primerlearning
patreon
Simulating a single brain cell
🕑 Added 2025-09-27 15:07:54 +0000 UTCHey everybody,
This video focuses on single artificial neurons, building up the idea of gradient descent and applying the single-neuron system to two kinds of applications: A distance predictor for a cannon, and a mango classifier based on length and mass.
I had originally planned to also cover more complex systems and to talk about AI more broadly, but I decided to cut scope. Even this smaller video is the longest video I’ve made on this channel, and I think it turned out to be a good conceptual stopping point. The next video will continue the idea, addressing the multi-layer perceptron, backpropagation, and how deep learning allows approximation of any continuous function (assuming you have the right combination of data, compute, and algorithms).
For those eligible for the patron-only live streams, we’ll be doing the stream at 10 am US Central time on Saturday, October 4. I’ll make a new post for that, which will also let you view the video-on-demand if you can’t make it for the stream. Please use this post to let me know any questions you would like to discuss during the stream. You can also ask them live, but if you ask early, I’ll have a chance to prepare, possibly setting up an animation or some code we could explore.
Thanks, as always, for your support.
Justin
Comments
Sean
Best explanation of the concept I have come across
shikaji
"begging for money" section of the video did its job :D all your work has been great but i felt this video in particular was so incredibly well explained and demonstrated i'm more than happy to throw in some money as support. well-earned! excellent choice to cut scope and focus on a single neuron first. i watched plenty of AI / LLM / neural neutwork explainer videos but never saw this step focused on in detail like this.
