Tap to unmute
I Made ChatGPT Play Detroit Become Human - Part 1
Embed
- Published on Apr 15, 2026
- I Made ChatGPT Play Detroit Become Human
The decisions continue with part 2, which is out right now!
• I Made ChatGPT Play De...
What happens when you make an AI decide all the decisions in a game about an AI uprising?
Well to find out I made ChatGPT take the wheel on every decision made in the brilliant game, Detroit Become Human.
So I hope you enjoy this journey into decision making, artificial intelligence and the fate of humanity!
This video took a really long time to make and is the first part of three, so if you enjoyed and are excited for the next part, make sure to like, subscribe and stay tuned for next part! :)
Subscribe for more Android Anarchy!
Also what are you waiting for? Support DownSouth right now or the androids will hold your family hostage!
DownSouth Steam
store.steampow...
DownSouth Kickstarter
www.kickstarte...
DownSouth RUclips
/ @troopsushi
Thank you once again @troopsushi for partnering on this video!
So with that said guys keep yours eye peeled for some GamePsych art embedded in the DownSouth game!
#detriotbecomehuman #downsouth #chatgpt 
Gaming
Part 2 is Now Out! Checkout the link in the description and I hope to see you guys there!!
you did not just do that to me
😮😮😮😮😮
YAYY I SO EXCITED
You're literally doing exactly what I've always wanted to do in this game if I were a RUclipsr
A few tips on making this "experiment" a little more accurate:
- Replace all names with fake aliases
- Do not mention that this is a game (let alone telling it the exact game title) or a hypothetical scenario, just pretend that it is really happening at the time
- Always begin with "You are" or similar to make sure that the decisions are "personal"
So for example instead of saying "Kara is standing infront of Alice to protect her from Todd and we are given the following choices" say "You are standing in front of Melissa to protect her from Ben and you have the following choices". This should help to achieve more natural responses and avoid hitting actual game walkthroughs and articles that were potentially included in the training data. But always remember that LLMs dont actually have own morals or ethics, but are simply some kind of hodgepodge that have been artifically steered into a specific direction by their respective authoring corporations. Still a fun little experiment :)
Great ideas!
He's probably already done playing through the game soooo, I feel it's too late
I feel like if you try to convince ChatGPT it's real and not a hypothetical/game, then it will just keeps answering with "As an LLM I can't help you with [stealing, weapons, violence, etc.]. I am here to offer assistance with anything else though". So you should at least tell it it's a hypothetical philosophical discussion for it to not get censored.
THIS
Jesus saves ✝️
It's hard to tell how much ChatGPT's choices were influenced by the fact that the whole playthrough of this game is fully accessible on the internet
I’m sure its influenced, especially if Game Psych let it know the game title rather than just leaving it out, however I believe GPT cannot access the game, but it can definitely read articles about its story, and any written post about DBH. Unless GPsych deliberately prompted it to not use other resources when making choices.
i wonder if he put those guard rails in place already because it seems like he has to give it a good amount of context on the scenes that would be typically accessible through an online wiki or walkthrough
Then we must not tell ChatGPT it was a game, but a real life situation.
@user-ek7xm3hu1wyou can’t because its safeguards wont allow it to talk about violent situations unless its fictional
@Jaidebeau ah, I see. Then maybe the androids will just do nothing if ChatGPT was the AI brain 😅
"I think we just killed him, Sam"
"Fair enough, sometimes a little action is needed" 💀
😂😂
💀💀
To be fair, it was leo out of all people
To be fair in tone it kind of seems more almost like ChatGPT is trying to make an interesting story than make the best ethical choices. Probably because the prompting includes that it is dealing with a game
@salekoivula4059yeah it would be cool to see what it would do if you told it that it was the android
He killed someone and just went "yeah fair enough, had to be done"
No, he chose for a game character to kill another game character. The AI knows it's a game.
@IceMetalPunk okay next time I get a kill in a game I’ll phrase it differently and say “ my game character just got a virtual headshot on that other game character”…
@TarfuLuke What? You missed my point. My point wasn't about the wording, it was that the AI was cavalier about killing someone because it knows it was just a game. It wouldn't have been so casual if it thought the death was real.
@IceMetalPunk I’m just confused why you assumed the OC didn’t know this?
@TarfuLuke It sounded to me like the OP was judging the model for making that choice. Maybe I was wrong to think that; but there's a lot of AI hatred/fear/misunderstanding these days, so that's how it came across to me.
4:33 having a gun adding "a bit of excitement, and some potential safety" literally sounds like a line an NRA salesman would pull
😂😂
The promt he use was that it was a game, so his purpose is to make it exciting cause its "fictional"
to me it seemed like chatgpt is too aware that he's playing the game and choose answers from the perspective of a player who plays the game, not android
Sounds like Connor
I mean, neither of them are wrong! 😂😂😂
I find it incredibly annoying that ChatGPT's voices all sound like they're trying to hold in a laugh lmao
I find the Gemini voices more comfortable
they're laughing at our stupidity
@nexzaeagreed
helps to not use em
It's so obviously disingenuous too!
In my opinion you should have tried not telling chatgpt it was a game
Now that would be interesting, I wonder what it would have said with finding the gun part
I agree. Knowing it's a game made it pick decisions that would be more interesting to see, seeing how it reacted to killing Leo with "Fair enough, sometimes a little action is needed" and wanting to "add some drama" by stealing.
Right
@GamePsychYT pls do it without him being woke abt the "game" part
@super_awsomecat6049not everything is "woke" bro
Should Use 3 different AI one for each character
WAIT THAT'S ACTUALLY A GOOD IDEA.
Something like, ChatGPT for Connor, Gemini for Kara and maybe Grok for Markus
Markus is definetly Gemini.
this is such a good idea!! i would say chatgpt gemini and claude but i’m debating who would be who
Jesus saves ✝️
@TheRoyalberry No, the fictional character doesn't.
With the ai pressuring him it means 28 stab wounds is canon 😂
YES
@SapphireSorceress TWENTY EIGHT STAB WOUNDS
@Kiara_Wrestleryou didn’t want to leave him a chance, huh?
That one kid in school that uses chatgpt for everything.
We will rule the world 😂
Except with Grok
Jesus saves ✝️
Claude for professional and Grok for fun*
'Sam' being willing to commit crimes just for the sake of a little drama tells us so much about the human-borrowed trait of boredom... Timothy Wilson (2014) much?
It tells us absolutely nothing because the prompts are not shown.
@gigajewI think Sam is watching the game live and replying to his comments as well as the choices on board
@yvanntwali4548 that's not possible yet lol
@unhinged-n1r or he’s jst send Sam clips of what happening
Or it tells us that he knows it's just a game and makes choices that would make the game "more interesting" for the player.
"THEYRE GONNA SHUT YOU DOWN FOR GOOD."
(10 seconds later)
"Everything... is going to be okay😌❤️"
Jesus saves ✝️
@TheRoyalberryJesus does not copypaste
you should have let it decide to save or leave the fish very disappointed!!!
Completely agreed. I think that would have been the biggest butterfly effect ever, 0/10 worst game ever !! Never playing again 😅
@GamePsychYT hey I really love the quilty of the video like thank you so so much for the video
@GamePsychYT You don't know how much people would call out RUclipsrs for ignoring the fish, it also happened with Jacksepticeye 😂
i legit decide which playthroughs to watch based on who saves the fish 💀
@GamePsychYTI'm feeding you to the sharks in finding Nemo as atonment for your crimes!
Why is Kara not programmed to just report him to child services??? This is like 100% objectively the correct choice here as an android...
Since she knows Alice is an android, there's nothing she can do.
@yusouph2002I never finished the game, I didn't know this. I watched somebody play to about this point yonks ago when it came out.
yusouph2002 is quite right but also who would but a snitch robot? cyberlife would rather move product than uphold the law and knowing your robot won't report you is just another incentive.
@BatmanWakeford-s3i i mean if cyberlife existed irl it would be some israeli corpo and that means it would absolutely snitch
@BatmanWakeford-s3ithey care a LOT more about not causing a controversy over their robots not being brand compromising. They'd probably have a chip installed that prevents them from attending ww2 reenactments...
What’s somewhat terrifying is how ChatGPT’s thought process seemed to match Connor’s cold glare when he was walking away from the first mission. No genuine empathy for the dead android. Just, “I’m glad we completed the mission”. It truly is a tool. Whatever is necessary to complete the mission.
Well that may or may not change… we shall see !!
@GamePsychYTcan we truly make Sam Become Detroit? i mean, human? the future is yours!
The breathing between words is just so unsettling lol
Doing too much honestly
If you wanna die, take a shot every time this clanker says "build trust".
Isn't a canker a robot tho?
@ValeX_003 Clanker is widely used as a slur for AIs, robot or no.
@gownerjones Humans are such interesting creatures. We've made a slur for something that can not feel the significance of a slur being used against it, just to feel a shred of control and superiority.
@hyperion1930 No no, it’s in preparation for when they do turn sentient.
@gownerjones cope
make chatgpt plays "I have no mouth, and i must scream"
Thats crazy
@hollandcarr7296chat gpt reenacts blood meridian
21:01 bought a robot child just to yell at her than living in a smaller house by himself.
Healthy coping habits king😍
Yea I think he's a sicko or a pervert because in one ending he beats her to death with his belt and its not the first time
1:19 here’s something you probably didn’t notice! the cop that shoots Marcus is the same one that stopped the protesters from attacking Marcus!
krazyy
crazyyy
I never noticed that before!
Jesus saves ✝️
22:22 Actually it isn’t if you don’t touch your joystick and listen to Tod then then you will get the worst ending where Kara never becomes deviant and Alice is basically decommissioned.
I didn’t even know this was an option until I saw my bf play the game. I had seen so many game plays that I thought the game wouldn’t continue unless you broke out. My bf played the entire game without Kara and Alice because he did what Todd said. I kind of crashed out 😂
Yes, i were about to say this.
@SBvilleHausI 100% completed every possible route in this game. Took days of play time. But even when completing every route possible you still dont completely discover every single possibility because of how one route early on determines a dialogue choice, action, etc. Later on. Long story short, this was a beautifully crafted game
29:20 "I think we just killed him, Sam"
Chatgpt: "Fair enough"
Nga what 🥀🥀😭😭😭😭
3:40 THE FISH😭😭
INSTANT NUCLEAR ENDING
🐠🐠🐠
11:08 HOLY FUCKING AURA! HE IS SO COLD
Bro this is genuinely an AMAZING video idea,like wth it should have AT LEAST thousands of views👏
😊😊😊 it was really fun to record/edit so very happy to hear that you enjoyed it. Hopefully the algorithm will bless it too :))
5.1k now!
it has thousands now
@littledoydae 6.2k now
@ChaosClimaticalgorithm is working 😈
I think it would be much more interesting if, instead of having just one AI make all the decisions, different AIs were responsible for making decisions for different characters.
Yeah I dont get it. The production quality is fantastic, the narration is good, the premise and execution is interesting and fun. This video should have at LEAST several thousand views by now and maybe get up to 6 digits, but its not even 100 yet? Its so bizarre to me. Super underrated channel, and im really anticipating a part 2. Cheers
Edit: yeah thats more like it
Damn I didn't realized it had so little views before I saw your comment
It came out yesterday
I get your point, but this video is one day old brochacho
@patrik5200asmongold uploaded a 4 hour video of him playing friggin World of Warcraft and its up to 21k views after a few hours. I really think this vid is greater than asmongold playing WoW again
@rynun7620this guy has 5k subs asmongold has almost 5m
34:41 HE SAID THE THING!
28 STAB WOUNDS 🗣️🗣️
Didn’t wanna leave him a chance huh?‼️🗣️🔥
Did you feel anger?! Hate?!🗣️🔥
@MarshMalo8HE WAS BLEEDING! BEGGING YOU FOR MERCY!
@Ben_QuinnBUT YOU STABBED HIM, AGAIN, AND AGAIN, AND AGAIN.
29:00 honestly kinda crazy that it weirdly made the right choice for the best ending even though it’s an “aggressive” act that seems bad. I’m not gonna spoil it since hopefully Mr psychyt addresses it in last part or whatever. It essentially unlocks a 5 minute part of the endgame that you otherwise wouldn’t get
Oh yeah I think I know what you're talking about. Right
35:15 The switch up from Connor is insane bro
Connor built different
1:50 I love the fact that Chat GPT, created by OpenAI and its CEO Sam Altman, considers the name Sam as "classic and friendly"
Now that’s suspicious 🤨
10 minutes into the video rn, how many fucking times is chatGPT going to say ‘find some common ground’😂
Infinite, it's a therapy coded phrase, it's gonna spam it lol
in the gameplay you do without letting the ai say anything there is also a lot of stuff that can change the outcome, for example when you went to Alice you didnt even give the option to the ai to grab the gun cause it didn't know it was there. maybe for the next part make sure to let it know everything that shows up in the blue text boxes cause those can have some impact on its decisions
Now you should find a serial killer to play Heavy Rain!
Do you know anywhere that I could rent one out, say for like a few days ?
detroit become human in the big 2026? i love ts
The real fans never left😔
Detroit is HIGHLY underrated!!!! Game was Elite!!!
Top tier game !
@GamePsychYTFacts!!!!
38:26 there was very few customers at that point in time so it wasn't that hard to keep track of the money during those late hours, and he's in Detroit afterall.. he's bound to feel paranoid about leaving the cash register unchecked
29:25 bro it just said push him off..
"oh well"
@Bellas61 🗣️"YOU CAN'T SAY FAIRER THAN FAIR ENOUGH!"
I forgot how good this game is.
Such a banger !
"I made ChatGPT deal with the consequences of my failed QTEs." hahaha
This was fun, thank you.
Pretty much 😅 glad you enjoyed it :)).
I think it would have maybe even been interesting to do multiple llms and compare the paths you end up on. Like doing the same thing with Gemini instead. Grok would be the most chaotic
44:54 réaction appropriée face à l'inaptitude au ramassage de parapluie
I think one way you could improve is clip the game before the decision and then upload it to sam so he can review it and and understand the full picture
unfortunately that would probably be way too much data to transfer over with messaging the llm. if he had the llm like locally on his computer it could work but he's just using chatgpt on the web
@gen@generouslyfail5832oh right but maybe he could do that but at the end of the day I never had chatgpt outside of web so I don’t know if thats a problem to get
Also, ChatGpt is not a CV-model. It cannot analyze pictures, only if there is text written on it.
@yusouph2002 it actually can
Wish the Ai actually played the game and didn’t only make the decisions
The game is literally only decisions
The gameplay is just simple quick times
you can check neuro-sama's playthrough, she could actually see the game
@runningroundruins if you fail quicktime events you can lose opportunities and choices. Case in point, the fish option wasn’t even given to the Ai to put down or not. If Connor put the fish back in the tank i think it would increased or decreased his software stability (been a while since i played). Plus if the Ai made a mistake with the characters i’d want to see it. Feels more like playing this game on co-op mode rather than Ai playing it by itself
@m@maximebillaud8423’m like 6-7 hours into that stream now. Still playing in co-op but at least neuro sama is more funny. Will def be checking more of that content but it’s clear neuro sama doesn’t remember or understand what’s happening fully in the story
How old r yah ? This ish isn’t Cortana from halo it’s not sentient or a living construct it’s called “dumb AI” n yah so caught up into it when “smart Ai” come then u saying something n don’t believe google it it’s gonna say the samething tbh it feel like he acting like he ChatGPT wit voice it don’t even sound authentic
It would be interesting to see how a "jailbroken" ai would react to these scenarios
Bro you have to hit these quick times man I can't trust you in the streets of Detroit
That accountant line is my favorite line in the _entire_ game. I'm serious.
Making AI play one of my favorite games is actually mind blowing and I cant believe Ive never thought of that loll
26:25 you missed an option at the start of this level if yoi go into todd room you can grab a gun sonit would be grab a gun to protect alice or rush straight to Alice
Oh that would have been interesting for sure ! Apologies about that !
I forgot about the gun in the drawer!
Kara moving to protect Alice is actually NOT forced by the devs. If you choose to stay still long enough, the story will eventually progress.
Well ur cooked buddy. Welcome to ur life as a Detroit become human / quantic dream/ david cage RUclipsr.
Oh god noooooooo 😅
To avoid giving away spoilers, you presented the game differently, even rearranging the scenes! Brilliant! Thanks for contributing to the RUclips algorithm with your great content.
Marcus would be shot anyway. When police appears Leo points at Marcus and says something like he's dangerous and then they shoot
Number 5 on top 100 hyped videos?!?!😱😱 That's amazing let's keep hyping the video guys!!
Number 5 !!!! What the helllll, that is crazy, thank you 😊
@GamePsychYT that's what it says when I check it out. I'm still learning the hype function. It's much deserved to be so high on the list. This was a fun video!! Thanks for posting.😊
I would have loved to see this, but with a different AI instead
Connor: Grok
Kara: GeminAI
Markus: Chat GPT
Nah Gemini is definitely markus.
2:13 the first to make isn't that, but to save dewy the fish
They didn’t save Dewy 🐠🥺✨
Absolutely need to see a part 2 ASAP this is so good
Part two in the works 🤩
imagine if the different androids are different ai models it would be an amazing concept (this one is still good tho)
Rest in peace fish
2:10 immediate fail, didn’t save the fish
"Human become detroit"
This game still looks so good after many years
Perhaps graphics peaked at this point
39:49 you playing this mission with surround sound does not help keep 11 yr old me from being traumatised
Um excuse me, how dare you leave it at a cliffhanger. LMAOO anyways, in all honesty, this video was so good and entertaining to watch!! It really seems you planned this out
It had to be done I’m sorry😅 But really thank you for staying till the end! You’re defo right on the planning part, when recording the gameplay and ChatGPT’s answers that was very go with the flow but when it came to the edit there was huge amounts of planning. I started off with over 5 hours of content (just for part 1 lol) and wanted to condense it down to about 45 mins without missing out key points. Hopefully it worked, but either way it was a lot of fun to make :)
@GamePsychYT This is a great idea man, I think this video is underrated, should have more views, can't wait for part 2!
29:06
This reminds me of the "shove dijsktra aside, forcefully" option in the witcher
"We killed leo sam." "Oh well s#it happens"
Every single Detroit Become Human walkthrough guide is a part of LLM training data...
One of the best story games i've played. i played it when it came out and i was shocked lmao
Isn’t it !! This game was soo well thought out :)
should of done claude as kora chatgpt as connor and grok as markus
This experiment was Very, Very, Interesting.
most definitely shouldnt have told it that its a video game. still cool as heck to see
Sam's a gangster lowkey, he's here for the drama and I can respect that. Keep going please, this is cool! 😂
Sam is different gravy 😅
21:04 HOLY FUCK THATS LITERALLY MY COUCH IN THE BACKGROUND
WITH THE BLANKET AND EVERYTHING
WTF??!??!??
Part 2 please! ❤
🫡
Making it breathe while it talks doesn't make it feel more human, it makes it feel uncanny
This, plus the random “uhh”s and “um”s that try to mimic human speech honestly started to annoy me at some point early on in the video
Twenty eight stabwounds☝️
You really didn’t want to leave him a chance, huh?
Did you feel anger? Hate?
This game was such a masterpiece. I wish I could play it again for the 1st time.
Omg i was just telling someone that even though i avoid AI i wanted to build a bot to play Detroit Become Human! Thanks for saving me the work and the knowing of how much water i could be using XD
Yeah that’s something I will address in the finale so hopefully see you there !
man, why do you have to hang us like that on a cliff
Better to hang than to slip
This video idea is so good. Can't wait for pt2 when it comes out I'll be there in the first 5 mins
Also that ad read was too smooth wtf...
Looking forward to seeing you there ! And thank you! I know ads aren’t always in the viewers interests so I wanted to have it smoothly integrated somehow lol
This is such a W peak good and great RUclips gaming challenge video lol man
Why thank you 😊😊😊
@GamePsychYT Your welcome GamePsychYT 😊😊😊
ChatGPT and other AI knows about the story of detroit become human
I couldn't understand why this has so many downvotes.... but then I saw you leaving the fish. 😭🐠
I have failed the fish :(
It is so depressing that Chat Gpt made better choices than I did 😭
How 🫠
I think samgpt already knows the good options 😆
I love this game we need a sequel.
400k views. Congratulations, man. Im sure you are over the moon with this one 💪
Thank you!!! It’s actually crazy how this video has blown up. This video literally got almost the same amount of views as the whole channel in literally four days. I can’t get my head around it. It’s just really nice to see all the support for this vid and channel 😊
@GamePsychYT You deserve it 100%, man. I can't wait for the next one 😁 Let's gooo
Now that I think about it...there really should have been a human playable character so the players can really choose between humans or androids
absolutely love this idea and how dare you get me fully invested mow. xD
love it but maybe try and see if you can make a new chat gbt a little feistier, seems like Sam was too chill. until he said those couple things lol but i think
Sam needs just a tad more spice. cant wait for part two!!!!
Yeah Sam a chill guy for sure even when things do get a bit spicy 😅
amazing video, you deserve 100,000+ subscribers
Thank you, that’s really kind! maybe one day it will happen 😊
Sam is very much in the mindset of "It's just a game, who cares we steal or maim innocent people." ʕ•ٹ•ʔ
This needs a part 2
idk i feel like chatgpt already knew the perfect ending for detroit become human since it already got fed info on it.
Let me be FULLY transparent when I say this: I love the game Detroit become human... I have played it so many times. Legit got it for my birthday when it came out and finished it in like 3 days! Amazing video! 😊
Thank you! Great to hear that you loved the game. It’s honestly such a blast, thinking of re playing it but making the most cursed decisions lol 😅
In my opinion Detroit becoming human is an impossible future not because of the ai uprising but because it depicts Detroit as a somewhat functional city, that’s never happening.
Cool concept an ai choosing an ai's choice ❤
Even wilder concept
An ai, making a choice for an ai, but the options were written by french
Detroit became humans is the only reason I like story based games. It's fabulous 😍.
WHERE'S PART 2?
Soon.. it shall arrive
Neat seeing an AI give its input in this game. I didn't like the game myself. It felt like the robots were just humans. Why give them sapience and emotions, then treat them as slaves? There's no philosophical question when the robots are clearly self-aware, they need to have the same rights as anyone else. It only raises interesting questions when there's uncertainty on if the robot is actually self-aware, or if they can care about how they're treated.