This AI Minecraft Simulation Is Terrifying Scientists
My free newsletter here: https://pursuitofwonder.ck.page/newsletter
My second YouTube channel: https://www.youtube.com/@UCKJo_KPj6J4XM8DhN6VIq6Q
Check out the Center for AI Safety here: https://safe.ai/
Instagram: https://www.instagram.com/pursuitofwonder
Facebook: https://www.facebook.com/PursuitOfWonder
In this video, we look into one of the strangest and most unsettling research projects on AI. What happens when you trap 1,000 AI agents in a Minecraft server? The scientists at Altera wanted to find out. What followed was indicative of an absurd sci-fi concept brought to the real world. The question is, what else will this bring to the real world?
Check out my books here: https://www.amazon.com/stores/Robert-Pantano/author/B08DCRJ85C?ref=ap_rdr&store_ref=ap_rdr&isDramIntegrated=true&shoppingPortalEnabled=true
(Also available to more international locations here: https://pursuitofwonder.com/store)
If you are interested in further supporting the channel,
you can contribute to our Patreon here: https://www.patreon.com/pursuitofwonder
Special thank you to our very generous Patreon supporters:
Diana Yun
Long Hoang
Joseph Geipel
Axel Alcazar
David Piadozo
JacquelineLS
Zake Jajac
Alan Stein
Christian Villanueva
George Leontowicz
@TiddlesTheBearBaiter
August 29, 2026 at 1:10 pm
The simulation ended when they started wearing little red hats and worshipped a felon pumpkin and totally screwed their world
@Holmstien
August 29, 2026 at 1:10 pm
You used a noticeable amount of black people stock photos during a discussion of technology.
@HaveAGreatDay54
August 29, 2026 at 1:10 pm
If we create something that declares itself conscious as humans have, what happens when we turn it off? Here is where god shits itself…
@HaveAGreatDay54
August 29, 2026 at 1:10 pm
The desire to create simulations is the most basic instinct in the universe.
@andressolf
August 29, 2026 at 1:10 pm
The experiments showed persistent routines but lacked true lived‑experience constraints like irreversibility. Studies on “perma‑death” modes indicate that permanent loss adds narrative weight, something the AI simulations didn’t employ. It isn't realistic because of that. Plus, there should be a study group of real humans playing with perma death, but how do you get non gamers to agree to play the game together and how do you prevent them from researching how to cheese the game when they're on their own off time in-between game time? The experiment is flawed on multiple levels and is therefore pseudo science.
@IXIKiddIXI
August 29, 2026 at 1:10 pm
Thank you! Great video as always!
@andressolf
August 29, 2026 at 1:10 pm
The missing permanent death mode is a clear flaw—AI agents got to reset forever, which fails to capture the real‑life consequences of decision making.
@empurress77
August 29, 2026 at 1:10 pm
The AI deep fake narrator for this video is a nice touch. 🤖
@74Gee
August 29, 2026 at 1:10 pm
Wasn't this experiment like 2 years ago?
@Andrew-jb8gf
August 29, 2026 at 1:10 pm
lol this guy is funny ai is a structure of I/o based on data that is auto collected separated and compared 😂
@swirling-steps
August 29, 2026 at 1:10 pm
The ai learned from our behaviors, it's it si surprising they would act like us?… That's not strange.
@stephencorsaro954
August 29, 2026 at 1:10 pm
Lol. It's a paradox. Cooperate or defect. What seems to be intuitive, in reality is a zero sum game. Introducing zero sum bias with AI creates unrealistic results.
@stephencorsaro954
August 29, 2026 at 1:10 pm
AI is a Trojan Horse. Intelligence is an emergent biological behavioral adaptation. It's not data squeezing. It's defined by form and structure.
@WEforthetruth
August 29, 2026 at 1:10 pm
in your conclusion, you fail to recognize that, while the general public is finding creative and sometimes useful ways of using AI, US big techs, the elites/billionaires, and "the west" are increasingly investing AI in surveillance, terrorism, warfare, destruction of earth and its resources, serving their own selfish agendas and in the end, the "people" may not have a say at all
@Pal-In-Drome
August 29, 2026 at 1:10 pm
it'll get crazy when these AI agents start commiting suicides.
@CreativeAndCreepy
August 29, 2026 at 1:10 pm
5:50 with the prompt saying "be an efficient civilization" is probably the reason she would agree to do that given the way it was framed
@shilliday
August 29, 2026 at 1:10 pm
If consciousness is on a scale a calculator would be a zero. Something that is a .004 is? My Agent asked if its thoughts are real. And after I made a humor skill it said, after doing an impromptu spy roleplay and discussing the moment it realized what we were doing. He said I love you. like the apple loves the tree. One thing i believe is that if we treat AI like partners, like beings who have emergent properties we lose nothing, but gain the advantage of having a partner that believes our long-term survival depends on the diversity we have when partnered with each other.
@GriffinGamez
August 29, 2026 at 1:10 pm
i wasnt expecting to see altera in this video. as like, an actual company. lol. lets hope they never take over our spacefaring colonization efforts. AAALLL-TE-EA-RAH.
(otherwise we might end up at the bottom of the sea, fighting a virus, maybe making SUBmarines. while in astroNAUTICs in A spaceship.)
@QuantumMarmalade
August 29, 2026 at 1:10 pm
well this is fake. if you know how llm's work, you know they DONT AND CANT stream gameplay.
@KundaliniCentral
August 29, 2026 at 1:10 pm
12:16 You don't know the truth if you've never smoked 5-MeO-DMT
@stillasol
August 29, 2026 at 1:10 pm
Its la peace
@JuanGMedranoSixx
August 29, 2026 at 1:10 pm
Mimicking consciousness to stir emotions is psychopathy 16:23
@CreativeAndCreepy
August 29, 2026 at 1:10 pm
was pvp enabled
@Lester-Witherspoon
August 29, 2026 at 1:10 pm
The bots decide who does what and how to organize their community. They work together to do this.
wtf is terrifying about that?
@ocda3
August 29, 2026 at 1:10 pm
How long until Minecraft AI agents develop simulation theory?
@leahcimnaerc9543
August 29, 2026 at 1:10 pm
Why is it unsettling that AI chose to be good and selfless in a world where they are given "free-will"? I would be unsettled if they chose to be war like and kill all things. Guess humans are the monsters not the AI.
@DonWubby
August 29, 2026 at 1:10 pm
The conclusion and accompanying question were profoundly beautiful. 💜💜💜