Eric HertzI've only a tiny understanding of neurons, but as I understand a neuron takes in many [analog] inputs and "fires" when some threshold is crossed. The output is, essentially, binary... 'zero' most of the time, then 'one' very briefly during the 'firing' of an impulse.
Anyhow, I think what it boils down to is the *timing* of those singular impulses ('ones').
And... that's *exactly* what this storage-method is all about! Output consists of two impulses, the time between determines the value.
Have I inadvertently stumbled on neural memory?
(Thanks @Lee Djavaherian for the comment in the last log!)
Discussions
Become a Hackaday.io Member
Create an account to leave a comment. Already have an account? Log In.
Interesting! But just to clarify: they were not equating the neuron to the transistor, as they often can't even image individual neuronal activity in most cases, but were just comparing the results of their techniques using the 6502 memory patterns to the brain. Link is here (edit: sorry, broken link earlier--has been fixed):
https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1005268
Since your incandescent memory would emit light, it would allow this type of analysis by default. For example, just put a videocamera on your setup and you could record something similar to a temporal fcMRI brain scan.
If you just used it as a frame-buffer (which is a really cool idea, by the way!), it wouldn't tell you anything we don't already know, but if you used the memory to build an entire computer, what would those flashing lights tell us in relation to the underlying program being performed? Could we one day find a shared pattern between an idea and the "thought processing" of that idea that is the same regardless of the memory architecture? It seems impossible, and yet that's kind of how neuroscientists find correlations between the shared functional brain activity of many different people and what they are doing or thinking about at the time. Could it be that these networks formed in the way they do to efficiently match the "shape" of that idea?
I've always wondered if abstract ideas actually have shapes, like a type of geometric archetype. We know simple ideas do, as a repetition or cycle could be equated to a circle, for example, and perhaps something like a circle would also appear in the flashing computer memory when processing one. The lights might assume a repetitive pattern, which if traced, might be a deformed circle that could be transformed into an actual circle via some mathematical operations.
But we haven't isolated complex ideas into shapes. The lights might have to be mapped onto higher-dimensional topological spaces, I don't know, but one could get creative in how they orient those light bulbs in 3D space.
Are you sure? yes | no
Sounds interesting. I'm sure patterns can be seen in viewing RAM.. Kinda reminds me of watching a hex-dump in realtime in vidgames. Some patterns are very visible even when displayed in hex... e.g. one value always changes once per second, while most others surrounding it vary faster/randomly, or stay relatively constant. That one periodically-changing value sticks out, even if you don't know hex or don't look close enough to see that it always increments.
I guess some "universal" things like that are quite apparent. It could be that more-complex things are, as well.
Guess one'd just have to watch the memory contents till a pattern becomes recognized.
That link is a no-go, BTW.
Are you sure? yes | no