I’m gonna click it even though I know what it is (not telling anyone what it is)
I clicked the link saw the thumbnail immediately closed the tab
1 Like
Too bad… I GOT ONE!! >:D
2 Likes
This is what happens when you breed two neural nets together. SO… MANY… ARRAYS!! I’m glad its a nice shade of orange and not like, radio pink.
4 Likes
I clicked it guys ): (TOO BAD I LOVE THIS SONG!!!) 
Its on my playlist Guys!
2 Likes
WHAT IN THE WORLD IS THIS??? omg this is a sentence.
1 Like
I’m so sorry to say but something has went terribly wrong with the AI frogs. They seem to evolve to have the same opinion on every color! Like, some frogs will say hate every time, and some will just say like every time. I’m worried something about the nets is broken, but its so hard to diagnose as I can’t just ‘console log’ the entire net to see what part of it is acting funny. : (
I have a few theories:
- The nets are too simple. With only 2 hidden layers with 4 perceptrons each, the frogs could just be… too dumb
- The randomness is not random. If the random function in makecode really is broken like some users have experienced before, then a ton of my code is broken.
- The user is inconsistent. I mean, so much of color favoritism relies on aspects like lighting, mood, surrounding colors, and just guessing! The frogs are trying their best, but perhaps they are just spouting the most common answer as they cannot find a pattern.
4 Likes
What…? 
2 Likes
Good news! I’ve ruled out #2, which is the most scary one. I logged all the weights of the nets and they are perfectly fine. I’m thinking its actually the activation function that is hurting the nets- i’m using a step function, but because color predictions are about probability, the frog is limited to ‘yes’ or ‘no’ when it should be a spectrum.
3 Likes
Update: The calculations were recursive anyway (Looping upon itself) meaning that it would only ever calculate the same answer because… Well because thats not how its supposed to work. I don’t know, it was confusing, but its fixed now!
2 Likes
OH NO! this cannot mean anything good.
2 Likes
in other coding laguages its much more worse than that 10000 lines of code
3 Likes
Don’t worry, i proved it false : 3
3 Likes
Ok just to notify you AI are not truly random (yet) so they are still encrypted so people can access the algorithm and change some things in the game
2 Likes
Update (Again… thats the third time today): The ‘mutation’ rate on these frogs is WAY too high. They either all end up screaming how much they hate the color or love the color- no inbetween! This means that they are mutating the weights so far they are actually somehow reaching above 1 or below -1… def going to fix that. I feel like i’m SO close to having the perfect Neural Networks!!
3 Likes
I’m a little confused:
Did you mean that the random number function is not random? I’m aware my AI’s are not random- they are trained to recognize color!
They were never encrypted- the code 100% open for everyone to see!
Ohh… I think i see what you mean here. The algorithm is not some pre-programmed line of math that every game shares- the AI is random at the beginning of the game and is personalized to the user. If someone went in and changed the code, it would just be a broken game.
2 Likes
this is how I image the frogs: I HATEE I HATEE THIS COLOR RAHHHH I I HATE IT. or I LOVEEEE IITTTTTT ITTTT OH I LOVE THIS COLOR I LOVE THE COLOR RRRRRR ANYONE WHO HATES IT IS WRONG!!!
2 Likes
Oh no my frogs are radicalizing 
5 Likes
What i meant they are not truly random and what i mean by that not your project i mean if your gonna do other AI project they are not truly random yet
Not all are shown code are shown but in terms of your project its 100% shown
Yes and i learned all those the hard way my AI project got hacked
2 Likes
and AI random means Psuedo-Random
1 Like