“To achieve great things, two things are needed: a plan and not quite enough time.”
―Leonard Bernstein
So I started building a brain emulation way back trying out the Hodgkin and Huxley neurones. Trying to get it all to work I had a few hiccups and at least one major bug in that it over reported accuracy.
I spent a while fixing it. This one works and was a stress test of the code because the Hodgkin and Huxley formula can produce infinity or NANs if not exactly balanced.
This is to show working draft. I started this it was to wade into the AI debate if AGI or ASI is even possible. This is a biologically inspired self growing, self architecturimg neural network.
https://youtu.be/jZc6bzcQKv8?si=hdokNaPtBMpO7F-G
I thought that was a cool project to get on with and with the visualisation you really see how it all builds itself out.
its accuracy still needs to go up. This one is over clocked as a test such that it's neurones fire too much to be stable (it still self improves) so you'd assume if made a mistake with any of the maths it would crash in 5 hours I ram it last night. It's still on and doing fine. What I might do is if it keeps self improving drop further videos here.
But I think as a argument that AGI is at least possible I think you can see from the visualisation the EEG changes shape along the bottom from rapid thinking to lower smooth brain waves so I think it's working as intended.
I think representing each neurone as a red, black or green dot works really well and you can really see the way neurone avalanches cascade through the system. It looks like it's fitting a pattern until the pattern changes and my understanding is that in human brain development it starts like that then the pattern becomes infinite and non repeating as it becomes more complicated and in our brains it's fully non repeating and very dynamic.
I think and this is me being arrogant here you can see there are shifts in the pattern as it grows and that's quite cool to watch.
Also hilariously the top comment on You tube is loving the noot noot sequence. Haha I wonder if he thinks I'm joking about the brain emulation. For the record there's a few out there I haven't looked at the performance of any of the others. But as per the quote at the top of the page... I didn't think I had enough time so I thought I'd start building
:-D
Add comment
Comments