kind of laborious work these machine-learning models can do more easily. Though "every" is a high bar, and you'd still have to read it by eye to check if everything it found is indeed an 's' or whether it missed any because of resolution or contrast problems. While such tasks might be done locally by an application with no network or data centre demand, the easy availability of a general purpose tool means that people are less likely to obtain and learn specific tools. You can install a free or cheap shareware PDF merging tool or just upload the PDFs to a website and not care that you've just given your data to an AI training slurping machine.

I'm so worried about the brain-drain aspect of AI.

Years ago, I read something to the effect of: before the moveable-type printing press, when books were relatively scarce and the population was largely illiterate, people relied on their memories. You had to remember facts, processes, everything you needed for everyday life. You learned by having a connection to another person, a teacher, a mentor, a parent, a neighbour. The social process of transmitting and creating ideas was foundational to how we remembered things and transmitted knowledge.

Once the printing press was invented and books became plentiful and cheap (and literacy increased substantially), we began to offload a lot of those memory tasks. In many ways it was good, because we could write down much more than we are capable of remembering as individuals. This let us have great leaps and bounds in the development of science and technology, for example. It made transmitting of ideas easier over long distances and long periods of time. We didn't need to be connected to our teachers by time or space anymore. The way we used our memory shifted from remembering things to remembering where to find things. It also removed some of the social process of learning, because instead of learning from a person directly, the teacher functions more as a facilitator, helping us navigate the information conserved in print from many other people.

That's where what I read ended. This would have been a good 20 years ago, just as Google was emerging as an everyday tool. We can continue this thought and extend it:

Once things like search engines came into the picture, the way we use our memories changed again. Now it's not even remembering where to find things, but instead we learn how to prompt the engine. Instead of knowing what book we need to look in, we ask a machine to find the book for us. Human interaction is pretty much removed at this point. With search engines (without LLM), you still need to go and read the material, synthesize and collate it, and generally understand it. You have to know what information to retain and what to discard, which requires critical thought. You know what something isn't as much as you know something is. You also need to remember what question you asked the engine to find the material again.

But with LLM, you don't even have to really know how to prompt the engine, or have an understanding of the material. The reading, synthesis and collation already has happened. We don't fully have to understand something to give an answer. With just simple search engines, we have long known there is a bias problem; with LLM that problem tends to increase exponentially. The need to critically think about material, already damaged by things like Google searches, is further eroded.

I think sometimes about how much less I know now, even though I'm definitely older and supposedly wiser. I have an exceptional memory due to my neurodivergence, and while age plays a factor in declining memory, I couldn't even tell you my kids' phone numbers. But it's not even just remembering details like that; it's understanding things.

I'm a young GenXer, and like a lot of you here I've lived the last 3 stages (books to search engines to LLM). Whereas 20-25 years ago, if I wanted to know something I would go to the library and get a 200 page book on the subject. If I needed a recipe, I called my grandmother. I could maybe read a couple more books on a topic. I could understand a topic of interest to me. We were kind of all in that same boat.

Nowdays, an AI summary on Google is good enough a lot of the time. I know it has impacted by ability to critically think, to write, to have confidence in what I'm doing (because I haven't had the chance to internalize something). I can see and feel the difference. I can see the difference in my oldest children (who are just now reaching adulthood and who were homeschooled in their earliest years of school) and my youngest (who always went to school and are still in primary school). How they process and synthesize information is totally different.

My two cents.

PS: Church work, as it were, appears to me to be a challenging terrain on which to deploy LLMs. Why? Constraints. What kind? Well, let's start with....

1. Power structures that incentivize inertia and discincentivize healthy change, and thereby discourage curiosity.

2. Internalized biases that don't necessarily get laid out clearly in prompting because they are like oxygen in the imagination.

3. Passive aggressive niceness culture that can also play an inhibiting role.

Before discovering Dune I thought that it had to do with Judith Butler and I was confused.

In my experience, AI rewards expertise far more than it replaces it.

B I N G O