jedediyah.com
Teacher ◦ Robots and mathy science ◦ He/Him
#ITeachMath #ITeachPhysics #TeamCompSci
#mtbos #PAEMST
Massachusetts
https://jedediyah.github.io/
295 posts
726 followers
689 following
Regular Contributor
Active Commenter
comment in response to
post
It's such a substantial topic with loads of prerequisites, so I'll be surprised if there is anything out there. But happy to hear about anything anyone finds!
comment in response to
post
3blue1brown's videos on LLMs are accessible and well done. Definitely a lot of info to cover, though.
I've never done a unit on LLMs, but do simple chat-bots, which is fun. They will hard-code responses to certain inputs, and use Levenshtein distance to determine which input to respond to.
comment in response to
post
🤣 😂
If you took a bite, could an outside observer watch that bite slowly fade into eternity??
All of the spaghetti you eat would get sphagetti-fied
comment in response to
post
Ok, this is rocking and rolling now:
Choose number of terms k and check if:
(2025 - ((k-1)^2+k-1)/2)/k
is an integer.
That feels way less gross than it looks. And fits nicely in a spreadsheet
comment in response to
post
Oh wait, I think it should be T(k-1)...
comment in response to
post
I left off the RHS of:
kn + T(k) = 2025
so that:
kn = 2025 - T(k)
n = (2025 - T(k))/k
comment in response to
post
2 terms:
n + n+1 = 2025
3 terms:
n + n+1 + n+2 = 2025
So for k terms, we will have:
kn + T(k)
where T(k) is the k^th triangular number
But many don't work! This is fun.
We also need to have k | [2025-T(k)]
comment in response to
post
Reminds me of those Serato vinyl records that are tme-stamp encoded to be able to control digital music files.
comment in response to
post
Yeah I think the dep/ind when the data was collected will always be that, regardless of later math.
We tend to intro problems as disconnected from where data come from, so we oversimplify:
x-axis = independent
because we lack context
comment in response to
post
I have always thought of dependent and independent variables in the context of experiment: there is something you control and you look for a response in another thing.
I think a dep/indep distinction is not that meaningful in the math.
comment in response to
post
Waves, in physics. Today was electromagnetic spectrum.
comment in response to
post
My son just taught me that algorithm!
Middle x2 on each axis.
comment in response to
post
Totally. I would so prefer a student be able to write a few lines of simple code that they comprehend. That's a huge win these days.
comment in response to
post
The most noticeable difference for me is the huge amount of time struggling to work with students who have some issue with a solution not working and I cannot even begin to explain the fix because they don't understand the most elementary concepts in the complex nonsense they copy-pasted. It's sad.
comment in response to
post
I spend very little time trying to detect if students have used AI.
I do spend a depressing amount of time saying things like:
"I see you used jQuery syntax but did not include the jQuery library. What inspired you to choose jQuery for this simple task? Oh, you don't know what jQuery is?"
comment in response to
post
"... students take my courses to learn the fundamental capabilities and limitations of computing devices."
The capabilities and limitations to do what? Compute? Compute what?
Is this how applied math works:
1. Abstract real-world situation into variables
2. Now that situation doesn't exist
comment in response to
post
Long, and don't skip the 'y', no matter what your spellchecker says 🤣
Thank you!!
comment in response to
post
Thanks! There's very little time put into this. I did one level just to see if it would work, then copy-pasted another level, added a thing, changed a thing. It's an unedited mess! But fun 😄
comment in response to
post
Yes, please!
My gmail is [my first name]
comment in response to
post
I always struggle with implied commonalities between students and employees. One significant difference between those two being that the product hardly ever matters for students. I already know the answers to the problems I'm posing.
Let's believe CEOs when they say they are optimizing financials.
comment in response to
post
The current LLM pipeline in my intro CS classes:
1. I say don't use AI
2. They use AI
3. Their project is riddled with bugs
4. They ask what's wrong
5. I tell them the issue and how to fix it, but they are in way over their heads and have no idea what I'm talking about
comment in response to
post
😂
This was the best image Ss could get from a state-of-the-art "reasoning" model after many iterations of telling it what was wrong and it just kept getting it wrong in the same way.
Then we asked it for a set of base-2 manipulatives...
comment in response to
post
Yes.
I think there has been a recent shift in MA in particular. Feels like there is developing top-down pressure that is entering school buildings now.
comment in response to
post
Maybe their "AI" is just a lookup table of what the weather was on that day last year...
Most of my arguments around AI converge toward mathematical modelers ignoring the constraints of mathematical modeling.
comment in response to
post
The copies I have found had only limited viewing, so it included the front cover and the back cover 🤣 but I'll peek around again.
comment in response to
post
I believe I've seen this most frequently recommended from @benjamindickman.bsky.social. Great text! I agree that I got a lot out of the first chapters.
comment in response to
post
'Scatter' is the the default plot style in most plotting utilities. People should use it.
The man needs to stop trying to put my data in a box!!
comment in response to
post
It was when I got to college that I saw HS math actually being used for something. Plus this other math I'd never seen before. It was amazing, and I liked math for the first time.
But when I look at the stuff that I personally used A LOT for "real" things, it's 90% trig with matrices.
comment in response to
post
ToC
comment in response to
post
Oh gosh, I don't even know where to start, but count me in!
comment in response to
post
"AI" had very different connotations then, but there were some really interesting things happening. I just didn't see how we were going to jump straight to solving society's biggest challenges.
comment in response to
post
Not clear, because the solutions didn't exist, but oh boy was it around the corner!
The ideas were really just dynamic decisions re: practice. Like, can't do fractions? Try fractions.
At the time, I think support vector machines were the biggest deal, and various Markov things were happening.
comment in response to
post
For those familiar with the decades of AI failures, AI failure may not be surprising. But with chatbots being so mainstream, AI is getting a fresh audience, and the promises are getting that wave of new hype.
But then, so many experts really believe it, too.
comment in response to
post
I was told almost 20 years ago to get out of education because it was almost solved (with AI), and teachers would be extinct in 5 years.
The same with self driving cars or any problem, the last 10% is so difficult that "almost solved" is meaningless.