Non, because I don’t like artifical intelligence and I don’t understand it much. Would be nice if you can make maybe a guide for how it actually works in a way that you are explaining it to a 9 year.
@MYOriginsWorkshop and @Wyv_n you two should consider arguing in the DM’s as if you guys argue publically then it’s pretty bad for your reputation.
Non, because I don’t like artifical intelligence and I don’t understand it much. Would be nice if you can make maybe a guide for how it actually works in a way that you are explaining it to a 9 year.
Artificial Intelligence is really broad term though. Although it can cover machine learning and deep learning, it is mainly concerned about mimicking human intelligence. You can actually count pathfinding as AI actually.
@MYOriginsWorkshop and @Wyv_n you two should consider arguing in the DM’s as if you guys argue publically then it’s pretty bad for your reputation.
Meh. He decided to bring it to the public. I thought it would be much more appropriate for me to move it to DM even though he started first. But I thought it would a great idea to leave it public to increase the “reply” count and so more people will join us in and see this post more.
I know about it’s literal definition yet don’t know how the machine actually learns. I do understand using fitness scoring to see which machine did better yet actually making it “learn” is a challenge.
Yes, I do know about that. Almost every algorithm can be considered AI. I have also developed some pathfinding algorithms on roblox.
That’s kind of down bad of you ngl… But whatever, I don’t want to argue about this.
I know about it’s literal definition yet don’t know how the machine actually learns. I do understand using fitness scoring to see which machine did better yet actually making it “learn” is a challenge.
I guess let’s explain it in more human term then.
So let’s say you have a cannon that shoots a ball upwards and lands at the exact same spot. From here, you want to learn to catch it perfectly. When the cannon shoots the ball, you will try to “predict” where the ball will land. Unfortunately, for your first time, you didn’t catch it because your hand is way too far away from the ball. The distance between the ball and your hand is what they call an “error”. Now you learn to have your hand to be closer to that location by large amount. So large “error”.
The next time the ball shoot at the same location, your hand nearly catches it, and hence lower “error”. So you move slightly closer. You will repeat these steps again until one of these happens:
a) You can’t improve your location, but you still able to catch it, just not perfectly.
b) You finally able to catch the ball.
Then, you use that location as your “prediction” where the ball lands.
That’s just how the machine learning and deep learning works…
Get it?
Your making it sound like roblox is a platform where only developers with credentials should be able to create. I hope you remember this platform is target towards kids, and for them to learn, not to be perfect.
No it won’t this whole platform is meant so people learn, yeah someones going to make a mistakes but thats what the devforum is for, to help each other.
This whole post just seems like a way to bump your library, and there for I actually see no reason for this post… I mean you yourself said the post was biased so there really isn’t an accurate comparison made in this post…
A bit, one question is how can you apply it into something like obstacle avoiding orrr something like sword fighting.
Your making it sound like roblox is a platform where only developers with credentials should be able to create. I hope you remember this platform is target towards kids, and for them to learn, not to be perfect.
I didn’t say the roblox games themselves should be developed by people who has credential, did I? I was specifically referring to the machine learning and deep learning libraries which would be used by researchers, which could lead problematic results if the library developers have no such credential in place. Do you want researchers to use Roblox as research grounds so that we have better resources? If yes, then these kind of stuff need to be held at high standard.
No it won’t this whole platform is meant so people learn, yeah someones going to make a mistakes but thats what the devforum is for, to help each other.
Oh, people are learning! Learning the existence of other people’s libraries if they aren’t aware of them yet.
This whole post just seems like a way to bump your library, and there for I actually see no reason for this post… I mean you yourself said the post was biased so there really isn’t an accurate comparison made in this post…
Then feel free to ignore it then. Let it slide into the abyss.
Almost same thing here. Although, this is getting into another territory called “reinforcement learning”.
obstacle avoiding
Let’s say we’re playing a game where we can only dash forward, left and right. And we have three tracks, where the trains can move towards us. We have left, middle and right. You have to repeat the game when you get hit by the train. The longer we survive, the better high score (rewards) we get. Crash? Then hurt (heavy punishment).
Now let’s say we are spawning in the middle track, and a train is coming right towards us at the middle track as well. We have two issues here:
Determining the distance between us and the train before we can switch tracks.
Determining the direction which to move.
Since this is your first time in your game, you can expect that you don’t know the train’s speed and what could get you killed. This is what happens when we create an untrained model a.k.a. “We don’t know anything about this game” phase.
First time you play, you didn’t react fast enough when the train crashes you. You know that the crashing hurts you, so you try to avoid it. However, because you aren’t used to the train’s speed, you crash against it every time due to low reaction time. The more you crash, however, you start to be able to “predict” when to move since you can now determine the speed.
Eventually, you avoid the first train. Unfortunately for you, that’s not where the game ends. So few more trains incoming! Now let’s say we have two trains are coming now. This gives us an issue:
Determine the sequence to avoid the two train.
Determine the timing when to avoid for each train.
At this point, you crash against them multiple times, but somehow, you started to learn a “pattern” to avoid particular situations. Sometimes you avoid, sometimes you crash. Eventually you get a point where the number of crashes begin to decrease and you get better high score.
That’s pretty much it!
I see, but how do you actually “predict” the direction to move in based off the previous generation?
For this part, it’s either they learn by chance or trial-and-error.
For chance, they might have gotten lucky that they did not crash into a train (i.e. randomly going left or right"). So, the model “learns” that particular sequence of movements leads to rewards.
For trial and error part, it might keep on crashing multiple times, leading to the model to less likely to choose that particular action. Then when the probability to choose that action is low enough, the model starts considering to choose other actions, which could potentially lead to point 1.
I see, thanks.
Remove this. You said “unknown” yourself. That’s just unfair and not meant aa a Ressource but advertisement. At least ask them per dm.
Another one went thinking that I am doing advertisement…
Here’s a short message from frames:
Your making it sound like roblox is a platform where only developers with credentials should be able to create. I hope you remember this platform is target towards kids, and for them to learn, not to be perfect.
My message:
I didn’t say the roblox games themselves should be developed by people who has credential, did I? I was specifically referring to the machine learning and deep learning libraries which would be used by researchers, which could lead problematic results if the library developers have no such credential in place. Do you want researchers to use Roblox as research grounds so that we have better resources? If yes, then these kind of stuff need to be held at high standard.
basically a promo for his data predict module i do not see a resource anywhere in this post 
Read the damn other replies BEFORE accusing me of promoting.
Consider this: Platforms like Unity and Godot are already being used in research because their contributors often have strong formal backgrounds, such as PhDs in machine learning or related fields. The resources such as Unity ML Agents and Godot RL agents shows that!
Do you want the reputation of the machine learning (or more generally, research) for Roblox as a whole be damaged as a result of incorrect results due to potential incorrect implementation due to informal backgrounds?
Hopefully your comment stays up there so it doesn’t age very well.
At the same time, if I really wanted to promote my own stuff, why did I give the visibility to the other libraries which could reduce the effects of my non existent promotions? Wouldn’t I just highlight my libraries only?
The big difference is, that this is not my message:Your making it sound like roblox is a platform where only developers with credentials should be able to create. I hope you remember this platform is target towards kids, and for them to learn, not to be perfect.
This is promotion. You have columns where you would only have information about yourself that has little to do with the topic, but looks good for you.