I would love to hear about your favorite entry-level visualizations that explain how neural nets work. I know about really impressive, detailed renderings (e.g., bbycroft.net/llm), but I'm looking for something appropriate for a first encounter. - ThreadSky

miriamposner.com • 5 days ago

I would love to hear about your favorite entry-level visualizations that explain how neural nets work. I know about really impressive, detailed renderings (e.g., https://bbycroft.net/llm), but I'm looking for something appropriate for a first encounter.

Comments

zoeleblanc.bsky.social•5 days ago

Sure you already know about this and it isn't really LLMs and neural nets, but just in case I really like this one for explaining machine learning http://www.r2d3.us/visual-intro-to-machine-learning-part-1/

miriamposner.com•5 days ago

I didn't know about this! Thank you so much!

miriamposner.com•5 days ago

There is so much garbage out there, it's impossible to sift through it.

ncecire.bsky.social•5 days ago

This statement truly applies to so many things

miriamposner.com•5 days ago

#evergreen

elbienmaspreciado.bsky.social•5 days ago

It has an ad near the beginning, and the second half gets a bit complex, but my students found this really helpful for understanding what a neuron is, how weights work, etc.., using the perception as a model https://youtu.be/l-9ALe3U-Fg?feature=shared

miriamposner.com•5 days ago

Oh, thank you, sounds promising!

elbienmaspreciado.bsky.social•5 days ago

*perceptron

sciencebanshee.bsky.social•5 days ago

Hear me out - Neural Networks for Babies.

It's colourful, clear, uses text sparingly and effectively, and I've managed to successfully teach MANY people the basics of neural nets with this.

Bonus: it's made out of cardboard, so if anyone gets frustrated/hungry, the pages are delightfully chewable

miriamposner.com•5 days ago

Honestly a good suggestion!

mthompsonbrusstar.com•5 days ago

I have repeatedly sent my students to the 3blue1brown videos on neural networks and I think it works well for most people with ~high school math?

miriamposner.com•5 days ago

Thank you, I'll take a look!

mthompsonbrusstar.com•5 days ago

Apologies for not posting the link! It’s here, and there are many in the series (e.g. “how does backpropagation work?” etc) if folks have calculus or more, although if I’m being honest the dynamic visualizations (eg on the segment on gradient descent) are good even without

https://youtu.be/aircAruvnKk

miriamposner.com•5 days ago

I was just watching this until I had to step into a meeting! It's excellent. Thank you!

angengea.blog•5 days ago

I really liked this series: https://www.youtube.com/playlist?list=PLZHQObOWTQDNU6R1_67000Dx_ZCJB-3pi

duvekot.bsky.social•5 days ago

I rather enjoyed this: https://www.youtube.com/watch?v=l-9ALe3U-Fg

miriamposner.com•5 days ago

You’re the second person to suggest this! I watched it today for the first time and found it really useful.

anirula.bsky.social•5 days ago

Where I found it most effective is to show people how the next words when they go to text something in their phone. The guess of what they mean next. That LLMs are essentially a massive version of that. And then they get the concept. After that I can explain more deeply.

miriamposner.com•5 days ago

Good idea!

anirula.bsky.social•5 days ago

Long career in tech. Didn’t work on AI and had concerns about the model they are using.

Then I lost my job to AI. Fun times. Fuck that bullshit. I love seeing people like Satya admit the current models aren’t profitable.

Comments

Posting Rules

Reply