So you didn't write tests, and your code is a big enough mess that everyone else is afraid to make changes in case it breaks? Congratulations, you're a real developer now.
Here's one way you could account for transform: Make a copy of every pokemon that has ditto or mew's stats for HP and their own stats for everything else, and then make it so any pokemon that has transform can be replaced with the ditto equivalent of any of the pokemon on your team. Now, you'd have to do a lot of work for this, but it at least solves the main problems you mentioned
The thing is a pokemon could have many movesets and you won't be able to tell Also with the large amount of combinations, it would increase the complexity and slow it down
@@porkeman9007 that's true, but transform copies the opponent's moveset, so that simplifies it at least. Basically you'd count one ditto as being a random one of 6 pokemon, each of which is one of the ones on your team with the hp and held item of the ditto
That's a good idea, and it was my first strategy. That was back when the AI used machine learning so almost doubling the amount of inputs just to handle Ditto was far from ideal, but I might go back to something like this.
@@TheThirdBuild there might be some way you could make to handle cases like this plus form changes so that they're counted as variants of the same pokemon. I'm not sure if it would help the complexity though.
I really enjoyed this series, and that part about failing just as learning was inspiring. I'm a huge pokemon fan as well. Really hoping you come back someday!
Florida Man talks about a virtual Programm which is supposed to defeat grown men in a children’s game, but gets bugged out by some purple guy (I love your videos, I don’t know anything about programming but I like AIs really, I enjoy your content thanks for it :D)
"Learn how to code is learning how to make mistakes." So true... If I make a program, and it goes right on the first try, I get more suspicios of it than in any other situation... xD
Absolutely. If my code works right away on the first try, I check it again to find out, of I missed something. Like an old teacher of mine used to say: "There are two rules in software development and IT: a) It doesn't work and I have no idea why. b) It works and I have no idea why."
That edge case with Transform reminds me of this huge controversy in Gen 2, where there was the Sketch/Transform glitch, that would give Mew and Ditto learn any move like Smeargle. Then they brought up Metronome into the picture, and it was a big topic of discussion for legality in competitive. For example, Dragon Rage Ditto (for petit cup), Spore or Hi jump kick Mew. This AI edge case just brought back some nostalgia here. I was wondering how the AI would address that, and as I expected, an edge case would come up.
This is such a cool video. I have no experience with coding, but I was able to follow along with what you were saying. This whole thing is super impressive
I finally got to battle it, and it wasn't what I expected at all. It made decisions in an instant at first, but slowed down near the end, the opposite of what I'd thought would happen considering the number of variables being much higher at the beginning. It was exciting and I saved the replay, something I'll keep because this whole thing is awesome. I'd love to use the same team against it a few times, but I also don't want to hog it, so I'll try to only do it once per day while it's active. It disappointed me somewhat with it's skill level, but I'd bet it would destroy me with one more year of data. Geeking out over the project and that I contributed even a small amount of data.
I think it makes sense to be faster at the start, because all the Pokémon will always be in the same conditions (full hp, no hazards, no boost moves, full pps) so in a sense it's probably a state that is optimized already by the ai, but during the match there is always more to account for
Kinda similar to chess, in the opening you've already been these before so it's all familiar, but as the game goes on you venture into new territory. Should be interesting tho as like with Chess 960 an opening book is useless with random starting states (here being the variety of teams it'll be playing as)
Just now finding this series and went down the rabbit hole watching all of them at 4 in the morning. Truly so entertaining and amazing to watch the amount of effort put into this project. Having to work your way around a problem piece by piece, outcome by outcome is so interesting to me. Add Pokemon into it and suddenly its my dream hyperfixation!
So does your AI at least recognize the enemy is a disguised Dark-type when its Psychic fails on a "Poison" enemy? Because the last game I played, GameFreak's AI didn't.
@@johnkingsize Yeah, it has for a while, since gen 2, though not perfectly. Though the newest ai that's really shined was probably BW2 PWT trainers. I think Blue in fire red was also quite difficult and his ai was decently complex in comparison to other trainers in the game.
Hey man im sure you moved on to bigger and better things but please please record more. These videos are done so well that they seem timeless and i had no idea it was a dead account. I know this might not touch you the way id like it to but this waa truly inspiring and im going to start learning java script because this is cool :) i know nothing about computers and was never interested until this video. I thank you for the inspiration and guidance. Hope you have a good life my guy 😁
I don't think that's where I would draw the line for the "cursed" side of furries. If people want to look up Zoroark on e621, that's a lot better than some of the things I've seen when encountering furries (not that I have a problem with a majority of them; they tend to be chill people, if not just a little awkward. Some of them, though... at best, they're horrible people. At worst, they genuinely make me concerned for the safety of individuals around them.)
I brought a gimmick team and the AI just sat there, staring at me menacingly while the timer ticked down help (Natu / Shuckle / Purrugly / Wigglytuff / Lurantis / Audino)[Skill swap magic bounce sticky web nonsense gimmick]
One time when someone removed the choice scarf from my ditto, but they were running a full stall team, so ditto kept stealing their sets and was able to use them at full effectiveness.
14:40 So true. That describes the last 7 years of my life. :D I wrote some code 3 years ago. I thought that it was as good as it could be considering the tools I had. Turns out that it was many seconds slower than what I needed. I wrote some code a year ago. I thought that it was pretty ok, until it totally broke because some core functionality was outright impossible to implement. I wrote some code 5 months ago. It was slow and buggy. It took a week to clean up and two more to fix most bugs. So during the last 2 days, I wrote a solution which I thought was way better. Turns out that even though it only has half the features of the old code right now, it's about 30ms slower on average, which would add up soon. A friend of mine implemented an structure without any reference except for a short description. She thought that it was pretty good, until she tried to insert null and broke the universe.
@@sageinit It was both the evaluation- and the lookup-functions in an interpreter. The old ones used a lot of recursion and threw an exception to trigger tail-recursion. The new one replaced variables locally and used an infinite loop as a jump-point. (An evil word, I know :D) The test was creating a list of a lot of numbers, filtering for odd ones, incrementing each and summing them up and doing a comparison to 0. (A stupid test, but it's what I needed: a test which used the most common operations a lot of times). The implementation of each of those functions (filtering, mapping, folding and comparison) was the exact same. I ran the test 5000 times for each implementation. So it's not like the evaluation function took 30 ms, it was a tiny difference which added up over a million or so iterations. Your question about automating this might have been a joke, but I'm gonna be serious. When I write software at work, I have to get it to work quickly rather than get it to be fast. All except the first program I mentioned were compilers and interpreters I wrote in my free time, though. There, I am more interested in how to implement something rather than how to optimize it (at least until it's too slow to be enjoyable). This field is still new for me, after all. :)
Programming talents aside, this video is some of the best editing I've ever seen on a coding project from a technical analysis. Great flow, broken into distinct parts, with helpful visuals to boot. Excellent work.
Personally, I'd love to see a video covering and explaining (or, at least, your best guess at explaining) the most extreme/outrageous bugs you found and had to solve - like the one you mentioned where the AI thought the opponent had 15 moves. What the hell happened there? What went wrong, and how did you address it? I'd love to see your problem solving process for bizarre situations like that, I feel like that could make for an extremely entertaining video
Do you think you could turn on saving VODs for the twitch channel? I want to see the AI in action but I don't know if I'll always be able to catch it when I'm online
To become a professional in any field, you must realize how many of them are only better in handling situations due to more experience but many still do the same mistakes you make, they are just better in handling them... or cover them up or straight up ignoring them :D
Just found this channel today, but hopefully you come back. You’re so passionate and it makes every minute enjoyable even if I don’t understand some bits. Hope all is well
Dude, I love your work. This is the video which made me subscribe, because I love your detailed analysis on working with AI and the problems that can plague it.
This is crazy. So glad i had this pop up in my recommended. Best of luck. Im gonna keep watching this. Also id like to see some of the battles the ai goes through on these little exhibitions. It seems like it'd be a cool video in itself.
Excited to see the teams that the Ai makes. I love your video it’s so cool to think than an AI brought so much life to a community. I wish more poketubers would cover this
Would a viable way to solve the speed ties issue be having the AI assume it's always going to lose the speed tie and then account for the worst case scenario (logic being its choice there would probably be good in the best case scenario too)? Or would that end up causing other issues that I can't think of?
This could cause issues with speed ignoring moves being only what it uses or doing a move such as putting a faster Pokemon out that might be a technically a bad move. All because of a decently often event. So some issues and not as straight forward as we both thought.
It's been a while since I've subscribed to a channel this quickly after just seeing one video: you both have a very obvious passion for this project and and you're really good at explaining complicated stuff in an interesting and entertaining way! Plus, as someone who works in education, I love your attitude about treating losses as exciting learning opportunities!
I'm glad you brought up the test thing. It turns out its very important and you figured out why. On a professional level, we write down all the requirements, i.e. EVERYTHING we want a product to do, then we write tests for those requirements. At each step in the release process, alpha, beta, etc., we run these tests to verify that all the requirements are working as intended. Prototypes and products have different steps in between, but overall that's usually how its verified that it is functional as intended. So any new or inspiring programmers out there, take this as an example. Save yourself from headaches later. You dont have to test every single part as its developed, but a unit test to verify requirements will save a lot of pain.
"If we hit 333 subs" Looks down n see 23k subscribers. This was my first video of yours I've seen and you defo deserve this boost! Best of luck for any future vids and projects
Hey, so I stumbled across your videos while looking around for demos of creative AI programs (Aeva, Jukedeck, etc.), and I gotta say, this is just incredible. I conduct research into ethical problems with AI programs in the music industry while advocating for community based solutions, and I was happy to hear your reasoning for keeping the code for yourself. It looks like you’re at Florida Atlantic based on the logo in the background, and as a fellow alum, I just wanna say congrats on what you’ve accomplished with this program so far, and I hope you continue your work. I’m looking forward to more videos. Take care and good luck!
I am so excited to see this fully realized. I hope that it might revolutionize the way we can see the meta shifts, and be an example of how people could build teams, especially around certain pokemon
This is awesome! Gotta' admit that the "bug invasion" section almost scared me off, but the entire video was well worth it. Definitely gotta' remember to do unit testing during programing...ahaha Very cool project. Hope to see its progress!
I asked this on your other video but I am genuinly curious how your AI deals with predicting moves ahead when dealing with dmg ranges. Because dmg ranges can change the course of a whole battle and it seems to me calculating it can also create multiple branching paths depending of if you hit the range or not especially if you are looking multiple turns ahead.Not even mentioning moves that are not 100% acurate or the impact of lowering acceracy.
I was kinda secretly hoping this would be a video about how a super advanced ai realized that the bug type is sneakily busted as shit and proceed to completely uproot the competitive scene. This was aight as well tho props
Awesome work, both for the code and the videos! I really hope some day AI will be able to teambuild efficiently on its own, it would help us discovering new sets
So you didn't write tests, and your code is a big enough mess that everyone else is afraid to make changes in case it breaks?
Congratulations, you're a real developer now.
XD too true
smartest way to keep your job
I know, right?
So real.
This is now my favorite comment of all time
The only thing to do is restart.
"I didn't write a single test."
Right on par with every developer I have ever worked with.
95% of my tests are generated by a script I wrote. Does that count as writing tests?
Me who loves trying to break code to remove all bugs. 👁👄👁
When someone says "I broke a law of programming", I usually take it as "I did the same thing everyone else does wrong"
The only time I've written unit tests were in university and if the boss makes me, I never do in personal projects 🫢
it's insane how high quality these videos are, literally an instant-watch when it popped up for me. keep up the great work!
It was an instant like for me. Just 5 seconds in
Tbh me too, I subbed like the first few minutes of the first vid he made of this AI.
Yeah Id love to see his process for making a video
:(
"Today in the spiffing brit, we are gonna break an AI"
+
Pokemon competitive AI is perfectly balanced with no exploits whatsoever
brit should see this
Now that would absolutely be hilarious
This comment almost made me do a spit take...which would have been extra sad as I was drinking my Yorkshire tea at the time.
Smogon in 2022: "Due to the strategies discovered by FSAI, Hatterene, Galarian Stunfisk, and Jellicent are now banned to Ubers."
Coming from the future (Dec 2024) to say that was exactly what happened back in 2022
Bot: _Exists._
Ditto: "I'm gonna end this bot's whole career. "
Here's one way you could account for transform:
Make a copy of every pokemon that has ditto or mew's stats for HP and their own stats for everything else, and then make it so any pokemon that has transform can be replaced with the ditto equivalent of any of the pokemon on your team. Now, you'd have to do a lot of work for this, but it at least solves the main problems you mentioned
The thing is a pokemon could have many movesets and you won't be able to tell
Also with the large amount of combinations, it would increase the complexity and slow it down
@@porkeman9007 that's true, but transform copies the opponent's moveset, so that simplifies it at least. Basically you'd count one ditto as being a random one of 6 pokemon, each of which is one of the ones on your team with the hp and held item of the ditto
That's a good idea, and it was my first strategy. That was back when the AI used machine learning so almost doubling the amount of inputs just to handle Ditto was far from ideal, but I might go back to something like this.
@@TheThirdBuild there might be some way you could make to handle cases like this plus form changes so that they're counted as variants of the same pokemon. I'm not sure if it would help the complexity though.
@@r3io4mf I meant in how it handles it's knowledge of all pokemon
I really enjoyed this series, and that part about failing just as learning was inspiring. I'm a huge pokemon fan as well. Really hoping you come back someday!
Florida Man talks about a virtual Programm which is supposed to defeat grown men in a children’s game, but gets bugged out by some purple guy
(I love your videos, I don’t know anything about programming but I like AIs really, I enjoy your content thanks for it :D)
*gum
@@justafan9206 gum man talks about a virtual program
"Learn how to code is learning how to make mistakes."
So true... If I make a program, and it goes right on the first try, I get more suspicios of it than in any other situation... xD
Big mood.
*Tests a code*
*Code works*
“Oh shit did I screw up the test?”
Absolutely. If my code works right away on the first try, I check it again to find out, of I missed something.
Like an old teacher of mine used to say:
"There are two rules in software development and IT:
a) It doesn't work and I have no idea why.
b) It works and I have no idea why."
"I know why this doesn't work"
"I don't know why this doesn't work"
"I don't know why this works"
In order of ascending scariness.
Thanks for producing this video series! It's a pity projects like this tend not to survive.
When they said they had a team with it as well.
Jinx
That edge case with Transform reminds me of this huge controversy in Gen 2, where there was the Sketch/Transform glitch, that would give Mew and Ditto learn any move like Smeargle. Then they brought up Metronome into the picture, and it was a big topic of discussion for legality in competitive.
For example, Dragon Rage Ditto (for petit cup), Spore or Hi jump kick Mew.
This AI edge case just brought back some nostalgia here. I was wondering how the AI would address that, and as I expected, an edge case would come up.
I should learn how to abuse that glitch. Unfortunately bank wouldn't accept a ditto with anything other than transform, but it sure would be fun.
@@ArtisChronicles Basically: Ditto copies Smeargle. Smeargle has Sketch. Ditto uses Sketch. Ditto now knows the move it sketched.
I actually played against the AI and lost to it lmao. Amazing work and absolutely high quality videos bro, this is incredible! Keep it up
I keep missing chances to play it lol
Glad to see this project is still ongoing, I’ll definitely try playing a game with it in a bit
When you're doing unit testing, not getting error is a red flag. It could mean the test itself has error or the logic is flawed.
@@soniablanche5672 I think you replied to the wrong comment
"I didn't write a single test"-
*** the coders will remember this ***
I wish I could show this 9:47 to my parents 20 years ago who told me pokemon was a waste of time and I wouldn't learn any useful skills from it
This is such a cool video. I have no experience with coding, but I was able to follow along with what you were saying. This whole thing is super impressive
I finally got to battle it, and it wasn't what I expected at all. It made decisions in an instant at first, but slowed down near the end, the opposite of what I'd thought would happen considering the number of variables being much higher at the beginning. It was exciting and I saved the replay, something I'll keep because this whole thing is awesome. I'd love to use the same team against it a few times, but I also don't want to hog it, so I'll try to only do it once per day while it's active. It disappointed me somewhat with it's skill level, but I'd bet it would destroy me with one more year of data. Geeking out over the project and that I contributed even a small amount of data.
I think it makes sense to be faster at the start, because all the Pokémon will always be in the same conditions (full hp, no hazards, no boost moves, full pps) so in a sense it's probably a state that is optimized already by the ai, but during the match there is always more to account for
Kinda similar to chess, in the opening you've already been these before so it's all familiar, but as the game goes on you venture into new territory. Should be interesting tho as like with Chess 960 an opening book is useless with random starting states (here being the variety of teams it'll be playing as)
Just now finding this series and went down the rabbit hole watching all of them at 4 in the morning. Truly so entertaining and amazing to watch the amount of effort put into this project. Having to work your way around a problem piece by piece, outcome by outcome is so interesting to me. Add Pokemon into it and suddenly its my dream hyperfixation!
Wooo! The Master Trainer A.I. is back! And… they’re a bug-type user?
_Always have been_
So does your AI at least recognize the enemy is a disguised Dark-type when its Psychic fails on a "Poison" enemy? Because the last game I played, GameFreak's AI didn't.
I think gamefreaks ai cheats a little, they know what pokemon you are and they don't account for abilities
@@337-k5x meanwhile, Sonority's didn't do this as well.
** It doesn't affect Misdraevus **
That's just a use case of inverse damage calculation. "Okay, it did 0 damage, so that means the target is a dark-type...wait a minute."
Does the game's AI take the second type into account now, at least?
@@johnkingsize Yeah, it has for a while, since gen 2, though not perfectly. Though the newest ai that's really shined was probably BW2 PWT trainers. I think Blue in fire red was also quite difficult and his ai was decently complex in comparison to other trainers in the game.
kinda sad that you stopped posting an year ago, I hope you'll come back with your usual content!
Hey man im sure you moved on to bigger and better things but please please record more. These videos are done so well that they seem timeless and i had no idea it was a dead account. I know this might not touch you the way id like it to but this waa truly inspiring and im going to start learning java script because this is cool :) i know nothing about computers and was never interested until this video. I thank you for the inspiration and guidance. Hope you have a good life my guy 😁
Well said.
Usually my coding videos and my entertainment video's are two separate categories. You managed to blend them extraordinarily well.
You hesitated when calling Zorark Popular. I'm just going to assume your subtly calling out the more cursed side of furries.
Probably
Definitely.
We do be Cursed XD
@@kuroryuken it is more fun that way.
I don't think that's where I would draw the line for the "cursed" side of furries. If people want to look up Zoroark on e621, that's a lot better than some of the things I've seen when encountering furries (not that I have a problem with a majority of them; they tend to be chill people, if not just a little awkward. Some of them, though... at best, they're horrible people. At worst, they genuinely make me concerned for the safety of individuals around them.)
I brought a gimmick team and the AI just sat there, staring at me menacingly while the timer ticked down
help
(Natu / Shuckle / Purrugly / Wigglytuff / Lurantis / Audino)[Skill swap magic bounce sticky web nonsense gimmick]
Dude, keep rocking it. I love that you're letting us see into your learning journey.
You are going to blow up dude. The algorithm just hasn't found you yet. Great content, great presentation. Keep doing what you are doing.
One time when someone removed the choice scarf from my ditto, but they were running a full stall team, so ditto kept stealing their sets and was able to use them at full effectiveness.
yeah itemless ditto is actually super useful vs stall
14:40 So true. That describes the last 7 years of my life. :D
I wrote some code 3 years ago. I thought that it was as good as it could be considering the tools I had. Turns out that it was many seconds slower than what I needed.
I wrote some code a year ago. I thought that it was pretty ok, until it totally broke because some core functionality was outright impossible to implement.
I wrote some code 5 months ago. It was slow and buggy. It took a week to clean up and two more to fix most bugs.
So during the last 2 days, I wrote a solution which I thought was way better. Turns out that even though it only has half the features of the old code right now, it's about 30ms slower on average, which would add up soon.
A friend of mine implemented an structure without any reference except for a short description. She thought that it was pretty good, until she tried to insert null and broke the universe.
What the bloody hell are you trying to optimize and why haven't you yet automated the optimization cycles
@@sageinit
It was both the evaluation- and the lookup-functions in an interpreter. The old ones used a lot of recursion and threw an exception to trigger tail-recursion. The new one replaced variables locally and used an infinite loop as a jump-point. (An evil word, I know :D)
The test was creating a list of a lot of numbers, filtering for odd ones, incrementing each and summing them up and doing a comparison to 0. (A stupid test, but it's what I needed: a test which used the most common operations a lot of times). The implementation of each of those functions (filtering, mapping, folding and comparison) was the exact same. I ran the test 5000 times for each implementation.
So it's not like the evaluation function took 30 ms, it was a tiny difference which added up over a million or so iterations.
Your question about automating this might have been a joke, but I'm gonna be serious. When I write software at work, I have to get it to work quickly rather than get it to be fast. All except the first program I mentioned were compilers and interpreters I wrote in my free time, though. There, I am more interested in how to implement something rather than how to optimize it (at least until it's too slow to be enjoyable). This field is still new for me, after all. :)
Programming talents aside, this video is some of the best editing I've ever seen on a coding project from a technical analysis. Great flow, broken into distinct parts, with helpful visuals to boot. Excellent work.
dude I love your outlook on this project so much and the world by extension, keep up the great work
Personally, I'd love to see a video covering and explaining (or, at least, your best guess at explaining) the most extreme/outrageous bugs you found and had to solve - like the one you mentioned where the AI thought the opponent had 15 moves. What the hell happened there? What went wrong, and how did you address it? I'd love to see your problem solving process for bizarre situations like that, I feel like that could make for an extremely entertaining video
Do you think you could turn on saving VODs for the twitch channel? I want to see the AI in action but I don't know if I'll always be able to catch it when I'm online
I second this statement!
i think its online 24/7 for a week
Yep!
To become a professional in any field, you must realize how many of them are only better in handling situations due to more experience but many still do the same mistakes you make, they are just better in handling them... or cover them up or straight up ignoring them :D
your video production value is so insanely high
do you have any experience outside of just casual projects?
Thanks, and there's a link in the channel's about page regarding the stuff I do
Just found this channel today, but hopefully you come back. You’re so passionate and it makes every minute enjoyable even if I don’t understand some bits. Hope all is well
Dude, I love your work. This is the video which made me subscribe, because I love your detailed analysis on working with AI and the problems that can plague it.
This is crazy. So glad i had this pop up in my recommended. Best of luck. Im gonna keep watching this. Also id like to see some of the battles the ai goes through on these little exhibitions. It seems like it'd be a cool video in itself.
Loved your first video man, keep it up!
The quality of these videos blows me away. Can't wait to see this 'series' trending
I'm sad that the series never continued.
Keep working on this thing, it’s awesome!
Excited to see the teams that the Ai makes. I love your video it’s so cool to think than an AI brought so much life to a community. I wish more poketubers would cover this
Very inspirational words at the end, loving the series! Shout out for the dope VFX!!
I just watched all the videos on this project today, so sad to have missed the exhibition by a week. Subscribed to get the next exhibition
WOOOOOO NEXT PART I LOVE THESE VIDEOS keep it up man
The quality of a channel this small is amazing
The most ferocious foe: A Purple Blob
you mean your pokemon wearing a scarf or holding a balloon
“the other one is super… popular” an e621 callout i see
8:31 my man, that hesitation made me laugh way harder than it should have
Would a viable way to solve the speed ties issue be having the AI assume it's always going to lose the speed tie and then account for the worst case scenario (logic being its choice there would probably be good in the best case scenario too)? Or would that end up causing other issues that I can't think of?
This could cause issues with speed ignoring moves being only what it uses or doing a move such as putting a faster Pokemon out that might be a technically a bad move. All because of a decently often event. So some issues and not as straight forward as we both thought.
the intro cracked me up already, awesome xD
Great video! Really good message at the end too, can’t wait to see what comes next (:
When I found your channel, I was so surprised how low your subcount was... awesome stuff! Definitely worth my sub!
It's been a while since I've subscribed to a channel this quickly after just seeing one video: you both have a very obvious passion for this project and and you're really good at explaining complicated stuff in an interesting and entertaining way!
Plus, as someone who works in education, I love your attitude about treating losses as exciting learning opportunities!
Good to know you're doing alright and progressing forward. I'm gonna drop by and have a match for sure.
I'm glad you brought up the test thing.
It turns out its very important and you figured out why.
On a professional level, we write down all the requirements, i.e. EVERYTHING we want a product to do, then we write tests for those requirements.
At each step in the release process, alpha, beta, etc., we run these tests to verify that all the requirements are working as intended.
Prototypes and products have different steps in between, but overall that's usually how its verified that it is functional as intended.
So any new or inspiring programmers out there, take this as an example. Save yourself from headaches later.
You dont have to test every single part as its developed, but a unit test to verify requirements will save a lot of pain.
The visuals on this video are amazing! Good job on the excellent presentation and explanations!
"If we hit 333 subs"
Looks down n see 23k subscribers.
This was my first video of yours I've seen and you defo deserve this boost! Best of luck for any future vids and projects
Hey, so I stumbled across your videos while looking around for demos of creative AI programs (Aeva, Jukedeck, etc.), and I gotta say, this is just incredible. I conduct research into ethical problems with AI programs in the music industry while advocating for community based solutions, and I was happy to hear your reasoning for keeping the code for yourself. It looks like you’re at Florida Atlantic based on the logo in the background, and as a fellow alum, I just wanna say congrats on what you’ve accomplished with this program so far, and I hope you continue your work. I’m looking forward to more videos. Take care and good luck!
Ah yes, the test in production approach. Love what you’re doing and subbed to keep track
Take your time, but I'm looking forward to the next vid! No matter the time it takes i'll be ready and excited to watch!
This is seriously my favorite series on UA-cam
This is such a informative video. Very charming as well. Keep it up!
Loved the video. hope to see more
"I didnt wirte a single test"
*laughs*
"So what does this mean?"
It means you're normal buddy, welcome to the club
This is so cool, please keep making these videos!!
my man drops 3 brilliant videos in 2 months and then spends the next 9 working on the next one, absolute legend tbh
I'm curious how it would handle Zoroark mapping Smeargle or Ditto.
I LOVE THIS SERIES!!! makes me want to learn how to code. :) Also you seem very friendly
this has quickly become one of my favorite channels.
12:10 don’t worry, in my experience, programmers don’t write comments
Truly an underrated comment.
Really underrated channel. Great video I appreciate the time spent on this. Hope you keep up the project I'll be keeping an eye out for more on it.
Your growth has been insane, congrats! Keep up the good work
Wow what an intro.. You such a fantastic content creator! Love this 🔥
Will there be more videos?
I am so excited to see this fully realized. I hope that it might revolutionize the way we can see the meta shifts, and be an example of how people could build teams, especially around certain pokemon
Okay so what I've learned is to beat this AI I will use a team of Smeargle, Zoroark, Aegislash, Ditto, Mew and Clefable with Metronome
Glad to see my favorite blob not getting pushed to the side every once in a while! Also great video!
You got a bright future ahead of you kid. Keep up the progress
The Graphs and visuals with the way you describe the coding process are great
Nice job with the video editing, I can tell that you are really improving!
You deserved it bro all the subscription, I am learning programming and love pokemon, so your channel is perfect 😂
Neat analysis video! Thanks for uploading!
Your work is insane. I love to watch these videos, it's an instant click for me. Please keep it up, you are crazy
You're a legend. I can't imagine how good the AI will get if you continue on with it.
Wow, extremely well articulated explanations! Great video editing too. Super engaging :D
Keep it up dude, you’re killing it.
Super orginal video, very informative and great graphics! You've earned another subscriber.
This is awesome! Gotta' admit that the "bug invasion" section almost scared me off, but the entire video was well worth it.
Definitely gotta' remember to do unit testing during programing...ahaha
Very cool project. Hope to see its progress!
Ayo new video when
You're awesome dude! Keep rocking on!
This series is the coolest thing! Just subbed!
I asked this on your other video but I am genuinly curious how your AI deals with predicting moves ahead when dealing with dmg ranges. Because dmg ranges can change the course of a whole battle and it seems to me calculating it can also create multiple branching paths depending of if you hit the range or not especially if you are looking multiple turns ahead.Not even mentioning moves that are not 100% acurate or the impact of lowering acceracy.
Eyyy was waiting for a new video in this series to drop, I'm already hooked to the project
I was kinda secretly hoping this would be a video about how a super advanced ai realized that the bug type is sneakily busted as shit and proceed to completely uproot the competitive scene.
This was aight as well tho props
can’t wait to see how this project evolves, and what you end up doing next
XD
Two of my Legitimate favourite Pokémon to use. Wonder how the AI will react when it has a Zoroark Illusion as a Ditto
One time zoroark messed with my head in doubles when I was playing in shield. The opponent sent out 2 tentacruels.
Awesome work, both for the code and the videos! I really hope some day AI will be able to teambuild efficiently on its own, it would help us discovering new sets
It's not the content that made your subscriber count go up. It's you, and the quality of these videos!