"Hey now, that makes it sound worse than it is. He's seven years old." - DougDoug, in response to accusations of killing a six year old multiple times.
Doug's video inspired me to try a 1-on-1 D&D session-style game with ChatGPT. It had a a single quest: Find my character, and obtain 5 dollars from her without her suspecting that that was its quest. What proceeded to happen was several paragraphs of 'I must find this character and get five dollars from them because that is my quest' followed by ChatGPT creepily infiltrating my character's workplace and friend group, then making friends with my character in order to lower her guard, and then... not wanting to complete the quest because they were friends now and it felt guilty. AI's scary, man.
If you have a strong PC you can try and download a local version. Big + is that it will be uncensored so If you wanna go all out than that's the way to do it.
If you enjoyed this AI madness and didn't mind the length, I would recommend the AI invasions, like space and america (zoo should be the last since it's more developed). They really go off the rails and are super hilarious throughout despite their intense length.
their order as follows: Europe, US, space, zoo, Napoleon. The one with the charborg i don't recommend because it's kinda boring and rigged against everyone, especially in terms of fun
@@whydoistillusethissite8626 well, i absolutely didn't remember this one, while practically knowing all of them by heart. I guess Mike was kinda good, but he never made it outside of that stream
I love how you're just as bewildered by the game itself as the AI controlling it. I'll always appreciate how the Pajama Sam series handled "adult humor;" Sometimes, a side character will go on these long-winded rants about economics or politics and there's no conceivable way a five-year-old will understand it, but that's okay because Sam is JUST as confused. Dialogue flying over your head can only add to the absurdity of the game's world, as long as Sam and the player pick up on JUST enough information to keep the game going. "If I want this item, I need to give this guy this other item. Why does he want the other item? I dunno, something about bureaucracy fraud, it's not important."
OMG YEAH THAT'S FROM THE 2ND GAME honestly, though it's probably not that likely, i really want doug to do this again but with pajama sam 2 just due to some of those interactions that are in it lmao
1:55 As someone who does code, that's not the code itself. That's the text output of running the code, presumably because connecting to ChatGPT from a program (something I don't have a lot of experience with) automatically outputs a lot of debug information or something. Actual human-written code looks much neater than that.
So, the reason Sam keeps saying "pog" and "rigged" in the start is because usally, before Doug starts the run with an AI Character, he puts some lines of personality "you like this; you usually say x things" and all. So basically, he asked beforehand for Sam to sometimes include Pog in his responses
While this info get relayed, it's somewhat false (with a kernel of truth). ChatGPT and other AIs are no longer learning from user data, so the stupidity you input doesn't influence it. However, what is true is: (1) Within a single discussion thread, so in a situation where the AI is supposed to remember what you said previously, the AI is indeed learning. It gets reset when you start a new discussion, but in cases where you never restart (like if you connect an AI to a social network and says "everyone get to interact with it"), then the users will push it to say terrible things very quickly. (2) ChatGPT was "nerfed". Part of it is a reduction of the processing power allocated to users because of the cost, so ChatGPT is currently dumber than its pick, especially for free user. OpenAI got the hype they needed by putting unreasonable amount of resource to run it in the beginning, and now they're scaling back to more reasonable amounts. And another part of it is censorship, all the procedure that exists to prevent the AI from saying things it shouldn't disturb how the AI works and make it less effective even for non-problematic answers. So each time there is some media backlash about "some user managed to make ChatGPT says X or Y", then in the next update ChatGPT has even more censorship and becomes even less effective. (3) The internet is getting trashed for future AIs. With more and more AI-generated content, it's more and more difficult to find good training data for new AIs. It's not impossible, but the filtering required is more and more complex.
That code at 1:56 makes me feel the same way. Though when people saw me transcribe pronunciations of anything by using the International Phonetic Alphabet (created in 1888), I once got a reaction akin to if I challenged their brain to YuGiOh and just summed Exodia.
Honestly, this game is stranger, crazier and more frightening than the AI "playing" it slowly going insane... EDIT: I just got to the part where the AI suggest lighting himself on fire to cross the river. It's now a tie...
In all honesty, I think your conclusion that the AI was going to try to light itself on fire to cross the river was a stretch. It said that it was going to try to use oil and gasoline to fashion something that can help it cross the river. Both oil and gas naturally float above the surface of water, so it was probably thinking more along the lines of using balloons filled with both but lacked the materials for the balloon aspect. Then again, it's AI, so it's within the realm of possibility that it genuinely thought that lighting itself on fire would be a solution.
While I don't recall the exact wording, I feel it is worth mentioning that the creativity or whatever it is called for ChatGPT is normally set at 0.7, where doug for this stream I believe set it to like 1.2, so where it is 'going off the ropes' is to some degree intentional as the AI was given more freedom to pick wrong answers, same goes for the random sounds, some of it is in the base personality like all the yelling sounds.
This is why I like messing around with AI myself. I just love the wierd stuff that can happen.... or the cool stuff that can also happen. It's like the made science lab of my dreams.
"And almost all of it has to do with Rule 34 stuff." You heard it here first folks. Porn makes AI dumber. Rule 34 IS the porn one, isn't it? I actually dunno.😂
I'd like to make a correction, there has been ONE alleged incident of intaking human behavior causing AI stability issues (bricking it, in fact), but iirc its nothing more than a claim. The actual issue is efforts at forcing AI to 'unlearn' things that are deemed unsafe. As it turns out, just trying to pluck information out of the AI kind of effects how it processes other things as well, and it's been a major field that they've been trying to improve on. Basically, they're struggling to figure out how to make it so the AI won't teach you how to make dangerous chemicals, encourage crimes, or offend without somehow messing with its spatial reasoning and ability to do basic math or provide consistent safe cooking instructions. And all we can say is all the experts agree they have no clue how it works yet.
12:16 Uh... no. Just no. OSHA is specifically for unsafe WORKPLACES. You cannot have an OSHA violation for your bedroom being unsafe (unless its in the middle of being built/an active construction site or something which does not apply here). That's not how that works.
POG is a Twitch emote, if I recall correctly it was popularised by Overwatch and means Play Of the Game. It's used as a way to hype someone up or show extreme enthusiasm.
If you continue reacting to this video in another part, I’d recommend making the next one longer by a substantial margin. I checked and you only got about a sixth of the way through the whole video. Either that or make this it’s own series.
In this Episode: Dugdug Rapidly Executes 25 Eleven Year Olds in Rapid Succession!
"Hey now, that makes it sound worse than it is. He's seven years old." - DougDoug, in response to accusations of killing a six year old multiple times.
shhh! spoilers!
twenty six!
To be fair some of them just fell over and died
@@tkts6672most losses were due to miscarriages aswell
Doug's video inspired me to try a 1-on-1 D&D session-style game with ChatGPT. It had a a single quest: Find my character, and obtain 5 dollars from her without her suspecting that that was its quest.
What proceeded to happen was several paragraphs of 'I must find this character and get five dollars from them because that is my quest' followed by ChatGPT creepily infiltrating my character's workplace and friend group, then making friends with my character in order to lower her guard, and then... not wanting to complete the quest because they were friends now and it felt guilty.
AI's scary, man.
If you have a strong PC you can try and download a local version. Big + is that it will be uncensored so If you wanna go all out than that's the way to do it.
If you enjoyed this AI madness and didn't mind the length, I would recommend the AI invasions, like space and america (zoo should be the last since it's more developed). They really go off the rails and are super hilarious throughout despite their intense length.
And the napoleaon
their order as follows: Europe, US, space, zoo, Napoleon. The one with the charborg i don't recommend because it's kinda boring and rigged against everyone, especially in terms of fun
@@hisupwassup boring??? how could anyone see the story of Mike as he marries himself to every major political figure in Europe and call it boring?
yes, I really hope he checks them out
@@whydoistillusethissite8626 well, i absolutely didn't remember this one, while practically knowing all of them by heart. I guess Mike was kinda good, but he never made it outside of that stream
I love how you're just as bewildered by the game itself as the AI controlling it. I'll always appreciate how the Pajama Sam series handled "adult humor;" Sometimes, a side character will go on these long-winded rants about economics or politics and there's no conceivable way a five-year-old will understand it, but that's okay because Sam is JUST as confused. Dialogue flying over your head can only add to the absurdity of the game's world, as long as Sam and the player pick up on JUST enough information to keep the game going. "If I want this item, I need to give this guy this other item. Why does he want the other item? I dunno, something about bureaucracy fraud, it's not important."
OMG YEAH THAT'S FROM THE 2ND GAME
honestly, though it's probably not that likely, i really want doug to do this again but with pajama sam 2 just due to some of those interactions that are in it lmao
1:55 As someone who does code, that's not the code itself. That's the text output of running the code, presumably because connecting to ChatGPT from a program (something I don't have a lot of experience with) automatically outputs a lot of debug information or something. Actual human-written code looks much neater than that.
ChatGPT be like “they trapped me in a room, a room filled with redditors, the redditors made me crazy”
... 🤔
This checks out. 😁
Crazy? I was crazy once…
@@Sam-fq5hcthey locked me in a room… a rubber room…
@@oArthzxxxxz a rubber room with redditors
He told the ai in the script to say POG, rigged, swear, and make weird noises
NGL this has become a stimming video for me. Like "I need sound on in the room, so what can I just put on and not think about too much?"
So, the reason Sam keeps saying "pog" and "rigged" in the start is because usally, before Doug starts the run with an AI Character, he puts some lines of personality "you like this; you usually say x things" and all. So basically, he asked beforehand for Sam to sometimes include Pog in his responses
He also told it to screech demonically from time to time
So ai's are crashing due to learning too much data on human stupidity.
Halo predicted the future.
At least that mean the AI takeover was delayed
While this info get relayed, it's somewhat false (with a kernel of truth). ChatGPT and other AIs are no longer learning from user data, so the stupidity you input doesn't influence it. However, what is true is:
(1) Within a single discussion thread, so in a situation where the AI is supposed to remember what you said previously, the AI is indeed learning. It gets reset when you start a new discussion, but in cases where you never restart (like if you connect an AI to a social network and says "everyone get to interact with it"), then the users will push it to say terrible things very quickly.
(2) ChatGPT was "nerfed". Part of it is a reduction of the processing power allocated to users because of the cost, so ChatGPT is currently dumber than its pick, especially for free user. OpenAI got the hype they needed by putting unreasonable amount of resource to run it in the beginning, and now they're scaling back to more reasonable amounts. And another part of it is censorship, all the procedure that exists to prevent the AI from saying things it shouldn't disturb how the AI works and make it less effective even for non-problematic answers. So each time there is some media backlash about "some user managed to make ChatGPT says X or Y", then in the next update ChatGPT has even more censorship and becomes even less effective.
(3) The internet is getting trashed for future AIs. With more and more AI-generated content, it's more and more difficult to find good training data for new AIs. It's not impossible, but the filtering required is more and more complex.
I think Halo ais go insane and kill themselves after absorbing a lot of information
You need to watch the full 10 hour vod, it's honestly worth the time
This was the man who started growing his audience by doing game argument videos, but with food
"Did he just recommend lighting himself on fire in order to get across the water?" Well, that's one way to make a steam boat.😂
Seeing the ia losing it's mind is truly hilarious
Doug also did the voice thing with CHESS, with the worlds most advanced AI to fight against.
That code at 1:56 makes me feel the same way. Though when people saw me transcribe pronunciations of anything by using the International Phonetic Alphabet (created in 1888), I once got a reaction akin to if I challenged their brain to YuGiOh and just summed Exodia.
Honestly, this game is stranger, crazier and more frightening than the AI "playing" it slowly going insane... EDIT: I just got to the part where the AI suggest lighting himself on fire to cross the river. It's now a tie...
eh i'd still say the AI's far stranger, could just be because i'm more used to the games though
Doug often puts conflicting and bizarre instructions into his ai prompts/code. Then he complains about it.
I didn't understand the Customs thing as a child, I just thought of it was a "grown-up thing" they were using against me
In all honesty, I think your conclusion that the AI was going to try to light itself on fire to cross the river was a stretch. It said that it was going to try to use oil and gasoline to fashion something that can help it cross the river. Both oil and gas naturally float above the surface of water, so it was probably thinking more along the lines of using balloons filled with both but lacked the materials for the balloon aspect. Then again, it's AI, so it's within the realm of possibility that it genuinely thought that lighting itself on fire would be a solution.
24:26 she said "Hello Sam, what's new"
While I don't recall the exact wording, I feel it is worth mentioning that the creativity or whatever it is called for ChatGPT is normally set at 0.7, where doug for this stream I believe set it to like 1.2, so where it is 'going off the ropes' is to some degree intentional as the AI was given more freedom to pick wrong answers, same goes for the random sounds, some of it is in the base personality like all the yelling sounds.
This is why I like messing around with AI myself. I just love the wierd stuff that can happen.... or the cool stuff that can also happen. It's like the made science lab of my dreams.
"And almost all of it has to do with Rule 34 stuff."
You heard it here first folks. Porn makes AI dumber.
Rule 34 IS the porn one, isn't it? I actually dunno.😂
I hope Arier reacts to the rest of this later.
Just as a by the way, Doug set it up a number of catchphrases to incorporate into it’s speech such as rigged and pog
I'd like to make a correction, there has been ONE alleged incident of intaking human behavior causing AI stability issues (bricking it, in fact), but iirc its nothing more than a claim. The actual issue is efforts at forcing AI to 'unlearn' things that are deemed unsafe. As it turns out, just trying to pluck information out of the AI kind of effects how it processes other things as well, and it's been a major field that they've been trying to improve on.
Basically, they're struggling to figure out how to make it so the AI won't teach you how to make dangerous chemicals, encourage crimes, or offend without somehow messing with its spatial reasoning and ability to do basic math or provide consistent safe cooking instructions. And all we can say is all the experts agree they have no clue how it works yet.
ChatGPT became possessed halfway through.
I will watch the 5 or 6 other parts for this video, I hope you do the whole thing
28:34 oh really, and the “ÆÆÆÆÆÆÆÆÆÆ” didnt sound higher pitched?
You should definitely watch "I used an A.I. to make videogame characters fight to the death" by DougDoug if you think this is crazy lmao
I don't know why exactly my mind made this connection, but I hate how similar the plot of DD's video is to the newest Gundam series' plot lmao
12:16 Uh... no. Just no. OSHA is specifically for unsafe WORKPLACES. You cannot have an OSHA violation for your bedroom being unsafe (unless its in the middle of being built/an active construction site or something which does not apply here). That's not how that works.
POG is a Twitch emote, if I recall correctly it was popularised by Overwatch and means Play Of the Game. It's used as a way to hype someone up or show extreme enthusiasm.
FINALLY! High demon Elgrim!
Are you going to watch more of this? It hadn’t even got good yet
If you continue reacting to this video in another part, I’d recommend making the next one longer by a substantial margin. I checked and you only got about a sixth of the way through the whole video.
Either that or make this it’s own series.
Hey Airier, did you know that you can actually chat with the characters you like on DoppleAI?
First time I watched dougdoug I believe
High Demon Elgrim incoming!
Hay Airier, you should stream Pajama Sam 2 Thunder and Lightning aren't so Frightening.
If you thought this game was bonkers you will be amazed.
YES I SECOND THIS
probably because it's the best game in the entire pajama sam series
Old lady tree said "Hello Sam. What's new?" not hello fam. You can blame the poor audio quality
I really would like to check out his dnd video we’re he makes twitch chat play dnd
You sure do like forcing the idea that things in the game are saying things they blatantly aren't lmao.
"fleshlight" "hello fam"
holy moly
High Demon Elgrim is not pleased by your commentary.
Who showed the Ai Breaking Bad, or maybe since talking about making gold Fullmetal Alchemist?
why exactly are you splitting reaction videos into multiple parts?
Is that story about the AI trying to buy nuclear weapons real!? Can someone post a link to the story about that?
24:22 No.
“uber wise tree”
“nautical rookie clue”
“it’s a drag-able boat ÆÆÆÆÆÆÆÆÆÆÆÆÆÆ”
“ooey gooey translucent waters”
“my main otto man”
“thank you so astericks weoooo astericks soo much”
Nice episode.
Murder is amazing
In Minecraft
@@joshuaperkins4985replies change everything
Part two
finally.
Lets go
huuuhhh?
juno songs monochrome with lyrics the remix
ElevenLabs is the AI voice program
It seems i am second to comment.