So basically the red cars have totally random parameters and the other types of cars (yellow, cyan and navy) copy the parameters of the car that has gone the furthest yet and change the values of those parameters up or down by a notch (10, 1 and 0.1 %).
If you wanna learn more about how to create the neurons network, I greatly advise the video by 3blue1brown on the subject.
Good Luck !!!
well most people don't take any stock in using randomness in machine learning. however although im not sure about this if you generate a random seed at the beginning. the computer generates a predestined series of random numbers at run time as long as it has enough memory to remember all the numbers.
randomize numbers 1 3 5 7 12 21 28 39 47 52
all those numbers will be run through regardless of changing conditions
randomize every second
random number 1
randomize every second
random number 3
you could argue that it doesn't make a difference whether random generation happens at the beginning or constantly because the method for random generation is the same. however I believe randomizing only once can never factor in changing conditions if the algorithm isn't designed to constantly change the seed.
I was actually thinking of donating to some games that use machine learning in their games as a way of buying games that are fun but also could potentially improve society beyond just being fun. I couldn't find any that are based in letting the computer make random decisions.
information learned from machine learning is saved to the harddrive and is swappable between computers. many would argue that the computer is just the files you download to the harddrive. I would argue its the processor itself. I believe that computers only do the instructions theyre told to do. however I still think its an open question whether a computer can learn the way people with memory loss learn. for example when people have alztheimers they can still talk and maybe even remember some old skills but theyve still lost their memory. essentially by repeating a task enough the robot could get an understanding of how to do it without thinking about it assuming that the program is still in memory and that the computer hasn't been shut off. after that the computer has lost everything. again I don't know where this memory would be stored and I'm also contradicting myself because this requires that the robot is doing things that it isnt explictly being told to do. I don't think anyone really knows what a computer is aware of when a person uses it. I know it has no control over the operating system. computers aren't sentient. they can't initiate anything. that doesn't mean there isn't something alive in there. computers have gotten to the point where some of them have either surpassed our intelligence or gotten close and even if they hadn't we can write programs that simplify human language so that a robot wouldn't have to exert all its brain power to make use of it.
