For me, that's probably actually harder than reinforcement.
Is anyone actually working on this? Or have all the contestants abandoned it?
5 Days till the deadline.
I haven't started because I don't know what I'd do for it.
i want to join
I already posted about this:
Joining for the contest has been closed.
Refer to this article that a lot of people helped me on:
oh god its tmr? I barely even worked on it...
Has anyone done anything yet? I think this challenge was a communication bust, what point is everyone at?
im making an ai that if you say something 3 times that isnt in the word list, it gets added. Im done with that part, but not with starting the convo your self
Not really AI or ML though, but, I make things that aren't AIs, just interpretations of them, and call them AIs even though they aren't. So I guess I shouldn't be a hypocrite.
a machine cant actually learn. Basically, if you say something as a reply to the bot, and it has already existed in the work list 2 other times, it will learn that it is an appropriate response, and there is now a chance for the bot to say that.
I mean, yeah. You're kinda right about it.
Basically ML or AI programs don't learn, they run scripts that force the AI or ML program to learn. It's like how your teacher teaches you in school about concepts like English, Math, Social Studies, and Science.
I have thought about this because I published a few papers about how AI works (or so I think). If you want I can tell you how I think AI works.
Please, go ahead!
Okay! (Sorry for the long delay, I was on a piano lesson and off to go do some mountain hiking)
Anyways, here's my opinion about how AI works. You don't have to teach it or take any notes because I am NOT a college graduate and definitely don't have a PhD or Scholarship from MIT. So if anything, you probably shouldn't teach information that hasn't been reevaluated or anything from a Middle Schooler.
So, let's disclose a few things. We have 3 modules: The Memory Module, Thinking Module, and Reset Module. The order matters because it indicates which part is supposed to be coded first or which part COMES first. Personally, I think the Memory Module comes first even though the Thinking Module and the Memory Module are one Module combined together which forms the basic AI.
Anyways, enough with that. I keep saying "anyways", so please get used to me saying that over and over and over. I know it's intimidating but shut u-
The Memory Module, as the name implies, is the Memory for the AI. This Memory is not Classified; it can be accessed by anything, even the user. Anyways, it's just a list full of variables like floats, booleans, integers, strings, and other whatnot. You can choose to delete all this information and replace it with other pieces of information. I get emotional just talking about the Memory Module.
The Thinking Module is the BIGGEST piece of the AI. It's full of if statements that are full of more if statements. And if all those statements return a true output, then it will run through two states - the "Before" State and the "After" State.
The Before State is basically when the Memory is empty and so the AI has no idea what to do. So it makes a random output in hopes of gathering more information.
Once the Memory Module starts filling up with a bunch of information that is kept in variables, the "After" State kicks in and the AI starts to use the Memory Module to make outputs.
After the Thinking Module comes the Reset Module. I've talked to a lot of people outside of Snap! and they are very confused about the Reset Module and think that it is unnecessary, but you have to understand this: Memory is NOT infinite. And so you have to do some Garbage Collection here and there and delete some information. And that's what the Reset Module does. It resets the Memory Module and the entire AI, meaning that the Thinking Module is affected.
I should probably go more in-depth about all this stuff sooner by editing my original articles that I made when I was 10.
Yep.
You basically just described RAM, the CPU, and free()
U g h. Give them an inch, they take a mile.
I will extend the deadline to the 30th.
This will be the last extension. If no one submits anything by then, I won't know what to do.
Wanted to highlight a presentation from SnapCon where Ken Kahn shows how to create deep neural networks in Snap. Might be a good starting point or spring board for ideas:
Yeah, but it also needs a brain and such. It's not all if statements. 