News from gopher9

Because the recent release of LLMs has been too vigorous, I organized recent notable models from the news. Some may find the diagram useful, so please allow me to distribute it.

Add RWKV.

Something I've caught a lot of flak for on here is saying that I dislike people suggesting "How to Prove It" or "Book of Proof." To be honest, I despise this entire genre of textbook. I think it makes people get a number of false ideas about what math is.

Using truth tables to introduce logic is arguably outright harmful:

Linear algebra and differential calculus.

Also some probability theory to understand VAEs, and also some signal processing to undestand papers like

RWKV works rather well on 4090.

It really is not great for teaching. And I would go so far as to argue that mathematics education was held back in part because it was used so much as a textbook in prior centuries.

It's still pretty solid for a more than 2000 years old textbook.

Whether a | b means that a divides b or that a is divisible by b. Whichever it is, I know a ⋮ b means the other one, but the fact that common notations exist for both makes it harder to remember what either one actually is.

Knuth suggests using a \ b for this.

Visual Group Theory has a lot of pretty pictures.

Because of the whole anti ai art movement and their claim that ai generated art is just a collage of art work or photos from actual artists’ works.

And thanks to collage diffusion their claims may finally come true.

Some very classic book:

Yes, there is:

Oh wow, I didn't know about this project! After a first look this seems to be very close to what I had in mind. Even though the covered topics are closer to abstract algebra I guess one can imagine how it would look if more basic topics were covered instead.

It's a weird combination of requiring no prerequisites but all the mathematical maturity. Every type theory construction could be linked to a mathematical idea, but these ideas are subtle and the very first chapter contains all of them.

Consider

I’m an econ guy so I see all these in non-calc applications. What’s the spooky math that goes with η?

Programming language theory is a pretty spooky field of math :P

ξ and η are sometimes used as fancy x and y. For example:

So much talk about ideal ways to make Rust GUIs... so many UI libraries.....so few (GTK only?) complete libraries with enough widgets to make anything beyond a toy. I realize theorizing is fun (I love it too), but I sure wish somebody would just settle on an approach, even if not ideal, and complete enough of a UI lib to make it possible to create a full UI in rust for a serious program (without resorting to JS/CSS/HTML). That said, people are free to work on what they want of course, I certainly do... just vocalizing what I wish would happen. :-)

Slint seems to go this way.

There's the casuals, and then there's the people who just wanted to either troll the artists, or legitimately screw them over. While I do know people who made mistakes like these, people who have been lying their butts out about this whilst profiting over it is just straight up wrong.

By the way... what distinguishes good art from bad art? A Jackson Pollock is considered good art, even though probably every average Joe could spill a couple buckets of paint over a canvas.

Control. A skilled artist has much more control over the end result, making a lot of subtle deliberate choices in the process.

Most important thing is the concept of mathematical proof, but it's hard to encode in a sentence. And most other things make litte sense without proofs.

You may also like a larger view:

Ha-ha, I cannot disagree with You there. I perhaps meant not right this instance, but for a project in the future, by then I probably understand a lot more, albeit I am curious to learn in terminology I can currently understand.

Here's another one:

Neural networks are by design black boxes. You get great performance in exchange of explainability. This does not mean though that you have no control over the result.

at one point she talks about how we could make do without imaginary numbers.

I would rather ask a dual question: why people introduce numbers in a type oblivious way? Make it clear that we construct a new type of numbers, and all the mystery just goes away.

There's a paper that does that and also other transformations as well:

Phyrgian. Or if you're feeling extra spicy, Phyrgian Dominant!

Another fun option: locrian dominant. Major third in the root helps to stabilize it a little bit.

I joined the Sigmoid Mastodon. It's a wasteland of people posting AI "art," pseudo-intellectual gibberish about AI, and nonsense that belongs on the worst parts of LinkedIn.

Did you take a look as Mathstodon? There are some actuall mathematicians and computer scientists there, so maybe it's a better place to look at.

MachineLearning

math

I find it disheartening that so many people are completely trampling the discussion here by just vomiting the most obvious pseudo-answers without going into any depth, i.e. "this is just another technology lol". That is, for some people here, any kind of musical technology is exactly the same in that respect. No one is discussing the impacts of any of that technology in the course of music.

I have to upvote this, but also I'll try to defend AI a little bit.

“Obscure Code” seems to actually improve readability. While “Random Whitespace” is a simple but extremely painful transformation.