When I get back into the classroom in January, one thing I’m looking forward to is the ability to generate lessons with sophisticated computer graphics by simply clicking a button. So fancy! So informative!
The irony of teaching about water conservation with AI does not escape me.
How much water does Atlantis have? It’s apparently located in the Atlantic.
My brain broked.
I’m 95% sure that’s a shitpost and not a real assignment.
AI says:
Odd that they would specify temperatures on two different scales in the same paragraph.
Well, I learned something at least.
I can spell better than that AI.
Somebody needs to give the AI a dictionary.
Also, Atlantis is in the wrong place.
The AI has it at the equator, between South America and Africa. It should be a lot further north.
“USS DEIN
PLOW”
It could be a shitpost, but then, ask yourself: could any AI do any better?
Unheimlich und unsinnig.
Uncanny and nonsensical.
In recent months I saw a few articles about use of AI in school, the fears for pupils cheating their way through school without learning anything – never saw anything about the risk of teachers using AI.
I can so imagine one of my old teachers pulling shit like that. Guy was so fscking lazy, never preparing any lesson, at best going through a chapter of the textbook with us and then simply taking the tests from some random textbook or model lesson plan that had little to do with what our textbooks covered.
If this guy was still teaching (I doubt he’s still alive, last time I saw him his alcoholism had seriously marked him), I’m sure his students would get AI generated exams.
Just for the record: I had some excellent teachers, dedicated teachers too, of course.
Isn’t this sorta cherry picking a worst case scenario of using AI if it was even legit?
My own experience with using AI in Canva for generation of visual output leads me to realize there are shortfalls. One must sort wheat from much chaff. But it has worked pretty good with some amusing mistakes cast aside. I do like three eared rabbits though. I don’t ask AI to render textual output when doing images. That never works out well for me. That’s frustrating. But I can craft my own text boxes to put in as modular components after AI generates images based on my specifications. And I have to be careful in what I specify. In some cases the uncanny valley actually works in an appealing way.
My experience with AI doing some comparative analysis of a couple similarly themed books was not super disappointing. And AI did well at generating discussion questions for a book I was reading.
Mostly for my uses I’ve queried AI and it throws things at the wall and based on that output I determine what sticks. I don’t dwell on the hallucinations and harp on that.
I low-key want to change my username to “Aval Babity.”
Rain Rain Barkels was my favorite forgotten 80s New Wave band.
PZ Myers @ #7 — “could any AI do any better?” I gather it depends on the prompt…now known as “prompt engineering”…and the context…now known as “context engineering”…and the number of parameters or tokens you’re willing to pay for. The more tokens the more costly, and also the more time to generate the results. Success depends on the person working with the model. It really isn’t magic, just statistics…despite the hype. AI/LLM are just tools. Don’t blame the tools.
Who uses a bucket and a half of water for their Bisic Ilcomations these days? When I was younger we only got one and were grateful for it!
@13 I don’t blame the tools, I blame the tools using them. I’m sticking to the tried and true approach of ten thousand monkeys and ten thousand typewriters.
I’m on LinkedIn, and people will frequently post AI-generated drivel as “useful infographics”. I download them and show them to my students for a laugh. If I was looking to hire someone, I’d check their profile to see if they’d posted such nonsense without proofreading it.
I think I’d far rather Russell;s teapot* exists over this AI’s also (almost certainly if not provably) magically planetary mass & I’m guessing not to scale water bottle here..
Although if it actually did exist and could be poured out it sure would come in handy for terraforming the Martian and Cytherean surfaces..
Also, be intresting to determine the temperature of the water in that larger than Earth size (&mass?) bottle & its expansion & contraction and the effects on the plastic of that & the solar radiation esp UV here.. There’s some surreal story or three in that ain’t there?
Oh & now imagine that’s carbonated soft drink not water and shaken up by its rotation / tumbling / micrometeorite impacts and that cap gets taken off and off it jets…
.* See : https://en.wikipedia.org/wiki/Russell%27s_teapot
I mean, one of them shows the structure of a “nucletoide” as a carbon bonded to an R group, two water molecules, and double-bonded to another carbon. That second carbon is also double bonded to another oxygen, and also double-bonded to an “OI” (whatever that is). This gets turned into an amino acid, which is apparently a carbon bonded to R, P, OH2, and double-bonded to N=NH.
And glucose contains an what would be a hydroquinone (except with triple bonds instead of double bonds) bonded to a five-membered ring with multiple double bonds and one hydrogen.
@8. Tethys : Mochiron.
Of course.
https://sknihongo.com/en/jisho/%E5%8B%BF%E8%AB%96-mochiron/
Honto*, I learnt some things about a Nihongo word I’ve known & liked since the 1990’s from that link just googled.
.* Truly. En Eigo / Anglais / English.
@8. Tethys : Mochiron.
Of course.
https://sknihongo.com/en/jisho/%E5%8B%BF%E8%AB%96-mochiron/
Honto*, I learnt some things about a Nihongo word I’ve known & liked since the 1990’s from that link just googled.
.* Truly. En Eigo / Anglais / English.
@8. Tethys : Mochiron.
Of course.
https://sknihongo.com/en/jisho/%E5%8B%BF%E8%AB%96-mochiron/
Honto*, I learnt some things about a Nihongo word I’ve known & liked since the 1990’s from that link just googled.
.* Truly. En Eigo / Anglais / English.
Dóh! Sorry. Dunno why my last comment just went thro’ thrice..
What happened to 2 and 3? I thought it was 4 that there is no.
Man, three tries at spelling “availability” and still couldn’t get it right. Sad.
bcw bcw @ #15 — “I’m sticking to the tried and true approach of ten thousand monkeys and ten thousand typewriters.” That’s great…that is, if you don’t mind cleaning up all that monkey poop for nada. It’ll never get you the millions of protein structure predictions that AI/LLM systems are achieving. But yeah, as you say, “blame the tools using them”.
The poster should want to smash the teacher. The computer didn’t assign this to the students.
I’m supposing the original post that PZ is quoting was AI generated #25 Walter Solomon.
I have seen some AI assist stories with AI author comments and AI criticisms lately. This whole thing looks like one. After all who is doing most of the criticism of AI slop? Humans do! Doing bad commentary is proof that you are human. Doing stupid commentary is also proof you are human. The creators of AI slop are in an “attention span” economy where they sell advertising and targeted propaganda, so any kind of attention they get is good, and it helps them do their fake. Therefore the AI criticism of this utter crap diagram is poorly done because most English Major type criticisms are poorly done and because they’re just not that good at it.
This particular fake commenter is prompted by a poster who is aiming to grab teacher attention, so it blames students even though the fake-teacher it itself created is the problem. Or it just can’t tell and it got spammed all over the place and is reproduced the most where it gets the attention.
I believe the correct response it to ignore it or block it, unless you think your fellow humans have been fooled or confused.
“It could be a shitpost, but then, ask yourself: could any AI do any better?”
Yes. Of course.
Depending on what one wants to do.
But not by merely pressing a button.
Assuming this is legit, I feel sorry for the teacher in question. How heart breaking it must be to not be able to spend even a moment looking at this before shoveling it out because you are so time poor, not to mention probably actually poor, to do any better.
I dunno, I think this could be an excellent teacher’s aid. Think about all the things a good teacher could use this one image to present.
AI sourcing and limitations: Clearly something went wrong here and it’s easy to point out.
Implied data but no actual data: Definitely something to avoid in their own projects, whether AI assisted or not. Keep data presentations targeted and relevant.
Recognizing bad faith arguments: This doesn’t prove anything other than a nonsense infographic can be created with AI assistance. Can easily transition to a “Where’s Waldo?” about finding the ways this shows you it’s a bad faith argument if someone tried to use it to “prove” anything to you.
And on top of all that it easily opens up a conversation about climate change, politics, lies, and why today’s students are likely to hear many different versions of that story throughout their lives.
While it would fit most readily into a general science class, nearly any class could benefit from some of these topics coming up.
Sending it home to parents might be a bit questionable though. Some parents apparently want their children to grow up doing as they’re told, believing what they’re told, and generally being easily manipulated benighted ignoramuses who have no idea how to defend themselves against misinformation. The job of a teacher is always to prevent that outcome from occurring.
On the subject of AI-generated work that’s possibly a bit closer to home, someone’s been posting on Tiktok photos from a textbook that was allegedly written by the professor, but all of the images of the human musculature are generated by ChatGPT, and are predictably wrong, with mislabeled muscles and bones, completely wrong bone structure, etc. https://www.tiktok.com/@liaaa.bia/photo/7557473261169298743
True story: Yesterday, for reasons that will remain obscure, I wanted a picture of Peter Cushing at age 27 when his movie career was beginning, so googled “Peter Cushing age 27”.
The “AI” summary said (paraphrased, I nearly screenshot it but didn’t):
X-D
Artificial “Intelligence”