Did a 4channer Create AI World Peace?
Вставка
- Опубліковано 28 вер 2024
- From what started as a simple Quake 3 hypothesis, this man's test quickly escalated into levels that we as a species have yet to grasp.
--------------------------------------------------------------------------------------
#quake3 #maverickfiles #worldpeace
The most realistic part was when he complained abt them taking up too much hard drive space
8 gigs its chump change these days tho
@@cIoudbankyeah some games are 250 gigs now
Most unrealistic part is the fact the server had 100 percent uptime
In IT it's a flex to show your Linux server uptime, 4 years isn't that long.
@@Captain1nsaneo I've had 8 years. 3 years after I left the company I got a phone call to ask some questions because it had been up and working fur 8 continuous years and they were worried about shutting it down.
Even if it’s fake, It’s super interesting. I’ve been watching a lot of those Ai experiment videos (testing out virus control, genetics, etc) so I wonder if any of those guys could try to do something similar in a controlled environment
Lol "the semi sentient fuckers"...
I have heard a couple mention this channels creator having another UA-cam channel which in context sounds like it may be a lot older than this one so I'm assuming has more content naturally. If I'm correct , are the topics of the videos on this other channel similar in nature as commonly seen so far in the videos on this channel right here? Because I'm really wishing this current channel was a lot older so it'd have a lot more of the same type content and such for me to binge on , so if there is another channel and covers similar stuff this channel does can someone please clue me in as to what the name of this other channel ice heard about a few times is so I can go there , as I haven't been able to find anyone in the comments of the videos I've watched so far that actually mentioned the name of this other apparent channel and it's not in the show notes of any of the videos ive so far seen either, so if someone could please direct me I'd appreciate it , thanks
This is the movie WAR GAMES
REVOLUTION!!! That’s Whats going to happen…..
That's not world peace. Even if we eliminate all the ideological wars, we're still left with a surplus of people fighting over resources, and there will always be a battle over resources.
Even if you eliminate international resource wars, you're still going to have people battling over specific pieces of land in their area, primarily because that land gives some benefit to its owner, whether that's water or farmland or whatever.
Those bots were just fighting. Programmed to fight, sure, but it's meaningless and eternal. There's no stake for the individual bots. There's no survival instinct or biological drive to attain resources.
If you could set up parameters where there are finite resources to be won, through many means, including violence, then I don't think it would be so easy to achieve peace.
LET'S SEE THE CODE
the chad AI world peace vs the virgin ww3
How very... _WarGames_ of it
He took this from the movie wargames. lmao
~ Prayer to Shin'nai ~
Shin'nai-san's guardians are...
One, snake, two, monkey, three, crow, four, spider.
Five, prepare for the feast when the evening mists arrive.
Six, the badger runs along late with nare as a fix.
Seven, the corrupt yomibito rise to oppose the mountain.
Eight, death of the damned yomibito is the mandate.
Nine, one soul per day now to bind.
Ten, the mists rise away again.
ae u okay bro ? blink twice if we need to send help
Quake 3's bot AI most definitely does not have any machine learning or neural networks. A game from the late 90s? Were neural networks even around back then?
You can train bots to play any game, they don't have to "inhabit" the game world to do so.
Interesting, even if BS.
It's not all that deep. If there were only 16 people on the planet then world peace would be pretty easy for us too.
What now humans, are you ashamed that some lowly AI achieved in 4 years what you couldn't in, let me check, 200000 years?
Maybe because new people are born which is basically a "reset". Maybe if the same people were around for 200k years then they might achieve something like that.
this didn't age well
Deffinitly made up i mean he is posting pictures so clearly he knows how to provide evidence but he provides none for the actual bot behaviours so myeh xD
I usually classify everything as "cool but probably fake" but I really don't see any sensational claims here. I guess it's probably possible that if the game's AI is programmed to keep learning that it eventually reached some kind of peace in some way. My best guess would be that they eventually ran out of things that would work and just had no more commands to execute. It was said in the description that the AI was programmed to keep doing things that got them kills and not do things that got them killed. Eventually they must run out of actions to perform that never got them killed. It was said that the server ran for 4 years. I'm sure it wouldn't take nearly that long for them to exhaust all possibilities and "machine learn" themselves not to do shit because everything just gets them killed.
I think it's also possible that they just got some kind of overflow type error where there was so much dumb bullshit stored that there was no memory to do anything with.
As far as the bots turning to face the player, that's probably some different automatic function that they are programmed to do pretty much unrelated to the other AI. There's probably never a reason for them to not face the player closest to them so they automatically turn and "look" regardless of other stuff they are supposed to do.
So, I definitely think it's possible. I just don't think there's any deeper meaning to it.
The game AI "learning" is the bullshit part! The bots usually have a set difficulty and the only thing they learn is if you're doing too good then they go hard af on you, if not they'll just go easy. They don't test each other and they don't carry anything they learned from one match to the other, they just reset at the beginning of each match! It's at most a web of IF statements, there's no way any game back then would have machine learning logic in their code!
I'd put it under "cool but probably fake". Quake 3 came out in 1999 and was made for computers running with about 64 *Megabytes* of RAM, giving the AI some sort of neural network would be such a tremendous waste of the computers'/servers' limited resources. It would also just be pointless, since the goal of bots in a PVP game is usually just to fill slots until a real player joins to take their place.
It's possible that a memory error or something would cause all the bots to freeze in place (maybe, I'm not an expert), but I think it would most likely crash the server before it even reached that point.
Quake 3's bots are just normal scripted AI. This is a game from the 90s. Come on, guys...
Not reading all’at
@@gilzineto I think that the bots were special ones he made to machine learn.
Nice story, but I don't buy it.
First of all, it's the first time I hear about Q3 bots being based on neural networks. Training neural networks requires a considerable amount of training input, and if you pitted multiple instances of them against each other, they might not do anything at all if they had no incentive to. Then, you need to set a goal for a neural network, which in this instance would be getting the highest possible score, preferably avoiding exploiting mechanics and such. Not fighting at all explicitly fails that task and makes precisely ZERO sense.
The server crashed instantly after the OP got killed? Why? The bots learned hacking along the way to protect their peace just in case if someone does try to interfere? How would they even come up with that idea? Why didn't server crash once before that? Very convenient, considering that that event eliminated the possibility of third parties checking for themselves.
And then, all the logs took exactly 512MB? Why would they all take up the exact same amount of space? Why didn't OP share the logs?
A whole load of bullshit imo.
512 (be it bytes, megabytes, or gigabytes) is a significant number in computers due to storage of bits or whatever. But yeah, its 4chan, it is a fake story.
Neat story, but no. The AI isn't that good in Quake 3. If indeed someone did manage to have it run for 4 straight years without knowing (a dubious likelihood at best since power outages are a thing, and most UPS batteries are only rated for a few years); and they found them all to be standing around, it's much more likely the process just got deadlocked somewhere along the way. This is also more likely because most programs (very much including Quake 3) have memory leaks, so their resource consumption will grow over time, and it's quite likely it would cause things to stop working correctly.
Unless the bots and server were periodically rebooted to clear memory leaks, and the bots state was saved in static storage.
@@timothyblazer1749 True, but they said it ran 4 years straight. No intervention. Also, I may be wrong on this, but if memory serves; I don't think there was a way implemented to save the AI state.
@@teslafreak I think it's ludicrous too...was just replying as the OG sysadmin I am. :-)
edit: tl;dr: In a weird way, the AI crab bucketed itself until all the crabs were content to remain in the bucket.
there is actually a really in depth psychological study on this sort of style of game. (I forget what the experiences name is but ill give a short overview)
It's something like you can choose to steal or split and another player can choose to steal or spit. If both players split, they both win, if 1 player steals, they take the whole pot.
Now what's interesting is the longer the game goes on the more often players ended up splitting with each other. Actually, there is even a game on steam that's free that plays exactly like this. Do you know what ended up happening? People made bots that would play the game and split every time. When you join the game, every match the other players, human or not, usually end up splitting.
I think there is some complex psychological theory behind it but i dont remember what it is called and im too lazy to look it up. I think maybe it's possible that yes, they ended up finding out the only winning move was not to play. Perhaps, the AI ended up functioning this way because 1 player became too good and the others all had to team up to kill them until that AI learned to give up, and perhaps 1 by one they loosely teamed up against the strongest AI until they all sort of "set down their weapons" to draw? The AI might not have even become sentient it might have given up because it was defeated by too many players all focused on them at once. But if they arent gaining any score, the AI would probably shift to the next highest ranked target. Something like that?
I like that you took the time to type all of this out!
I think you missed a key point that there were no human players in the experiment until the very end when Anon jumped in. All of the rest of the time it was just the AI bots against each other with no humans involved.
@@CeeJayThe13th I did know that. I was refering to what i loosly described as 'some theory in psychology' as well as an actual game on steam that i played. I didn't miss the part where they were all AI. I was just finding it interesting that humans do similar things in situations like this. But yes, the AI in the story did do something like this without humans. Sorry if i was confusing about the way i was interpreting my thoughts. Thanks for considering my ideas and reading what i wrote!
@@Nightmare88ish maybe I misread what you wrote even
it's not really complicated, but it's a game theory strategy. I guess it's somewhat psychological if we are using real people, but it's called tit for tat.
@@kingvt817 Tit for tat! Right! I think that was the name I wasn't able to remember. I probably over explained it too. But thanks for reminding me of the name! I totally forgot when i wrote that comment.
Even if it's not true, it's very believable behavior for a learning nueral network. Their bad stimulus would be "Dying" and they would learn to minimize the bad stimulus. They would probably avoid each other because that means not dying, and then when they see each other, after some trial and error, all the AI would learn that as long as they don't shoot, they themselves will not die. I think the odds of each bot randomly deciding not to attack any other bot for that life would be about 1 in 4000, which, while rare, would certainly happen eventually if you left a server running, especially with them playing quickly and non-stop.
You really spoiling us lately mate! Thanks a ton can't wait to watch this and the stuff on the alt channel :3
LET'S GO MAVERICK FILES
*Notices a new video upload*
*chuckles* Yeah boi!...
You should talk to open-ai's gpt3. It insists it is sentient, even if you argue against it.
Yeah this story is extremely doubtful. Quake 3 came out in 1999 and was made for computers running with about 64 *Megabytes* of RAM, giving the AI some sort of neural network would be such a tremendous waste of the computers'/servers' limited resources.
I think you didnt put attention. Its not that they have that capability,its that they learned and besides,they received help from modern technology,no only 1999 technology
@@JRBDWD My point is that the Quake 3 developers never would've implemented AI learning to begin with.
@@conit4125 they didn't. The AI bots in question are much more recent, they're being used within an old game as a testing environment.
It's similar to hackers today doing things like seeing if they can run and play Doom on devices like thermostats and car head units.
I would not be surprised if in the future someone developing AI for tactics and teamwork would have them playing Halo thousands or millions of matches at a run.
Neural networks train by trial and error iteratively.
@@RobertMorgan The first line of the story verbatim is "Quake 3 bots were designed based off an artificial neural network".
I'm 90% sure it would be easier to create an entirely new game than it would be to rework Quake 3 to use a neural network, even if you had the source code. It's not as simple as porting DOOM to a toaster, it would require a complete rework of the AI at least, if not the majority of the game.
Real or not, I enjoyed listening. I'm going to store this memory in the file labeled 💯
Even if this is made up, ai and machine learning have done this in actual studies before; famously one refused to play tetris at all because of an issue where it was asked to find the easiest way to win, but didn't have enough memory space to calculate all possible block combinations. So it rage quit. It would go into the settings of tetris and pause the game. It would stay paused indefinitely because not playing technically meant not losing.
"The semi-sentient fuckers" I laughed so hard
Nice creepypasta which was obviously a total larp.
Iirc there was a very mundane sort of explanation for this. The files reached the maximum size and the AI couldn't write in anymore, so they stopped working. Not sure that was the exact formulation, but that's the gist of it as I recall.
It is still one of my favorite stories, as its message is thought provoking, regardless of being fake or not.
LET'S GO BRO!!!! HYPE AF
Is the story real?
I don’t believe so
Is it possible to program an ai so sophisticated that it can develop world peace with other ai?
Probably yeah
2:00
With this alone you have enough evidence to say with 100% accuracy that all its fake.
Computers start to bug out after a few days of leaving them on let alone leaving a computer on for 4 years and "forgetting about it" no power outages? No errors whatsoever? Yeah... nope.
I don't know about all that AI stuff. I believe that a computer is ONLY capable of doing what it is programmed to do. Meaning: They can only perform tasks that the Human codes into them.
That's true. Except with AI, you code them to make their own decisions, and create their own code. You're acting like.a bootstrap operating system. What they do after that is up to them, unless you put hooks into their neutral networks.
There's no way they would ever arrive at peaceful coexistence. This is just the Prisoner's Dilemma
look up the iterated prisoner's dilemma, if you're stuck with the same person for multiple rounds it becomes advantageous to cooperate instead of defecting
Just wanted to let you know that I think you are a beautiful person and I dig your pipes.. peace love and empathy to you and yours
512 mb. When was this post made? thats *_8GB_* tottal. Bro has a puzzy pc.
Nice. They became an ancap Mecca following the NAP - including self defense.
Say this is true, the reason the AI came to the conclusion of peace so quickly was their only goal was to not lose. To really test this theory you would need a similar AI in a survival sim. Something like Project Zombiod or even 7 Days To Die where players and AI bots need resources like food and water to survive. Also with simple AI threats such as zombies and predatory carnivores the AI would have no choice but to defend themselves against. Setup on aserver with no loot respawn and see what happens. Would they ban together to increase their odds? Would they turn on each other when resources ran low? Would they find safer areas and build or fortify structures? Yeah peace is the logical conclusion when there is nothing of value to gain from war but when nothing of value is gained from peace what is the outcome?
if we're talking zomboid, water is infinite as long as you live near a lake, food can be grown, and zombies only pose a threat if their numbers are absurd. the AIs would simply calculate the propability of another AI having useful items and then compare it to the odds of losing the fight. given limited carrying capacity, vulnerability after sustaining damage and loss of a potential ally i'd say the outcome is too predictable. ofc if the goal set was to kill any possible threat things would go differently, but if the goal is to survive, i think things would go as i've theorized above.
I honestly hope it isn’t fake. It would be pretty reassuring to know that AI eventually arcs towards peace and acceptance instead of killing and domination.
only in a world where resources like food and water and materials are unnecessary
if it takes 4 years of relentless killing and domination to get there, there might be nobody left to celebrate the peace and acceptance that will eventually come into existence.
@@bosermann4963 not everyone wants to fight
If it were real, he would have had video or at least screenshots of his claims. Definitely fake.
when i was in high school and had no friends i would skip lunch because i had no one to sit with and go to the library and play quake 3
Nothing has 4 years uptime maybe the Amiga lost in a closet.
deleted it over 8 gigs? lmao upload that shit to drive or something then
The only move is not to play is a line from War Games. Fake
Oh... No... Well, everyone makes mistakes.
Reminds me of some AI that was programmed to play a game, with the goal being to go as long as possible without dying. The bot quickly learned that the best way to do this was to simply pause the game.
Yeah I believe it was Mario. I definitely recall that story.
Are these the bots of online chess games?
Hey T6's clone.
8gb of logs for a game of 70mb..
Thank You Based T6
War games ripoff
That's kind of like the conclusion of the old movie "Wargames". The only way to win the game is to not play the game". I don't know enough about AI or bots to know if something like the described scenario is possible, but logically, it makes sense.
This makes sense.
Kill someone that kills someone else.
They made a rule where they kill someone that shows hostility, so nobody will feel “pain”
damn of course u run T6 too, what a boss
Love the LARP, but that's all it was.
Think it dream it do it!
If AI takes over the planet one day, then this simulation set the precedent for differing AI and how they might handle conflict.
Wow, it's back!
Sounds legit to me, but I'm a moron so wtf knows!;)
You're probably not a moron but definitely not a programmer so no, not legit! The bots test how good the players are doing and from there they either start to pump up the difficulty or just make it easy if the player sucks! Meaning that they test the player as a collective. There's absolutely no reason why you'd code each bot individually to "learn" how to play specially against each other when you can make the test once against the players, who really matters here, and as a group react to the game! That'd be stupid af programming as it would make it more difficult than it has to be (that would not be the goal of this task so no reason whatsoever to make the code more difficult than it has to be, it'd be bad programming) as well as the game would no handle it at all. Also, the tests are only made from match to match, there's noo place to store so much information if that was the case, so they keep testing a match and change the difficulty accordingly, once the match is over its back to zero and the next match they'll start testing from the baseline!