Yes, thank you, thank you!
I will record some commentary this week for the final release that will help demystify. If reading it via text doesn't click, it will click better when I'm going over it live, so stay tuned for the full thing.
But this puzzle is like a thread. The more you pull on it, the more more you unravel, the deeper it goes. This is basically a picture of what's running the whole world right now. It's about far more than guessing numbers. This is how Google searches work. This is how Snapchat filters work. This is voice recognition, handwriting recognition, it's all up in your phones. It's how you think, it's literally what your dreams are made of. It is literally a digital brain.
You know how if you place a magnet in the middle of a bunch or iron filings and they all kinda line up along the magnetic field? That basic structure is the magnet, society is the filed iron. This is not only the structure of how your brain works, abd how AI works, it's how information is processed on a primal level. Even on a macro scale, ie, if those were people instead of neurons, it still demonstrates how information is processed and kinda boiled down, and just like in real life can lead to different conclusions with different biases.
In fact, if you imagine the neurons as a tribe of binary people, it's a bit easier to understand.The basic structure is actually simpler than it looks. There are several layers, each denoted by different colored glass. And they do, indeed process in order. It's goes like this:
First row (yellow) is the retina. It's a direct copy of what's on the screen, it's just separated out in a line. These are direct witnesses to the original "event". They all saw something. Some of them are alarmed (on) and sone are not (off).
Second layer is called V1 (red). V1 is the closest friends to some of the neurons in the retina. Each one is looking at three other retina neurons in this case. These will generally be 3 in a row horizontally or vertically, some diagonals. Neurons #1,2,3 might activate number 41 while 42 gets activated by #2,3,4 for example. There can be overlaps.
Second layer, called V2 (blue)t same thing, but goes a step higher. These villagers aren't as easily moved as the last group. But they know if #41 and #78 both happen to be lit up (fir one example) then there actually might be something going on, so they light up. They're forming slightly more complicated shapes now.
We skip v3 here because v3 is about processing colors. We're keeping things "simple" in this Minecraft works, so we just don't need it. A reminders, this is all based on our real brains and how we really process visual information. "The chicken or the egg" thing here is the brain cane first. Computers using this method came second.
Moving on to v4 (green). These are the "beat reporters" of the village. They're really not easily moved. They only bother noticing certain patterns that are relevant. The chain often ends here unless it recognizes some basic shapes. But if these guys get lit up, you've got yourself a "story" abd they send their info over to the last layer, IT.
The IT (purple, right under the big screen) is a little bit different. They're more willing to listen to everyone from v4 because they know v4 has done some homework. In other words, they're OR gates instead of AND gates. This is where the guess is determined.
This is what I needed that last circuit for. Now you've got a bunch of abstract shapes that all could be parts of a number, but how does it make a choice? The answer is biases. For example, a number 1 and a number 7 aren't too far apart, shaoe-wise, so you'd think any 7 might look like a 1 with some extra stuff. And it does, so when you activate both, the 7 takes precedence because it has more features. Same thing with an 8 vs a 9. They share a bunch of the same features, so if something has everything a 9 would have, but also that one thing and 8 would have, then only the 8 fires and the brain has made a decision.
After that, it just goes to the big screen. It was originally just fir the output fir the player, but looking back and seeing what looked like all those neurons in a big theater looking up to the big screen in the sky to tell them the truth, man, it done did my head in. Imagine different screens with different biases. It's not an exact copy of the real truth, it's always dissolved, and, depending on how those biases are set, ie depending on which news source you prefer, the "truth" could be entirely different things.
Anyway, fascinating stuff abd I hope to write these for a living. This one here is a known algorithm made in MIT. What makes these things so magical is how they're figured out. You can have any random data on the retina. Let's go with, say stock market data. You train the algorithm. So, let's say when a stock rises, that's a yes. If it drops, it's a no. You let it train itself one tons of examples and it will pick up on patterns a human wouldn't even think about, based just on what patterns tend to keep popping up. You won't need to know what those patterns look like or why, the algorithm does it all for you. It literally "learns" the same way WE learn , only it doesn't have a faulty memory. I can literally predict the stock market out to a certain degree. Not with perfect accuracy, but with enough to make consistent profits. This is actually how how weather is forecast. It can only guess so far because chaos and all, but it works! It picks up patterns out of nothing all on its own and makes genuinely accurate predictions (I'm working on a Bitcoin bot next
) There's really no end in sight to what they can do.
