AI & the environment

Blegging for some knowledge

In unpacking this new (but very old!) form of education, one of the things we need to do is get clear on how AI can help with it — if at all!

Imaginary Interlocutor: Did Kieran Egan write a lot about AI?

His books were written before LLMs came out, so no, not at all.

I.I.: Then why are we talking about this?

Egan pointed out that the ecstasy of learning comes when the deep meshy muck of information is brought out into the classroom, and begin to congeal in new ways inside people’s heads.

Usually, we talk about this in terms of how the information is structured. We say we need to return it from the academic-speak of PHILOSOPHIC (👩‍🔬) understanding back into the humans-everywhere speak of MYTHIC (🧙‍♂️) and ROMANTIC (🦹‍♂️) understanding. And this is true! When we’re able to do this, we revivify, revitalize, rehumanize, and re-enchant the content being learned.¹

But the other side of this is that you want to have a LOT of information. In order for students to be infected with the love of physiology, or computer science, or math, or poetry, or whatever, they need to have gained a lot of knowledge about it.

Egan put it thusly:

The more you know about something, the more interesting it becomes... The person without the intellectual resources deep knowledge can provide is much more likely to be bored.

– Kieran Egan, Learning in Depth: A Simple Innovation That Can Transform Schooling

So in order to make it matter to themselves (and then to their students), teachers need to become mini-experts on what they’re teaching. Alas, contemporary teacher training is, let’s say, not so good at helping new teachers develop the deep intellectual resources about their subjects.

In the past, this has been a major limitation of how Egan education can spread.

But now we have LLMs — the large language models like ChatGPT, Claude, and so forth. I use them all the stinking time to gain knowledge of the things I’ll be teaching. What used to take hours of research now takes seconds. They are, warts and all, ludicrously efficient at delivering mostly-true information about abstruse topics.

But oh, there are warts.

I’d like your help. I’d like to take just one aspect of something people worry about with LLMs — their environmental impact — and see if what agreements we can, in the comments, come to.

I’d like to aim for two things:

1. what numbers can we agree on? (e.g. “using one normal LLM search takes x amount of electricity”)

2. how can we make contextual, human sense of those numbers? (e.g. “a half-hour of normal conversation with an AI takes __ as much energy as streaming TV”)

To help with this, I’d like to suggest a few things:

1. stay on topic — don’t talk about other AI issues (e.g. how accurate/inaccurate they are, fears of being paperclipped, effects on kids)

2. share where you’re getting your numbers from

3. don’t defame anyone else’s source (e.g. “well that person is a known climate-denier” or “once shoved a dog down a flight of stairs” or whatever) — provide what you think to be a better source

I’ll delete any posts that in my mind egregiously violate #1; if you’re on the border, I might ask you to edit it. Be over-the-top kind to people you disagree with. If it helps, put on a nice set of clothes before hitting “reply”.

Imaginary Interlocutor: Brandon, how do you stand on this?

Lightly. My starting point for this is Andy Masley’s recent post on this question:

Why using ChatGPT is not bad for the environment - a cheat sheet

I recommend starting with it, too. He responds to objections like “A ChatGPT prompt uses too much energy/water”, “ChatGPT is ten times as bad as a Google search”, “Data centers are an environmental disaster”, and much more! Here’s one of his diagrams that really hit home for me:

h/t Andy Masley

If you read it, you might also want to read his replies to the criticisms he got on his original longer piece — and then you might want to read his longer piece itself!

Where Masley’s piece is wrong, I want to change my mind on this. Where it’s right, I’d like to be confident of it, so I can be bold in helping teachers make use of these tools that can help us rehumanize education.

This is an experiment to see how well our emerging little community can do with a controversial (but still fact-based) question. I’ll try to call out my favorite responses by name — feel free to crown yourself in glory!

1. I’m focusing mostly on elementary & middle school education here. A mind full of pure, inhuman, abstractions actually sits close to the crowning achievement of Egan education — I tend to focus on the opposite because I’m trying to ward against the classic mistake of educational Traditionalism, which is to mistake that goal for the path itself.