Can AI to Predict Human Randomness?

[@KiddyKene]

All code is on my discord: discord.gg/EFNmDnCgK7 Thx for watching

[@chro.]

@@KiddyKene you should make it a website so you can get more data.

[@RichConnerGMN]

or just, you know, let people take it without forcing them to join your discord for no good reason

[@KrishDaEpik]

@@RichConnerGMNwhat is he supposed to do? upload the code on youtube's site? 😭

[@innxir]

@@KrishDaEpik github exists ._.

[@renlog7755]

@@KrishDaEpik github exists,

[@kvasir1]

ai trying to guess your number: ❌ trying to guess the ai number: ✔

[@superproxocz]

@@kvasir1 gucci

[@Siwdvi]

Just put 69 over and over

[@Lantanoid-05]

​@@Siwdvi why 69🤨

[@Fifasher2K]

​@@Lantanoid-05 Because yes.

[@snorman1911]

Or 96, or 42.

[@helonoidea]

but you don't need to read my mind, i always go rock.

[@eduhund21]

@@helonoidea I will pick scissors EZ

[@krielo-v5t]

I always take rock to

[@asheep7797]

@@eduhund21i will pick paper. damn, 3-way tie.

[@ajdndbdjbdj]

Solid choice

[@Paradoxical-n4u]

/

[@GarciaGamingFR]

i made this but heh... it always works... code: while True: num = input("") print("Bot guessed correctly!") print("Bot number: " + num)

[@spamton8202]

cheater

[@ibrahimali3192]

cheater

[@GoldbergToastyBred]

nice "AI"

[@GarciaGamingFR]

@@GoldbergToastyBred thanks

[@T3rr4T3rr0r]

Damn insane ass AI

[@BurrritoYT]

I have no coding knowledge and even I was screaming “OVERFITTING” in my head

[@notajalapeno4442]

yup that is what happen he should have also ran the tests over the data from the discord

[@rishibellam738]

omg its burrito

[@palmberry5576]

Yeah, they accidentally turned the test data into the training data

[@jeremy5819]

I literally yelled this outloud when I saw the part where he said he made a bot to alter his hyperparameters to improve the win rate. Amazing idea but textbook overfitting lol (still impressive regardless tho lol)

[@arskiz]

Noob

[@SpaynjXD]

I love where this channel has gone. Slightly different content, but still the rock hard core principle of "Mental torture endured through funny memes in order to accomplish something mindlessly useless"

[@puppet9577]

43 likes or something and know repl let me fix that

[@SpaynjXD]

@@puppet9577 Thanks 👍

[@MetroTO.]

@@puppet9577 no relpy*

[@alphagama-fb6yn]

@@MetroTO. no reply*

[@MetroTO.]

@@alphagama-fb6yn no relelpy*

[@9qgaga]

Bro's swearing increased as his sanity decreased

[@CaptainTom4862]

13:02 LOOK GARY THERE I AM! I'M THAT LIFELESS VIRGIN!

[@amongus_pvp]

Oh no it's captain tom

[@lightyisreal]

W

[@HeroBattlesB]

oh cool your profile pictures from amphibia

[@creeperboy1013]

13:43 WOW LOOK AT THAT, ANOTHER LIFELESS VIRGIN :D

[@villss894]

WOW RARE SIGHTING GUYS!

[@alansmithee419]

23:55 "You're either gay for me or you have cancer what is it?" "I'm going to send my entire bod over to you and I want you to play with it." Well that answers that question.

[@Kwauhn.]

4:57 It's okay to gloat about achieving a basically functional GUI. We know your pain.

[@miguelcardoso5084]

you don't need to top all your past acomplishments each video, if you just make something interesting/cool that's always gonna be enough to please the masses

[@emmah1408]

Fr

[@w花b]

But that's a recipe for complacency on a personal level

[@franklydoodle350]

@@w花b You're both right. There is a balance with everything.

[@LucinhooFiles]

Oour pfp is nearly the same

[@markusfassbinder8275]

11:04 I never use those. NUM PAD SUPREMACY

[@jonaskarlsson5901]

if you type correctly aka resting your hands on the letter keys the distance to the top row number keys are actually closer than to the numpad

[@cocomonkilla]

​@@jonaskarlsson5901The numbers are closer to each other on the numpad and typing in longer numbers is much easier and more familiar of a motion

[@fa113n_l3af]

I’ve alway noticed that when I play against other people in rock paper scissors, (usually) the next turn they play is what beats their previous move. For example, someone plays rock and then plays paper. Paper beats rock.

[@greenberrygk]

@@fa113n_l3af also for some reason everyone starts with scissors paper rock

[@Tek]

That ending is crazy

[@remyrauschenberger1603]

At 1:00 i LITTERALLY picked 73 because ive watched that viratasium video (which u didnt give enough credit toooooo) and im so glad you did point it out

[@MasamuneX]

this lesson is on overfitting lmaoooo

[@ARMMilad]

How did you add the search btw 🤔

[@steampunkWizardStudios]

As a fellow programmer this is some of the most hack code I have ever seen but if it works it works and it works insane. I never would've thought AI could be used for this.

[@Crates-Media]

Don't do that. Please don't praise or endorse hacks. The world doesn't need any more programmers suggesting "if it works, it's good enough". I've spent an incalculable portion of my life inspecting, unpacking, rewiring and generally unfucking code written or cosigned by those types. If it works but isn't well-conceived, clear in premise, and elegant enough to parse by the most junior guy on the team, it's worse than useless.

[@Jiglias]

@@Crates-Media obviously looking at just the snippets of code in the video you can tell it's ass, but worse than useless is definitely an exaggeration, especially for something like this which will probably never be seen by any other person

[@jonaskarlsson5901]

@@Jiglias everyone that sees any code on the internet says that the code is bad no matter what. That's just psychology.

[@jus6464]

10:46 I fell asleep with KZbin on. I don’t know this channel. I know that this woke me up

[@Sandman-xs8ui]

😂😂😂😂😂😂

[@CryptiCryptoAIM]

Try a neural network ai and see if it is better (without ANY special tricks, like pattern recognising and let the ai work for it self)

[@CryptiCryptoAIM]

I know that this will require a lot of testing data, so maybe setup a website where we can go an try

[@CryptiCryptoAIM]

Also make sure to create a second network, that predicts if the number is predictable or just a randomness app or smth. like that.

[@liam8370]

i agree there where two problems with this video first they were using psychology to guess the answers which will only work as well as u know the subject instead they should've chosen neural network for this like u suggested secondly there was no available data for this project if they had gone with chess there would be tons of available data plus the already available chess bots and not to mention it would be more meaningful

[@dabbopabblo]

@@liam8370 That would be stupid. This was a video on predicting randomness, not predicting strategy. Chess is not a game played randomly, its played following a rule set that guides a players decisions.

[@liam8370]

@@dabbopabblo you'd be surprised to find out how random we become under pressure

[@Shad0wWarr10r]

Time to find correlation between a persons 100 first random numbers and racism

[@goodboi6540]

0:39 sciencephile the ai jumpscare

[@HalalaBrawlStars]

2:06. These numbers add up to 90%, so I wouldn’t take these too seriously, although rock is still the most.

[@blankqwerty1234]

The last 10% were people chosing random shit like "gun" or "nuke"

[@tperm7282]

Storing the raw data in the code is ATROCIOUS

[@poi-cd8yo]

the video captures the cuteness of people so well, those jokes, the tone friends talk with each other...

[@ninebreaker274]

go outside

[@НикитаЧумак-т2ч]

this guy is totally not an alien

[@brightblackhole2442]

it's impressive how the humans can be so friendly with each other in groups. they alter their speech patterns and mannerisms according to the human they are talking to. the people are cute and squishy like ants

[@Simigema]

“If you’re out of ideas in Minecraft, make a creature that self replicates and evolves, spreading around”

[@EidRos]

@@Simigema that's kinda cool

[@sphereguanzon2217]

I did. Its pretty cool tho

[@astral6749]

This is the kind of data science I like to see. Not too complicated and not too practical.

[@XanderYTV]

24:16 A Will Tennyson reference in my nerd side of youtube was not what i was expecting

[@Low_qualityKel]

every time he starts explaining what hes doing i think hes about to go "But instead of doing all of that-"

[@WoolyCow]

i havent seen any of your previous videos, but honestly this vid slapped! whatever stuff you keep making im down for it :D dropped a sub

[@nokts3823]

Cool video, it was a really fun watch. This looks like the kind of thing an RNN would be good at. Btw careful when you tweak your hyperparameters based on the accuracy! You are inadvertently training on the test set and the model probably won't generalize all that well. You could split your test set again and use one half for hyperparameter tuning and the save the other for the final testing.

[@I_Am_Transcendentem]

Kiddy Kene just shattered our illusion of free will

[@electricportals3644]

18:39 so what you are saying is that you made an AI program your AI better?

[@Gamicaleye]

SCIENCEPHILE THE AI!! 0:41

[@culemp4657]

Kudos to you for making this thing work. I am kind of sad that there aren't a lot of views on this vid. You clearly deserve a break for at least a MONTH or maybe even a YEAR!

[@AllExistence]

Number one rule for any input type experiment: make it save data.

[@Gnulcho]

Actually helped me understanding prediction algorithms a bit better. I initially thought that you would be going for text prediction, which would just be to assign the numbers to one specific word each, with everything else almost not changing at all. Then instead of bland numbers, you write stories. I did something like that about 15 years ago, when I was still learning to code. Although it took whole messages as tokens, instead of words or numbers. Not having any sane practices or experience, I ended up storing the memory info in a simple file folder structure. The file names were the input values, and the contents just had the user-generated answers to that input. Each of the answers there did get an own file of course. Another rather big difference to note is, that as a result of the scuffed memory storage, there wasn't really any referring to anything else than the last message. As a result, dialogue usually looked like this: USR: Hey BOT: Hey, back USR: How are you? BOT: Good USR: Good to hear, I'm fine too *BOT: Why would you care? You're a bot. ... *no answer recorded for the previous statement, so one at random is chosen instead

[@cleminition]

that goddamn lobotomy sound is like an instant headache glitch. i hear 1 second of it and i instantly get a headache, so now i have a headache thanks for that

[@felipebr13pr44]

i just realized the name of the progamn when it was 30:37 lmao- man, just sad that it isnt public atleast i think, i wish i could test that, probally some more people too, sounds cool, like maybe on a link on website that is the normal guessing progamn.

[@tookiecar1]

Program*

[@pratyushplayer0840]

N=int(input("enter a no.")) print ("the number you were thinking of",N)

[@cactumNOGD]

HE IS A GENIUS

[@ginqus]

you broke, like, at least 4 PEP8 rules here

[@pratyushplayer0840]

@@ginqus ?

[@ginqus]

@@pratyushplayer0840 Just google what PEP8 is. In short, it's a set of rules to make your python code look good and readable. You broke these: - Variable shouldn't be a single letter - Variable should be in camel_case - You should put spaces around "=" - Space after "print" is not needed - You should put space after ","

[@pratyushplayer0840]

@WoidDev bruh idk but works for me

[@scanzas]

This is the first video I've watched from your channel, and honestly, I would have thought it was a coding channel, not Minecraft. If that says anything, you're doing an amazing job with this type of content. I'd love to see more like it! 😉

[@Sowian69420]

15:09 sigma first and second digit is wild

[@protoncracker_xp]

Got your code, did 22.395%, highest on runs after fixing the AI algorithm. It's a nice code, and I believe it can increase for up to 50% or maybe 80% (spec). I know none of you actually care for what I'm about to say, but: The concept of improving AI performance to predict human randomness can be enhanced by refining the way data is tested and individual psyche can be understood by AI, with data being processed and applied to *individual* users. The creator's (Kiddy... it's weird calling him that) approach already showcases several common AI methods, such as data collection through his Discord community and prediction optimization through algorithms. However, further improvement can be realized by focusing on multiprocessing, individual-user weights, and better data optimization as I see it fit. The first thing I think is leveraging multiprocessing, so the AI can handle larger datasets and execute predictions faster. Parallel processing of data points allows for faster prediction cycles. Multi-threading could be applied to enable different layers of predictions simultaneously. And I don't mean both of them only on single-level, but on multi-level processing - Based on weights, you can have a kind of generational AI running that only executes once and has a few children that try to interpret the user's inputs and lookout for patterns of psychological awareness of a dataset. Any human can, arbitrarily, know that it's following a pattern and then choose another, this takes your current AI too much time to refigure it out and, because of that, increase the cost of the winrate. As seen in the video, the AI started learning patterns from users, but when human randomness increased, predictability dropped. By assigning weights to individual users based on past behavior (such as typing speed or input timing) and analysis of mass-data-effect, the AI could better distinguish user-specific patterns. Updating weights dynamically as the user continues to input data could also help the AI fine-tune predictions. Isn't like you have a GPT running there, it's lightwork, so enhancing the individual weights using a base as a starting point would be ideal. Also, a playground for test and data-collection, along with data effectiveness (weight applied to pattern, data pattern compression) could be ran by your subscribers to enhance the model. I doubt we could get to 80% on EVERYONE, but 20 to 50% is quite possible and 80% might not be a dream so distant.

[@protoncracker_xp]

Oh, and, nice video, keep it up.

[@sussdood]

12:35 gee, that suss guy that appears for a few frames seems really handsome

[@PLANT7-qy3iv]

Indeed

[@TVmommyUwU_YT]

1:38 Is no one gonna talk about "cuming up" text

[@LeFakeOfficiel]

I thought was the only one who saw it xd

[@ilomaniac1933]

0:50 that was my number you scared the hell outta me

[@Aqua2D]

sameee

[@SANSTERGANGSTER]

19:53 GİORNOS THEME GOES HARD🗣🗣🔥🗣🔥🗣🔥🗣🔥🗣

[@cheetosnour.scratch-learn]

get the number predictor thing to predict floats (ex. 8.561745) and integers (ex. 9856)

[@sanstheskele-bro6102]

I felt personally attacked when you said monkey and zoomed into 37, tge exact number i was thinking of

[@nithinbhandari3075]

1:09 I cannot believe my eyes, around 30 second ago i thought 73. Mathematicians are genius.

[@owenmacpherson5311]

it was a lot of stats…not them just knowin

[@corrupt_world2471]

Michael reeves enjoyer you must be

[@snibo1024]

At the ending I got surprised that it was you I have already seen your Minecraft videos bit I haven't recognize you and I totally enjoyed the entirety of keep up with that man

[@tachrayonic2982]

I wonder if a compression algorithm would be good for predicting patterns like this. One way of doing a compression algorithm is to take any repeated group of characters, and recursively collapsing them into single characters in a 'dictionary'. The bigger and more numerous groups of repeated characters you can collapse, the more effective your compression is. (They can then be decompressed without loss by looking at the dictionary and expanding the relevant characters) One you have this compressed data, you can interpret it in two ways: - Of the possible next characters, which one would fit best into the currently used dictionary entries. This requires analyzing the dictionary very closely, and I'd image it would be quite complex. - Of the possible next characters, which would compress to the smallest file after being appended to the original data. Let the compression do it's thing and see which one fits the bets. The problem being that you have to run the compression 101 times for each new character. As a side note, if you want to speed the data entry up you could always separate the data entry and the AI guessing. Grab hundreds of numbers from a participant at their own pace, then run your guessing algorithm over it.

[@tetra6354]

0:51 dude how did he guess my 65

[@Flitsch-j9d]

same

[@pranamyadangol294]

if this man decides to solve world hunger he could but he says to himself why do that when I can predict your mind with science.

[@leonardsalt]

First video of yours I see, I used to love Minecraft but kind of grew out of it. Love the content! Subscribed

[@septanesium]

This type of AI could unironically be used for really dominative (if that’s even a word) stuff and you choose to play numbers 😭

[@AnEnderNon]

wtf is dominative supposed to mean here

[@septanesium]

@@AnEnderNon a variation of “dominant”/“dominance”, sorry for the confusion

[@AnEnderNon]

@@septanesium yeah what stuff would be considered dominant i still dk wym

[@ehsanreghabi]

Love the video, keep it coming

[@Fiberoeatingfiber]

honey wake up kiddy kene just uploaded

[@nathanielwastaken_1]

15:06 “sigma_first_digit”😭🙏 bro has lost all of his sanity

[@aratictvAlt]

1:10 me choosing 37 because of that video

[@dadamaldasstuff1816]

Me actively avoiding numbers containing the digits 7 and 3 after watching a Veritasium video

[@Spyingcow]

i picked 73 because of personal reasons and now I'm scared

[@av3stube480]

This is the first video of yours that I watched, and I subscribed. So don't worry, at least one person likes this content :D

[@darqed]

I wonder what would happen if you ran it on actually random numbers. like yeah theoretically there shouldn't be any patterns but there's no such thing as true randomness so maybe with enough data itd find something?

[@AnEnderNon]

the randomness that is being used probably does not fit the patterns that he is using (distance away on keyboard etc) since its gonna be based on really complicated or a complicated physical system

[@maze7050]

@@AnEnderNon yeahhh, that’s what I was thinking, they use patterns like when you opened the program and shit, I don’t think it would be over 1%

[@mafi288]

I guess a neural Network would work best for this type of prediction. I would also be easy to get a big dataset to train on

[@lars0334]

Your first no minecraft video and yt recommended it! Love to found your channel

[@aunntaco]

0:53 CASUALLY GUESSES MY NUMBER ON HIS 3rd TRY WHAT

[@WALDtoon]

33:25 for those who are new to the stream calc is short for calculator, its a slang.

[@harryvpn1462]

2:00 this is completely wrong, scissors is, could show a random difference in cultures and their RPS choice

[@mikaskaltoft9761]

Not to mention the numbers only adding up to 90%

[@Egryzilon]

there's 2 Veritasium videos what talks about randomness the first one is with Michael (Vsauce), that talks about randomness, but in a more, philosofic way But the one the video mentions talks about that, says that, we incline to think of 37, when asked a random number between 1 to 100 because we see 3 and 7 as more random. Both are odd and primes numbers

[@marioplushyworld]

Idea: Try to make a AI in Minecraft that can win a game of bed wars against you

[@hilljan]

The easiest way to conceptuallize this is to think of many “random” numbers in your mind, then try to include random numbers from 0 to 10 as the digits. There is often going to be numbers you forget about for at least a split second,and a common one for me is 5.i’ve wondered how you would guess this pattern before, and I now have my answer.

[@_.simply.isaac._]

BRO IS GONNA END HUMANITY

[@liorshoshani]

That's amazing! Love this.❤

[@master_of_blinchiki]

1:26 I love how you used the Epic version of L's theme :D

[@technicalfool]

First number shown in the sequence: 23 Me: Aight guess I'm an NPC now.

[@cheetosnour.scratch-learn]

0:40 my number is π

[@unknown09132]

so 3

[@IJustKen]

@@unknown09132 .

[@AsiccAP]

yeah, probably a good idea to specify an integer

[@mangez_du_pudding]

5:45 🫵

[@Hotdogman287]

Bro your content is incredible l.

[@vengirgirem]

That's hilarious. The first number I thought of was 13 and that was literally tge first number you mentioned. You don't seem to need a computer to predict my randomness

[@ArielFawkwes]

Damn only 90k views, production that went into this is amazing, good luck with future videos

[@Gisulogia]

Fun fact nobody cares if you’re early

[@svallux]

[0:49] NO, WHAT, NU UH HOW, i mean i know how, but STILL WHAT. AND LS THEME PLAYING IN THE BACK TOO, TOO GOOD. i’m hooked in the first minute

[@_Teska]

1:05 no way no one said 69?

[@Justap1g]

@@_Teska you see, in the original video they eliminated 69 and 42 because they're funny numbers, so people were more likely to choose those instead of choosing random numbers

[@_Teska]

@@Justap1g :(

[@aminmagura]

​@@Justap1g 42 and 420 are not the same thing

[@NghiaBui-cq3xm]

im thinking of e^π

[@HassanOuni-h2e]

​​@@aminmagura 42 is the veritasium number. No one seid it was the same as 420

[@Gabriel-rg7cy]

I was watching the whole video thinking it was a new channel I didn't know yet, them the final came and my mind just exploded. Anyways, this video was great. I will surely watch more videos like that if you post

[@RiedElliott]

Now do it in Minecraft

[@xFanth23]

You have no hidden nodes, made me crack up

[@malikphill3324]

Here's what am thinking 10:27💀

[@Davisthedudeedits]

💀

[@Trufflo_]

liking, subscribing and commenting. This was too good!

[@kingdom2]

It's almost like he's still with us... Fly High Kiddy Kene 🕊

[@spaceguy20_12]

@@kingdom2 what? He’s still alive right?

[@kingdom2]

@@spaceguy20_12 Nope. drunk driver crashed into him.

[@spaceguy20_12]

@@kingdom2 that’s sad, but when?

[@kingdom2]

@@spaceguy20_12 did i ask. your tricks don't work on me.

[@allayallay-z1z]

When did it happen?

[@EggZu_]

the guy holding the gun going "i don't know what to guess" and shaking the maraca is the funniest shit

[@Luxof_]

i wonder how many people understood that part where tyler and kene were like "yeah he has 0 hidden layers" great video, i think that just like the human the robot may have been off by like 1 or 2 which is why it didnt get too far past 10%.

[@kubaniezyje225]

super underrated video, ur very smart bro. id be glad to watch more outside of minecraft coding

[@JetLag03123]

amazing video, keep up the great work! i really like the new style of content

[@Trard]

This video is a diamond on youtube. 10/10 good job

[@The_Lunartic]

23:45 this random Windows sound made me laugh so much. I think my humor is broken at this point...

[@redi4ka951]

The remix of L's theme on the background is just a cherry on top

[@paulogodoyp]

i prefer the alternate ending jokes aside sometimes the journey is more important than the destination think of how much you learned, it's incredible

[@wooftubeyt]

oh god the insane command block guy got his hands on more powerful code

[@scriptoblox]

The fact that this was recorded around 3 months ago

[@bomg123]

Oh your that minecraft guy also love your vids

[@redcrafted_]

I usally dont wirte comments under videos like this, however i wanted to say that this video was the most interasting and entertaining i have watched this year!! I whish you all the luck for this new jurney, cant wait for the next one!!

[@Arlogy0]

That programme is insane , I liked it

[@AirzGD]

I LOVE THIS. please do more