วีดีโอ

It would be cool to see a tour of this thing and how it works.

i hate ai, its going to be the death of culture and human expression. all videos in youtube will be made by ai, the "users" who made them will be bots, etc.

this gives me 1.4 minecraft vibes

I like that thing!

cool

Holly Crap That is complicated, epic video!

amongus

Goof

Amazing Creation Man, id love to see an update on it, it terrifying t see it chase you down like that

erm what the sigma lllllllllllllllllllllllllllllll lll ll lll llllllllllllllllllllllll lll lllllll lll lllllllllllllllllllllllll lll ll lllllllllllllllllllllllllllllll

Неплохо, но сыровато.

Don't know what's going on, but keep it up! I could never do something like this in Scratch.

What, Really? No one have done AI in scratch. If it really neurons, you appreciate a thousands of viewers. Also, I didn't understand, wdym by sounds at the end?

это блять индукция?

wow, so when you are planning to make fully functioning computer with DOOM on it?

this is really poorly optimized man, use some loops and idx variables to do the repeated calculations. is this neat alg?

you can create system that detects when rocket detached from launcher and starts thruster. just put 2 nor gates in rocket and 1 and gate in launcher, then connect and gate to 1 nor gate at rocket and connect this nor gate to another nor gate than connect second nor gate to thruster. How it works - when nor gate has no connections it gives 0 signal but if if has connection to inactive and gate it gives 1 and this output reverses by second nor gate so while rocket welded to launcher and gate from launcher connected to rocjet nor and it gives 1 and reverts it to 0 so thruster off but if you break cardboard piece on what rocket welded to launcher connection to and gate breaks and nor gate outputs 0 and reverses it to 1 so thruster on and flying rocket, so just weld rocket on cardboard with this system and enjoy rocket launcher that dont needs you to break blocks manualy /-------------------------\ /-------\ /-------------\ | | | | | | and gate nor gate norgate | launcher rocket thruster

read description for more info on it should i update it for modern combat?

obstacle* ;)

idk whats going on or why this got recommended to me but cool

what else should i make this ai do

Very nice

Can you please make a tutorial of if you duplicate a sprite with a variable with(For This Sprite Only)

he talks about the formula, but i have never seen a formula that just multiplies EVERYTHING TIMES EVERYTHING. Seems odd. Normally you have the activation function of the sum of the individual weighted inputs of a neuron.

why is The Joker making a neural network tutorial in scratch?

this is a good demo on how neural networks work but actually training on this would be really bad as it is quite slow

I was wondering how long it would be before someone made a computer in this game. I look forward to this.

thanks this is an easy explanation how neural network works.

Love the video but the microphone...

is this a neural network? at 0:54 there doesn't seem to be any mention of bias or activation functions and I think the equation is supposed to be a weighted sum so i1 * w1 + i2 * w2 + i3 * w3 ... i_n * w_n + b and then activation functions (if there are any). Edit: So, after watching the whole video, I'm a bit confused but I think there are some things that are incorrect. NOTE: I might be wrong as well, I'm not an ML expert but rather a hobbyist. 1) It seems like they're trying to do RL with neural networks and a genetic algorithm (ish) approach. At 1:28, I'm fairly certain this is a mistake but I'm not entirely sure. It seems like they're talking about the idea of a living penalty (a negative reward given at each timestep to incentivize the agent to achieve the task as quickly as possible). It should be a set reward to -0.1 at the start of the reward code and all the "change rewards" should be modified to "set rewards." The issue right now is that it's "change" instead of "set" so any existing reward given is slightly penalized by -0.1. This means that if the agent's distance is <50 away it would be canceled out and it would be 0 reward instead. 1a) Not sure if this is intentional, but the rewards seem to be stacking due to the way the if statements are set up. If the agent's distance is <10 away, its actual reward would be: 0.1 + 0.3 + 0.5 + 0.8 - 0.1 (assuming it's not touching sprite 2). If the intention was to setup a nonlinear reward function, then I would suggest plotting some datapoints into a graph and using a regression calculator (or just do it yourself) to make a nonlinear reward function that takes in distance (x) and outputs reward (y). 2) 2:22 seems to be code for mutations. The problem is that your code is selecting a random item, x, but then repeatedly multiplying the *same* item six times by a random value from -2.001 to 2.001. This is essentially an exponential(ish) function, albeit random. (Example: Let's say we got really lucky and got 2.001 each time. It would look like: x = x * 2.001 * 2.001 * 2.001 * 2.001 * 2.001 * 2.001, or x = x * 2.001^6). Also, mutations are supposed to be small changes throughout multiple connections. There's a balance here: too much change and the model's strategy keeps changing wildly while never fine-tuning its strategy, but too little change and the model never explores new strategies. 3) (This is really long already, but this is just a somewhat cursed mathematical explanation for how the neural network should actually look like, so feel free to skip if you knew this already) The formula they listed at 0:52 seems to be wrong. I'm not exactly sure what they're trying to do so I'll just list the correct formula (note, I don't think I'm using the standard correct notation, sorry) Here's an example: *My (cursed?) notation:* *Assumptions:* - The number of hidden nodes in each hidden layer is equal - The activation function for each hidden layer is ReLU - The activation function for the output layer is TanH i = input h_xy = hidden node y in layer x (E.g. hidden node 1 in layer 2) b_xy = bias for layer x and node y c_xy = connection from node x in the previous layer to node y in the next layer n = total number of input nodes H = number of hidden nodes in one hidden layer weighted_sum_xy = the weighted sum for layer x and node y O = number of neurons in the output layer o_x = output x (in the last layer) 1) *Layer 1 (Input) --> Layer 2 (First hidden layer):* *1a) Calculating weighted sum values* weighted_sum_21 = i1 * c_11 + i2 * c_21 + i3 * c_31 + ... i_n * c_n1 + b_21 weighted_sum_22 = i1 * c_12 + i2 * c_22 + i3 * c_32 + ... i_n * c_n2 + b_22 ... (and so on) weighted_sum_2H = i1 * c_1H + i2 * c_2H + i3 * c_3H + ... i_n * c_nH + b_2H *1b) Applying activation functions (very easy)* - Our activation function is ReLU (x if x > 0 else 0) h_21 = ReLU(weighted_sum_21) h_22 = ReLU(weighted_sum_22) ... (and so on) h_2H = ReLU(weighted_sum_2H) 2) *Layer 2 (hidden layer) --> Layer3 (hidden layer, if we were adding more hidden layers, otherwise this step would be optional)* *2a) Calculating weighted sum values (basically the same general process)* weighted_sum_31 = h_21 * c_11 + h_22 * c_21 + ... h_2H * c_H1 + b_31 weighted_sum_32 = h_21 * c_22 + h_22 * c_22 + ... h_2H * c_H2 + b_32 ... (and so on) weighted_sum_3H = h_21 * c_2H + h_22 * c_2H + ... h_2H * c_HH + b_3H *2b) Applying activation functions* h_31 = ReLU(weighted_sum_31) h_32 = ReLU(weighted_sum_32) ... (and so on) h_3H = ReLU(weighted_sum_3H) *3) Last hidden layer --> Output layer* *3a) Calculating the weighted sum values* weighted_sum_41 = h_31 * c_11 + h_32 * c_21 + ... h_3H * c_H1 + b_41 weighted_sum_42 = h_31 * c_12 + h_32 * c_22 + ... h_3H * c_H2 + b_42 ... (and so on) weighted_sum_4H = h_31 * c_1O + h_32 * c_2O + ... h_3H * c_HO + b_4O *3b) Applying activation functions* - Our activation function is sigmoid --> f(x) = 1 / (1 + e^-x) o_1 = sigmoid(weighted_sum_41) o_2 = sigmoid(weighted_sum_42) ... (and so on) o_O = sigmoid(weighted_sum_4O)

good

what happened here? can you explain please?

I got recommended this a day after I just made a neural network on scratch

read description for more info

wtf. why did i get this recommended 4 hours after upload lmao this is epic

O~O

From scratch? IN SCRATCH!

Фух, чуть не попал

can you make a tutorial on how to make a neural network in lua or scratch pls

NOOOOOOOO Loadout: gangnam style/10

Gangnam style

Played about 3 hrs of Payload yesterday, and neither of the teams ever had any bot issues. Everyone on both sides were happy and doing some casual/funny stuff in honor of the occasion.

What is this

whats the point of this

Can you publish the game

Why does it always start with 0 tho?

ermm actually math.round doesn't work that way it only takes one number and then preforms banker's rounding (0.5 => round up) on the parameter and then returns it 🤓if you looked up how the round function worked you would've known that and roblox's autofill should've only given you one parameter 🤓🤓🤓🤓👆👆👆👆

you dont need to make an arg to make tf2 scary😭😭😭😭😭

just dont touch harvest, its crawling with those freaks