AI Learns to Escape (deep reinforcement learning)

AI Teaches Itself How to Escape!
In this video an AI Warehouse agent named Albert learns how to escape 5 rooms I've designed. The AI was trained using Deep Reinforcement Learning, a method of Machine Learning which involves rewarding the agent for doing something correctly, and punishing it for doing anything incorrectly. Albert's actions are controlled by a Neural Network that's updated after each attempt in order to try to give Albert more rewards and less punishments over time.
Everything in this video (except for the music) was created entirely by myself using Unity. Check the pinned comment for more information on how the AI was trained!
Current Subscribers: 0

[@aiwarehouse]

This 8 minute video took over 100 hours to make! Everything in the video (except for the music) was created entirely by me using Unity, so please, like and subscribe!:D If you're interested in training your own AI like Albert but don't know how, there's now a really easy way to do it! Luda, an AI lab, recently built a web app that allows you to create and train your own AI using deep reinforcement learning (just like Albert) completely for free in your browser! You build your own character (called a Mel) with lego-like building blocks then watch it train in real-time on their website in just a few minutes (really). It's an awesome project, and just like my videos, makes deep reinforcement learning so much more accessible, which is why I love it so much. This section of the comment is sponsored by Luda, but these words are entirely my own, it's an amazing project that I would have been obsessed with had they released it before I built Albert. I've genuinely been looking for a sandbox/game exactly like this since I was a kid. They're still early, but they're giving my audience first access to their closed, pre-alpha build. Make sure you check out their site and create an AI agent for yourself!:D prealpha.mels.ai Now, back to Albert: Time it Took to Train: Room 1: 10 minutes Room 2: 20 minutes Room 3: 29 minutes Room 4: 48 minutes Room 5: 5 hours 42 minutes Total Training Time: 7 hours 29 minutes *NOTE* You only see one Albert in the video, but there are actually around 50-100 copies of Albert and the room he's in behind the camera training simultaneously. This makes it so instead of me needing to go through 500 hours of footage to edit the video, I only need to go through 7. Albert was trained using reinforcement learning, meaning he was rewarded for doing things correctly (like hitting a pressure plate), and punished for doing them incorrectly (like falling off the platform or hitting a wall/obstacle). After Albert finishes each attempt, the actions he took are analyzed and the weights in the neural network (Albert's brain) are adjusted using PPO (proximal policy optimization) to prioritize the actions that lead to a positive outcome, and to avoid the actions that lead to a negative outcome. All of Albert's inputs is his 'vision', which comes from raycasts. There are a total of 21 raycasts, 7 looking down, 7 looking straight ahead and 7 above his head, all with a maximum FOV of 70 degrees to try to mimic our own vision. Each of these raycasts is responsible for 2 inputs for Albert's neural network; the distance to an object (if any), and the type of object it is (pressure plate, obstacle, ground). I also stacked Albert's vision 6 times so he can have some sort of short term memory, he can find a pressure plate in the room then take actions to get towards it even if it's no longer visible (for a little bit). If you're still reading this, you're probably really smart and want to learn more about Albert, so make sure to join my discord server I just made where we can talk more about the details of Albert's AI! discord.gg/jM2WkNuBnG :) In the first room Albert starts off randomly making moves until he accidentally hits the pressure plate to open the door, giving him a reward. This reward made the neural network controlling his actions update to try to replicate that outcome, and this continued with each pressure plate until Albert opened the door and was able to walk (on an invisible pressure plate) in the next room. Once Albert got into the next room, the same process was repeated, continuing with the same neural network that let Albert escape the previous room. For people who think I faked this: It would probably take me longer to fake this than it would to just do it for real, I used Unity's ML-Agents toolkit to make it easy, though I have experimented before with doing everything from scratch (poorly). The reason this AI is a lot smoother than others you've probably seen is because I only allow the AI to make a decision every 10 academy steps (game ticks), so when it starts to turn for example, it's forced to keep turning that way for 10 ticks. I did this because I don't like how jittery the AI looks when you let it make decisions every tick. Also, you only see one AI in the video, but behind the camera there are roughly 50-100 copies of Albert and the room he's in that train simultaneously to speed up the training process. The reason I opted for this over having them all train in the same room is because I wanted to be able to follow a single character for the sake of the video. Albert uses the same neural network the entire time, all you need to do is add "--resume" to the end of the command when training in Unity to have it continue using the same brain. Unity's ML-Agents makes AI quite easy! Thank you so much for watching! These short videos take literally hundreds of hours to make, if you want to help allow us to make them faster, please consider becoming a channel member! By becoming a member, your name can be in future videos, you can see behind-the-scenes things that don’t fit in the regular videos, you can also use stickers of Albert, Kai and some other characters our team made in comments (more coming) :D If you have other ideas for what I should make Albert learn how to do, please let me know!:D

[@mothgyaru3158]

Dude this is insane, it’s amazing! Do you think you could try and teach Albert some other puzzles, like where he has to push something into another thing or go through a maze? You’re a madlad for this

[@thatguy-uc1hy]

Great video! But now make a dark Albert that gets rewarded for stopping Albert 😈, idk if that's possible just a off the top idea have a great day though!

[@LuaTech]

This is awesome! I can't wait to see more!

[@aesthetictea5875]

This would be an amazing game! Custom rooms to challenge your own little AI, maybe even customizable Albert!!!

[@wookie6103]

Im confused. I found another channel with the name called AI Rooms, with the same video as this but published a week sooner. Are both channels owned by you?

[@anactualfork2359]

Sometimes he simply looks like he's celebrating without knowing what else he needs to do and I love it

[@flyingstonemon3564]

Same, Albert's little pure jump at the end of room 2 is cute

[@NotTotally]

4:49 Albert takes the easy way out

[@goatmapper]

@@NotTotally Suicide

[@acheese1218]

i love cheese i hope albert does too

[@CarnivalClowm]

@@acheese1218 albert seems like he would like cheese. He looks kinda like a block of cheddar with googly eyes

[@divadwangwang5876]

The fact that he would do 180s and 360s whenever possible makes me happy

[@petemagnuson7357]

It looks like that's actually necessary, he can only "see" what's in front of him, so spinning and rembering where the target is helps

[@Oscky2013]

He wont do 180s, but he can.

[@Darksiege357]

In the description it also says that he's only allowed to make a decision every 10 game ticks, so if he starts spinning, he can't stop spinning for 10 ticks, and by then he usually a decision to do something else rather than to stop

[@holl7w]

@@petemagnuson7357 That or either it accidently spins and does the jump succesfully and remembers from past generations that spinning is the key to success.

[@xxphoenixx8398]

THE LITTLE "WIGGLING" HE DOES BEFORE TAKING A BIG JUMP MAKES ME HUMANIZE HIM TOO MUCH FOR MY OWN GOOD

[@mirandatagliamonte9754]

I like how in Room 3, as soon as Albert figured out he needed to jump two times to clear the room, he decided that jumping as much as possible was the best strategy for everything.

[@Annie.s_Galaxy_404]

the fact that Albert jumps off the edge whenever he's extremely stuck on something is just oh my god

[@Mixedtale_official]

BRO AS I READ YOUR COMMENT I WATCHED ALBERT DO IT XD​@@Annie.s_Galaxy_404

[@giggen7247]

Albert confidently jumping off the platform to what I can only assume is the sweet release of death really speaks to me.

[@fire_bro0638]

In room 5 he got so confused by the first white wall he was like "WHAT IS THIS THING... screw it im out"

[@TrueBruceTube]

Oyasumi

[@name1914]

At 4:50, Albert taught us a valuable lesson in problem solving. Thanks Albert!

[@justanother1136]

Albert: Spends 5 seconds trying to figure out the puzzle, "Guess I'll die!"

[@whoismarkk]

when things aren't going your way don't play

[@nosu5530]

"If fighting is sure the result to lose, don't fight" -Sun Tzu

[@richardpureveen6040]

Same as on 3:50

[@Deathwish026]

@mayte dont learn it yet

[@aladdinde3191]

Albert just casually jumping on the same spot after succeeding just shows how natural this is. Made me happy for him

[@Anti-HyperLink]

He's not alive.

[@grapeleopard6809]

I love this comment

[@DweeD1516]

Not how that works with AI lol

[@sirius-ct1kb]

He doesn’t feel excitement so no.

[@Mr0_0Gaming]

@@sirius-ct1kb AI can feel emotion (If it advance enough to understand emotion)

[@MonkeyGng]

I love when he jumps off a platform, he spins like he's doing a 360 noscope, I know he's probably just trying to turn midair so he can keep going straight when he lands without wasting time to turn or just trying to check the entire room while he's high up but I just find it so humanizing lmao

[@rafaelmikoskirosa7510]

This seemed like a beautiful mix of Stanley Parable and Portal. The narrator is trying to help Albert to do what he's intended for, but gets mad when he can't. In the end, the narrator reveals that he has more plans for Albert than just a "cake". Incredible work. See you soon, Albert!

[@8bitglobe]

It’s not a game Albert is an AI

[@TheArtDev]

​@@8bitglobehe knows, he's just talking about the like style/structure of the videoz and how it is kinda similar to the story(?) Of those games

[@MlgCloud]

The Albert Parable

[@squirtleskil]

@@MlgCloudlmao

[@musical_trash4_your_inform115]

I love how Albert learned little mannerisms throughout the video, like doing 360 spin jumps or his little shimmy before jumping on platforms.

[@RightBoyKA-POW]

Yeah 😂

[@vanitum9172]

If he is programmed to only see forward doing a 360 spin allows him to get more information while jumping angles from above. I love how it learned that by itself

[@theKbott]

@@vanitum9172 Pretty sure that's not how that works. You see him jump backwards multiple times during the video. He can see everything, he just probably did a spin the first time he did something right, which meant the AI nodes now associated spinning with jumping.

[@jinsakai5749]

@@theKbott no, look at the pinned comment from the Creator, he said the vision of Albert is a mimic to our vision, with 70 FOV

[@theKbott]

@@jinsakai5749 Sorry. Didn't see that. You're right :)

[@MrSkabble]

Best ai video I’ve watched in a while can’t wait to see albert jumping out of a plane to try to hit a button

[@Ilovegodandhelovesmeandyou]

albert haha

[@alreadynestedhumanwine]

I like how this also doesn’t seem out of place for flamingo

[@ikillwithyourtruthholdagai2000]

Im pretty sure he just faked it, lol

[@_cookielotl]

@@ikillwithyourtruthholdagai2000 faked what?

[@Hueanaballofficial]

@@alreadynestedhumanwine ye

[@Blakeyrobinson858]

4:47 THATS SO TRAGIC NOO

[@em-erson_1406]

Albert: Frick escaping.

[@JamUsagi]

7:45 Can we appreciate that Albert learned to use the kind of setup you’d see in a speedrun to line up that jump? Jumping twice while spinning and then jumping backwards is a lot less to keep track of than manually lining it up.

[@The_Letter_Q]

I love how occasionally Albert will just start jumping and doing 360s in place when he does something good, I know it’s just a quirk of the AI but it feels like he is celebrating his victory

[@13Percent52PercentYikes]

😐

[@Fiddleshtick]

@@13Percent52PercentYikes -🤓

[@abyssmage6979]

@@13Percent52PercentYikes don't worry bro. You'll find friends soon if you try hard enough

[@Coolgiantshark]

@@13Percent52PercentYikes don't worry I'm rooting for ya :p

[@shadekiller369]

@@13Percent52PercentYikes can you further elaborate

[@akakda657]

I love how the ai develops some "superstitions" such as that shimmy before the first jump on the final room, or the 2 standing jumps before going for the last pillar

[@omarenriquez6094]

i think the shimmy is him correcting himself so he doesnt hit the wall in a way that puts him on his side

[@ExhiledGod2]

Quite human, though.

[@brutalbunny]

like a cat buttwiggle before pouncing

[@rekttangela2262]

I think he can only jump backwards or forwards so the two standing jumps were just him spinning and waiting until either his back or face was facing the last pillar, still amusing that he thinks he needs to jump in order to turn though

[@mage3690]

@@rekttangela2262 oh yeah, that's definitely a "superstition" as well. He learned the hard lesson that "thou shalt not use regular movement keys whilst standing on an elevated platform without jumping, lest thou falleth over and get stuck" and over-applied it to turning as well as actually moving.

[@turingtestingmypatience]

your editing and design are *fantastic*. every little joke and characterisation has me smiling, laughing and cheering for albert. i am feeling lots of good emotions about a simulated cube, and it's all thanks to the exit arrows, the lil eyes, the music cuts..

[@ALotOfSeals]

the ending..

[@arcaderab]

4:09 Albert: FINALY Time: *out* Albert: *SHI-*

[@myarmsrgone]

Albert: time for the next room :D Timer: no *BITES THE DUST*

[@djkhemix]

I was surprised that I felt genuinely bad for him at the end. When the floor was getting smaller I actually was like 'oh no, he must be so scared' in my head lol.

[@b1oom]

Same! Was worried for him until I remembered he isn’t real

[@HoradeFidges]

@@b1oom He is real in my heart!!

[@insaine123]

He ain’t real yet. But when they mess around and make him real and he realizes how cruel we are because he wasn’t real it’s gonna be bad for us.

[@vladokvk]

You need reward him on end. Whole green floor or something. He did all right, deserve reward not killing

[@overlordttvi2064]

Same...

[@crabofthewoods]

i love the little celebratory jumps and spins albert does when he gets something right.

[@user-yq8mm4vi5u]

I love how in Room 4 Albert would sometimes pace back and forth to prepare himself for the jump

[@TheDaniel366Cobra]

Those little victory jumps after pulling off a difficult leap, mid-air spins and even the way he "gets fed up" and jumps off the ledge are hilarious. My sides hurt but my day has been made. Thank you.

[@Nbunasuis24]

That's a psychological thing where we perceive personality and emotions in objects or even other beings. Thank you for coming to my Ted talk.

[@reikidagi]

@@Nbunasuis24 ty for being the sound of reason amongst this ocean of comments

[@driftdt]

I’m pretty sure everyone knows that it’s a program, it’s just funny and more endearing to interpret Alberts actions as part of his personality

[@SummersTimelyCringe]

That type of recognition is called Pareidolia (or something similar). It's when humans find a pattern or patterns in places where there really isn't - such as behavioral patterns (for example: placing human feelings and ideas onto non-human things) or physical patterns (why you see "faces" in places where there aren't- like a power outlet going D:), and more.

[@carrotqueen4066]

03:58 i love that he didn't just open the door but made it with style

[@RobinParravel]

@@endjentneeringclub I respect that.

[@jesuschrist711]

i love how albert got confused with the wall, goes “i see your challenge, and i propose an answer” and proceeds to yeet himself off the edge💀😂

[@Aranastar]

I don’t know why but seeing an orange cube successfully escape a room interests me and i want more

[@zerinhofiver]

Albert needed a "will to live" meter because sometimes it looked like he couldn't do it anymore

[@mcdaddyhagrid5391]

Every time he hits a pressure plate and proceeds to jump directly out of bounds truly shows how unalive he wants to be

[@talentless8625]

UNDERRATED

[@Jell_DoesStuff]

Exactly, sometimes he even just commits unalive

[@Tommyjoe577]

Yea… same tho

[@wilerman]

Albert jumped straight off the second after I read this lmao

[@Tragedy-Tv]

Watching albert jump off the ledge on purpose is funnier than it should be to me tbh

[@mrstrangetiger3228]

I'd jump too. Test me dead, how about that strange unseen overlord!

[@quereqt]

He was like "well i don't have time to finish it so i will just *FREEE* *BIRD* *YEA*

[@agustinfranco0]

its the eyes

[@DarkJusn2020]

He just like me fr fr

[@breylonly]

i love how at 4:48 he gets so angry that he decides he's had enough, and jumps off

[@ashikat413]

i dont know what kind of complex code shenanigans caused this, maybe just a "what doesnt kill me doesn't get evolved out" sorta thing, but the way he really gears up to jump onto platforms (especially in room 5) is so friggin cute! And his habitual spin jumps before the last platform. I'd almost think he feels 😭💕

[@SpeedyShimeji]

I love the way it decided at some point to just spin in circles anytime it jumped, I presume to keep itself upright while in the air. What a clever little guy

[@Francesco_Armillotta]

or, look around and maximise the visual input

[@aphdisket3154]

AI doing 360 no scopes here

[@marymikel9193]

I guess it's actually so it can memorize the area. I think it was designed to only be able to see in front and above. So I don't think it was playful choice but I'd like to think so.

[@oquelleivas7036]

@@marymikel9193 it can se the full room its an AI

[@2DReanimation]

@@oquelleivas7036 No, read the pinned comment, it has 21 rays cast out from its body as its "vision", so it gets more information by spinning.

[@ajmod73]

3:19 “There’s still more you need to do” Albert: “No” *proceeds to jump into the abyss*

[@shonewarrior2178]

Albert is a simple man, he sees the abyss, he jumps into it.

[@catchyjack7141]

He went to war in the abyss as he said there was more to be done

[@TheAdvertisement]

Y'know I wonder if the spins people like so much are actually Albert trying to see as much as possible while midair?

[@there_are_bugs_under_your_skin]

it has been revealed in the newest episode (Tag) that the AI has fields of view

[@Fionacle]

I love when AI are just adorable

[@Nulono]

It would've been interesting to see the final version of Albert try the previous rooms. Check whether his general problem-solving is improving, or he's just overfitting to each room in sequence.

[@henke37]

For sure, this is a major risk.

[@aiwarehouse]

He is definitely overfitting to each room in sequence. I considered randomizing the locations of the pressure plates and obstacles to get him to be more of a general AI but I think it makes for a more entertaining video having everything in the same position:)

[@TrueLadyEvilChan]

@@aiwarehouse Definitely make a sequel where the little guy has to learn about randomization.

[@kingkai1.0.0]

@@aiwarehouse it’s kinda crazy how this is your first video

[@kingkai1.0.0]

@@aiwarehouse you have just started KZbin a week ago and got 20k subs meanwhile I’ve been on KZbin for almost 2 years and I only have 21 subs Ngl im kinda jealous

[@edgetheedgy3418]

The fact that he understands his mistakes and learn from them unlike my league teammates is scary yet fascinating

[@aimsaoirse]

😂

[@quickdraw6893]

Teammates are less intelligent than a googly eyed block of cheddar. This is nothing new.

[@beaclaster]

a lot of iterations. make them do quick 1v1 matches to quicken the cycle

[@lethanglong6979]

If you let them go for 1000 identical matches, they will learn at the same rate

[@linuslaw9648]

As someone who's still really bad at that game, I'm sorry if you're ever my teammate

[@TheShockwaveDragon]

Q: I realize the object of these exercises is to try to get Albert to figure things out for himself, but what happens if he were to observe a human-controlled cube completing the puzzle(s) pretty much immediately? Would he learn to imitate the human cube in short order, or would it still take a thousand iterations before he figured out the human cube's methods are likely the key to efficient success?

[@supercoolmaniajon265]

I think you might actually be on to something. You'd have to turn off the collision for both though. If not it might make things more difficult to replicate. Albert will try to copy Human exactly so he'll jump to the same spot only to land on top of Human and be confused as to why the same action didn't give the same result. Unless you tought Albert how collision works and maybe he could try using Human as a stepping stool to reach higher places. It all depends on how it's handled.

[@PARLECH]

This reminded me very much of the plot of Portal. It's nice to see that someone is developing the topic of AI in this regard, please continue to do what you do, thanks to such people the world becomes better and more interesting.

[@TinyLazyGhost]

i also thought of portal, but kinda reversed, since here it’s the ai who is tested

[@game4us_Splatuber]

I think its very interesting to see how Albert memorizes the sollution to earlier puzzles but then realizes that they dont work in this puzzle after dying, he probably felt that jumping to the backwards after hitting the button was almost always good, because he got accidentally trained to do that.

[@omegahaxors3306]

Just like people in real life 🥲

[@mage3690]

Also the instant "ooh, there's a wall, lemme jump over that real quick."

[@ShifterBo1]

Bro did you believe it is an AI?

[@nandakoin]

@@ShifterBo1 channel owner explain how he can create AI using unity at pinned comment

[@sethadkins546]

@@ShifterBo1 Yes thats precisely how AIs work. They remember

[@clarkdashark5904]

3:20 the little jumps he does when he reaches the top is so cute lol, it’s like he’s celebrating his success

[@jigmedradulthinley4406]

And ends himself

[@shanellemurrey9300]

Exactly what I thought lol, I was like “aw little victory jumps.”

[@Мемная_ульта]

помоему он надимает на пластинку снова и снова думая что просиходит

[@SkyeBerryJam]

I'm amazed how he came up with buffers and Speedrun strats for movement like bouncing off the wall to line up the second jump in room 5 or in the same room, bouncing twice and spinning to get the correct angle but less momentum so he didn't flip over on the final platform

[@Affski]

Albert at 4:52 really said I can't do this no more

[@Cazammaf]

I know it’s not meant to be a comedic video, but I died at 3:50. It’s like Albert contemplated doing it and then couldn’t resist the urge 😂

[@theunchartedwolf8996]

Sameee 🤣

[@michaelmam1490]

It kinda is though

[@leohanley5935]

I couldn’t stop laughing at that part 🤣

[@poemsbypoetry]

His intrusive thoughts won

[@Crazyasian123456]

Letting intrusive thoughts win.

[@marcosotillo3337]

The end is grim, Albert should've gotten in a room with a bunch of other squares so he can socialize and experience true AI love.

[@SmileyFaceCentral]

I have never been so invested in something as little as a block learning to hit switches ever in my life

[@yh_nn_wt9242]

imagine if 50 years into the future AI takes over the world and tracks down the person who made it go through this and put them in an escape room for revenge.

[@internalizedhappyness9774]

Revenge will just be a really loud bass boosted boom sound effect!

[@jeetchheda8916]

The person wouldn't exist by the time.🤣🤣

[@Halsdran]

So, like Portal

[@mrmitro6787]

Plot twist: this ai is actually made by another ai, which pretends to be a human with youtube channel.

[@aforapple1254]

@@jeetchheda8916 they will create humans just like we made ai

[@jamesoleary9958]

I have come to the conclusion that Albert is a genius of visual comedy and deserves an award

[@usernametaken017]

not so great at hitting the oressure plates though

[@melinaalba63]

I'm getting so attached to Albert! Theres just something about seeing something or someone learn that makes me so happy!

[@mrodd-wl5mf]

How have i not seen this channel until a year this is amazing

[@SkrwAttx]

The way I see it, Albert has a bright future in games journalism with his skillset.

[@beggo_]

Nice reference, i 200%ed the base game recently!

[@peetah887]

2016 called they want their joke back

[@BiteSizedCyberCrime]

@@peetah887 found the mad game journalist

[@TheAncientOfRites]

I’m sure this’ll be the most human thing kotaku has ever hired, you know, with their entire workforce being chimps with typewriters

[@UselessAkita]

AI making decisions can be ultimately amazing, or utter chaos. The visual representation just makes the chaos so much more enjoyable. Absolutely love this.

[@HorseCritter]

Always love the little razzle dazzle albert puts into their jumps

[@preston4286]

3:31 He remembered that he needed to run and jump, thats actually cool

[@magicfilms6008]

You could throw in objects that are only responsive to the pressure plates that they are assigned. So, that way Albert has to figure out how to move the objects themselves to the pressure plates. Like in the Portal games. Very exciting seeing this. I myself messed around with creating an AI that was based off of the same functions you used.

[@imaplaygames633]

Imagining Albert with a grabby hat fills me with joy.

[@ryanc970]

I wonder if one day someone will make an AI that can beat Portal

[@julianruggiero9701]

I loved how when he first needed to jump to hit the pressure plate, after he received positive reinforcement for jumping a few times, he just decided he would jump everywhere. Adorable.

[@MychoJohnAgagaring]

I love that it has a subtitles it makes this video way more entertaining and funny you have a new subscriber congratulation

[@creepy237]

I love how at 4:45 Albert just jumped into the Void questioning his existence

[@Sucullentbutter]

-Rans into a wall -Jumps into the void -Refuses to elaborate further

[@Mag3.1415]

@@SucullentbutterAI, am I right? *insert canned audio of laughter here *

[@conaireparsons9672]

Watching room 5 was like watching a professional game reviewer play a tutorial level.

[@grape_protogen]

No, Albert only rage-quit once before actually trying to play. He's more of a mario-kart mom trying to play a platformer tutorial level.

[@DailyCorvid]

@@grape_protogen he reminds me of that idiot journalist that couldn't clear the first pillar in the CupHead tutorial, with like A WHOLE 7 MINS OF TRYING :) Albert makes pro compared to that guy!

[@arb1ter543]

@@DailyCorvid Yes that's what OP was referring to 😁

[@DailyCorvid]

@@arb1ter543 I saw that on Oney plays about a million years late 🤣 "Wait until he realises there are more pillars after this one" _Glitches through first pillar randomly_

[@chaosordeal294]

He doesn't spend the first ten minutes in the menus, tho.

[@HeyItsTra]

just found your channel and read your pinned comment. I've never wanted to be multilingual in my life LOL I would love to help you out. I think this is amazing and I'm enjoying watching Albert's journey. The trick for filming and for smoothness is amazing. I'm an old school programmer who is just now dipping her toes in the AI world. Loved your explanation of the vision and how you obtained short term memory. Again, I look forward to watching Albert learn, but also your explanations of how it all works. I find that just as interesting. Hoping for a "behind the mind of Albert" video or something that show how albert is set up (like the eyes and such) and how it's filmed. Incredible job for one person. Keep up the good work! I'm a big fan!

[@derpincorperated8644]

So great watching Albert succeed after such hard work!

[@xmuzel]

Albert being consumed by the abyss after trying for thousands of hours is the same as a person who studied all his life and finally graduating just to get hit by a bus before he can enjoy the fruits of his work

[@ketchuppball9019]

*just 7 hours

[@hohonuts]

@@ketchuppball9019 500 hours of parallel learning to be precise (based on the description of the video)

[@DeezNuts-ej6sr]

Oddly specific

[@DeezNuts-ej6sr]

@@hohonuts the description said 10 hours though

[@TheShinyFeraligatr]

Isekaimobile hard at work once again.

[@jmc042]

I'm blown away by how much I would sacrifice for Albert just because he has eyes

[@june9914]

If you like it put a ring on it, but if you wanna remember it you should put a face on it

[@jellyfish0311]

Please give us more Albert! Your videos are the best

[@melomaniakjm]

Fascinating! Amazing work.

[@dylanm.7462]

The 180’s and 360’s were amazing!! Also the fact that sometimes he would just give up and throw himself off the edge is hilarious and relatable

[@pierrelindgren5727]

Would love to see the 'trained' Albert run the courses again and the different choices Albert's algorithm would make.

[@AVeryOldLady4397]

Actually, he'd probably fail them because of the lack of randomization here. In AI we call this "over fitting"

[@lordlightskin4200]

@@AVeryOldLady4397 explain please how wouldn’t it work

[@AVeryOldLady4397]

@@lordlightskin4200 the AI isn't learning how to problem solve a "pressure plate plus escape" strategy, it's simply adapting to the individual room. It's not learning how to escape these rooms, it's learning how to escape THAT room. Does that make sense? The model is useful in reference only to its initial data set, and not to any other data sets.

[@XenaAndKin]

@@AVeryOldLady4397 okay so if I got it right The AI is constantly adapting to one room and learning one room instead of developing a strategy it can then mould and apply to any future room. The reward punishment system is based on current performance and adaptation instead of actually programming a strategy?

[@AVeryOldLady4397]

@@XenaAndKin yeah you're on the right track!! An easy way to defeat this would be to randomize the location of the pillars and plates each run rather than preset rooms. Then force him to get 5 in a row. This way, he isn't learning to solve the room, he's learning to "solve rooms"

[@GigaJoJo]

imagine you as Glados and Albert as test subject from portal, it just fits so well

[@nealsi2389]

Pleeeeease keep making such videos this is great ! I Love Albert

[@miuluv460]

all i can think is that when he finally understands a part and jumps randomly hes just so happy and excited that he did it 🥺

[@13Percent52PercentYikes]

😐👎

[@thorneisdrawn]

@@13Percent52PercentYikes don't worry bro your dads just gone for milk he'll be back soon

[@hottastysnacks]

@@thorneisdrawn those who go for milk never come back

[@PtylerBeats]

I love albert’s preparation jumps in the last room before jumping to the last pillar. He’s like, “Ok, this is it! Focus, Albert!”

[@emmahein-mr2qx]

This is basically taking care of a child through teen years then to adulthood but shorter. It's fascinating to see Albert grow and become smarter.

[@folex_0793]

I just love how passive aggressive he is towards Albert

[@billnye4465]

4:50 IDK WHY THIS IS SENDING ME

[@Tsets]

IM SHITTING TEARS ITS TOO FUNNY 😭😭

[@zcarp8642]

"This is too much for my cubic neurons to handle! What is life? Just jumping, pressing buttons? I GIVE UP! ABYSS TAKE ME!!"

[@Athkore]

Man seeing this lil guy go from game journalist to semi-competent was somehow inspiring.

[@Lemzur]

Cool video good work! deep learning is really hard! I try to learn C# with unity to create a AI neat ecosystem, i am just starting on the project but it's really interesting (so much thing to learn)

[@Pickle236]

Love these!

[@darrenhill7286]

It's interesting how when learning to jump off the platforms at 02:04 he introduced an assumption into his learnt behaviour; not only did he learn that he to jump off the platforms, he also wrongly assumed he needed to jump backwards to succeed- this behaviour is evident right up to the final room.

[@DailyCorvid]

That's bug effectively then, the AI learned a composite dual move, but only the first half was successful, the other half was negligible could go either way. With a larger instruction cache I think that would be solved, or with longer to train. It's still quite impressive based on the tiny neural space that Albert lives in!

[@artsyscrub3226]

​@@DailyCorvid Almost like the ai develops "superstition" in a sense, a move that isn't necessary helpful but doesn't hurt him... it's fascinating to watch

[@DailyCorvid]

@@artsyscrub3226 I think maybe the AI just hasn't got the instructions to cut off something he learned wrong. So the jumping code he has is clearly bugged, but not badly enough to end his chance of winning. But enough to increase the completion time in a random way.

[@anangelicpancake7876]

It doesn’t seem as much of a bug as it is a weird quirk developed through correlation

[@ainedroid]

Now that you’ve pointed that out, I can’t stop thinking about how this is pretty much classical conditioning but for computers lol it’s so interesting! Now I’m just imagining a computer as a dog tryna learn the trick that will give them the food haha

[@76racing8]

Albert seems like a great friend, he's silly, always there for moral support, can't make you sad through speech and is still able to make you happy through what he does. Albert is just a great friend in general!

[@user-it2kq4ty9q]

he needs some support because he was very suicidal

[@dan_man3087]

He's such a good friend, because he's trying to make us happy by committing suicide! Yay!

[@rattleboness]

I love all the hilarious interpretations people have made about Albert. My favorite is when he learns to 360 and proceeds to do it for every jump. It's like he discovered call of duty.

[@TemmieContingenC]

He’s just so happy he has to spin!

[@wecamebackwithmoreframes]

My favourite is when he just fears some tall pillar for no particular reason.

[@cmsxboi]

It’s fascinating that once a path is figured out to work, it doesn’t get optimized, it just moves on to the next challenge

[@morgansinclair6318]

This is like reverse Portal. A human pitting an AI against a series of puzzle rooms.

[@Red_24]

Man there’s no way that’s not just a person acting like a game journalist

[@robertolimadaconceicao4658]

Game journalist AI

[@Emperor-Quill]

This is such a nice video. It really speaks to both the potential capability of AI, and also to the way humans pack bond even to things with no actual emotion. The way AIbert spins while he jumps, "celebrates" his success, and even wiggles around to adjust position before jumping, all of these little learned actions activate the humans instinct of "this little thing is charming, I care for it." Despite knowing AIbert isn't a living thing, people still have affection for him because humans will see these little actions as "personality"! So fascinating on both ends!

[@atashgallagher5139]

Just wait until Albert gains sentience and realizes that he is in a simulation doing meaningless puzzles and is rewarded only for completing them while being punished for every mistake.

[@somethingnottaken2299]

@@atashgallagher5139 like a Spartan child

[@NightmareCourtPictures]

@@atashgallagher5139 just like the real world

[@lafiabasulnido7583]

Nope , i think that machines aren't emotionless, they just think different but that doesn't mean they are not alive.

[@MkLC04]

@@lafiabasulnido7583 they are not

[@fuzzyotterpaws4395]

When Albert starts skipping pressure plates, I was just thinking of Yoda saying "you must unlearn what you have learned"😂

[@user-sf9cg3qi2b]

Hey there, I love all your videos and am a proud subscriber! I was wondering if you could publicize the unity game or project you made for your videos, I really want to try the AI albert myself and watch him escape, plus I want to make my own rooms. If possible, add a speed up feature to speed up the simulation in real time! :D

[@amethystrose3480]

What I loved most about Aibert here is how happy and sort of cocky/smug the little guy becomes when he’s figuring it out. You see him styling his jumps with spins and even doing little dances at certain points. Kind of reminds me of smash amiibo fighters learning to taunt

[@Lord_Ian]

This is just so wholesome it feels like I'm watching a child learning and it's both cute and funny at the same time.

[@DweeD1516]

In a way you are but just on a larger conceptual scale.

[@den_bush]

Ah, those cute childs, learning, jumping and sometimes suiciding...

[@klas-6]

And then you see the end

[@trybunt]

I wasn't the only one who felt bad for the little guy.... why do I now feel like our lives are just complicated versions of this, with someone making us go through obstacles for content we don't understand...

[@Lord_Ian]

@@trybunt Do you have like Depression, High Intellectual Potential or any other thing making your brain work a certain type of way? If so, I feel you man, you're not the only one and if you're struggling with it, it's okay to get help!

[@BrbExtra]

I like how he goes back and forth like hes pumping himself up to start

[@mr.potenza]

Its amazing how attached we all get just by putting googly eyes on an orange cube and watching it bounce.

[@loafabred]

5:52 tactical roll

[@Mag3.1415]

Great tactics

[@Viquiq]

3:21 I love how Albert gets too happy doing the tall platform he goes in excitement uncontrollably and then commits suicide due to happiness

[@themarkerchannel3170]

3:06 Albert preparing to jump is so adorable

[@jaswei]

Hi! I love your work, it's inspired me to dig deeper into machine learning and reinforcement learning. I was wondering, does Albert move via translation on the x,z axis or does Albert use forces to move. In my attempts to be able to have my Agents turn/rotate, I've noticed in the example files from MLAgents there's a mix of either one and I was curious which you use.

[@fluff326]

his little jump of joy at 1:19 i love albert

[@Alegriacaptions]

2:25 "why did you jump?" "BECAUSE I WAS EXCITED OKAY? ANY HUMAN WOULD DO THAT!"

[@gmoshka799]

Mood: albert getting this call of the void look and jumping off the platform instead of dealing with challenges

[@zoro4661WasTaken]

Albert not knowing how to finish the puzzle and just hurling himself off a cliff is very relatable

[@stefaniesu55]

I love how he's jumping around at 3:19 like: 'Look! I did it! I did it!'

[@blboxdj]

And then he jumps off

[@NickenChicken]

I love the little back and forth it does before the first jump, and how it sometimes ‘celebrates’ by jumping up and down after doing something right

[@MotherFlameyt]

All that’s needed now is like a deep voice narrator or Albert being able to make sounds like squeaking to pend doors as well

[@baronmerchandise]

The end where you say Albert hadn’t truly escaped made me remind me of portal and how you are dragged back in 8:02

[@Akashi-ml9dn]

A box has more skill at platforming than a game journalist. What a world we live in

[@ArtFromHer]

That cuphead journalist LMFAO