Can You Beat an AI Generated Escape Room?

This is the most difficult AI

generated

Minecraft

escape

room

created by the best AI, ChatGPT4. And today I am going to try to overcome it and prove that nothing, not even robots, is superior to humanity. As? Well, ChatGPT will generate 10 different cameras for me to solve, and I only get 3 strikes. That is, if I can't solve more than 3, I lose. If I solve 7 or more, I win and humanity wins. I mean, surely this shouldn't be difficult, AI can't capture the human creativity needed to create a Minecraft

escape

. What the hell is this? There is not a single placeable block.

I have slime, thread, feathers and seeds and ChatGPT wants me to go up 18 blocks to that exit. There is no way this is possible. I went through my craft recipes and all I could make was wool and thread. Without a workbench I was literally helpless. Even using the chest containing the items that weren't technically on the list wouldn't get me any closer. I had no idea how to solve this puzzle and would have to give up one strike in round 1, leaving me with only two more until I lose the challenge. This was a terrible start, but I couldn't see how this puzzle was feasible.

So with a score of 0-1, I asked ChatGPT for the solution to maybe get an idea of ​​what I missed. Create tracks without a crafting table, create chickens from eggs, grow wheat on nothing, and float with chickens. This solution was stupid. This escape

room

was never possible to begin with. And this challenge just got ridiculously harder. The AI ​​has no idea what it's talking about, so I'm faced with 10 potentially impossible escape rooms. Which meant I had to do everything in my power to win. Any dimension, any version, any potential bug that can be abused is fair game.

He was going to win against the AI, no matter what, for humanity. In room number two I have 15 blocks of dirt, a fishing rod, 4 snowballs and 2 iron blocks. And my task, build 25 blocks. Simply piling my land would leave me 10 blocks shorter. And turning my snowballs into snow and using my iron would still leave me 7 blocks away. ChatGPT's solution was absolutely atrocious and defied the laws of physics, so I was completely stuck. It seemed that he once again had a completely impossible task. If it weren't for one thing. The bar. Behind pearls, boats, and water, the fishing rod is probably the most OP item you can offer in an escape room.

Because? Well, if I make a tall enough pillar and break blocks as I go, I could use the fishing rod to pull the dirt towards me. Well, at least in theory. And after messing around for a few minutes, I developed a trampoline setup that allowed me to jump up and collect all the dirt I used to accumulate. And with a little climbing, there it was. I got the blocks and made it up. The score was 1-1. If you wanted to achieve a victory for humanity, you would need to win at least 6 of the next 8 rooms. The third room was less simple.

I had a decent mix of redstone supplies, but not enough to simply exit the 40-block room. But the slime block and the pistons could only mean one thing. A launcher. I tried to jump in to see how far I would go. Not bad, but I quickly realized the problem. Once I cast it once, I would lose my slime block. I could keep my crowbar but that wouldn't mean much. Then I tried to break the slime block while jumping. If I could hold it before moving on to the next piston, I could launch all four times. The problem is that as a piston fires it becomes a different block called a moving piston where the slime block is in a limbo state while moving.

Which means it's unbreakable as it throws me. That was a dead end. It seemed like I wasn't going to get anywhere until this happened. Bud feeding or, if you want to make it sound smart, almost connectivity. This piston shouldn't work, but some poorly placed gravel tricked it into upgrading via a redstone quirk. If I could make each piston extend twice this way, it would give me twice the height per piston. And so, one lever, two extended pistons. And if I used my other lever, I could go all the way up. But was it possible? It'd be enough?

Well, I had to try to find out. I spent about 15 minutes playing with levers and pistons trying to find a bud configuration. I went to the wiki to read more and even watched my own video from 3 years ago on quasi-connectivity to get the hang of how it works. Except after all I realized, for the type of mechanics I wanted to use, the only way to get 2 blocks out of each piston was to try to run my lever through the gravel. And after multiple failed attempts I realized something was coming. Even if it had extra levers, even if it propelled me perfectly towards the chest I shouldn't have.

Even if all that worked in my favor, I would still be 10 blocks away. This room seemed completely impossible. I would have to take another hit. Putting me at 1-2 in the first three rounds. But he had one more ace up his sleeve. Version changes. I could play escape room on any version of Minecraft, as long as I just used what they gave me. So I went for an update. 23w13a_or_b, the voting update. If you don't know, pressing V in this update will give you the chance to vote on random game-changing proposals. Now I could wait here for about an hour until the vote I'm looking for is among the 181 possible votes.

Or it could simply speed up the inevitable for your viewing experience. Inflatable castle. Which turns every slime block into an infinite bounce house and makes the score 2-1 for humanity. With 7 rooms left to decide if humans are better than robots, ChatGPT failed completely. His ideal solution to this 20-block puzzle was to build a cobblestone generator and use this iron, and not sticks, to craft a pickaxe, extract cobblestones, and columns. What he didn't realize is that you can actually burn all those other items in the lava. And just water it. And just like that the score is 3-1 for humanity.

If I win a little more than half of the remaining games, I can seal the victory. But just when I thought I was getting better. What is this? ChatGPT dropped this absolute bomb of an escape room. 35 blocks with only 12 solid blocks to use. Even if I could hit myself with a bow and arrow to increase each quartz and could use the chest, both repeaters, the lever and the bow as stools, I would still be 12 blocks from the exit. I checked the Quartz wiki page to see if there were any vote update rules or any version of Minecraft history that would make Quartz useful.

There was nothing. He was totally defenseless and just like that, it was 3-2. If I lose twice as many of the next 5 escape rooms, I would completely lose the challenge. So the next room had to be a win. Immediately these materials screamed one thing to me. Snowman. There had to be some way to use snow golems to get out of this escape room. The room was 15 blocks high and all the AI ​​gave me were 8 snowballs, 2 pumpkins, scissors, and 6 dirt blocks. Simply accumulating pillars would not be enough, I would be 6 blocks away, and if I built my two snows I would be 4 blocks away.

The first step I thought I had to do was cut up at least one pumpkin. Without this, I wouldn't be able to spawn any snowmen, and while I wouldn't be able to undo them, I wouldn't actually be losing any blocks, and in the worst case, I can restart the room. But then what? Snowmen can make layers of snow that could theoretically be formed into pillars, but the only way to collect any of them was with a silk touch or a shovel. Breaking them with scissors or anything else wouldn't drop anything. He was already at a dead end, but there had to be something he could do.

If I wanted to

beat

the AI ​​and prove that humans are superior, I had to win this room. Losing two in a row would be devastating, I would run out of strikes and have to complete the next 4 rooms perfectly. So I started researching. I searched the internet to try to figure out what I was dealing with. The history of snow golems, of snow layers, of snow blocks. Maybe there was some dark version of Minecraft where you could mine layers of snow by hand, or some mechanic about snow golems that made this room possible. And after a quick Google search, I found the answer to all my problems.

Game. Snow golems have a chance to drop between 0 and 15 snowballs per kill. And making them costs 8 snowballs. So every time I make a snow golem, I can cut off its head to get the pumpkin back, kill it and if I'm lucky enough, get back at least 9 snowballs and harvest infinite snowballs to make more snowmen and eventually , make enough snow. to get out of the pillar. But I didn't want to rely on bad luck, I wanted to be sure that this was going to work. So I did the math. Using very basic probability, I found the expected value of the snowballs, i.e. if I do this infinite times, what is the average number of snowballs I get per kill?

If it was more than 8, it would mean that you would make a net profit of snowballs per snowman, meaning you could safely bet and get out. Stay in school kids, it will make you better at Minecraft. However, the math didn't lie. 7.5 snowballs. He would be losing an average of 0.5 snowballs with each snow golem he created. Which should make sense, since between 0 and 15 there are more numbers below 8 than above 8, meaning below is more likely, but it was still discouraging. This bet would be a losing battle. There had to be another way to resolve this. And after some more research, I found another way out.

Minecraft Alpha 1.0.5, a unique version of the game where snowcaps could throw snowballs without a shovel. The version lasted a single day before it was patched, and the snow feature was immediately removed in the next version. But all it took was one day, the version existed. He could harvest infinite snow from a golem. So I started Minecraft alpha to test it out. And there I was destroyed. In such an old version of Minecraft I could only play one of the 6 worlds that the game generates for me. I couldn't wait to transfer my escape room. This version was impossible to play and I was once again at a dead end.

If I gave up here, it would mean a game tied 3-3. I would run out of strikes and any future mistakes would be my last. But humanity has to move forward. So I investigated even more. And I discovered something. The answer was never in the snow, but in the pumpkins. In 1.11, snowmen were given a tag that allows them to keep their cut state after logging back in. Do you know what that entails? It implies that before that, between 1.9 and 1.10, they did not retain their cutoff status. Which means logging in again would regenerate the pumpkin. If I kept cutting and cutting, I would have endless pumpkins.

The problem was that the ability to harvest pumpkins from snow golems wasn't added until version 1.16. Which means that if I wanted to solve this puzzle and achieve a long-fought victory for humanity, I would have to hack the snow golem into 1.16, pick up my pumpkin, change the world version to 1.10 or 1.11, log back in for the pumpkin is re

generated

. Change the world version back to 1.16, harvest the pumpkin, and repeat the cycle. Growing endless pumpkins that I could use to stack. There was light at the end of the tunnel, there was hope for humanity. So, ready to carry out my plan, I tried loading my escape room in 1.10.

Except one problem. Reestablished the world generation. Minecraft worlds are not intended to be downgraded, and reverting versions would cause the entire terrain to regenerate. So transferring the world's files was not going to be possible. No matter where he looked, it seemed like he couldn't solve this problem at all... It's over. I didn't want to keep recharging the world until I had perfect luck. That didn't feel like a victory. The score was now 3-3. The worst it had ever been. If he lost one more time in the next four rooms, the challenge would be over and the humans would have lost.

The following performances had to be perfect. Whatever it takes. 28 blocks to get up. And only 12 blocks in my name. I was absolutely devastated. Of all the escape rooms that chatGPT could generate, today it chose absolute violence. This set of elements seemed completely impossible. Now you might be saying that I can let cacti grow and form that way. And yes, that's true, except that I play all my escape rooms at a speed of 0 ticks. So if there's ever a crop problem, you can't just put cheese on it and let it grow. And I assume chatGPT wants it done the same way.

So if I want to win this battle against the AI ​​in a respectable way, I have to find a different route. My first instinct was to increase the damage. If I could use the cactus damage to jump an extra blockin the air, he could make double use of each cactus. This came with two major problems. The first is that this mechanic is incredibly difficult to achieve in single player. Still, I had no problems starting the world on a server and getting some lag to boost it more easily, but then I ran into problem two. It didn't matter if he could increase, ultimately it would be more efficient to simply accumulate.

Because sand can't float at higher altitudes, any sand you try to propel would have to be placed on top of other sand. Which means there was no point in trying to conserve blocks by boosting. I could turn my sand into 2 floating sandstones, but that would leave me even more resource-starved than before. This was really horrible. I had no way to

beat

the escape room and I was going to lose 3-4. The AI ​​was going to sweep me three times, back to back. Humanity was in its last stages. But he had one more ace up his sleeve.

Snapshot 22w13. Better known as the April Fools update one block at a time starting in 2022. The update completely revamped the way blocks could be placed. And while sand and cacti work normally in this update, bows and arrows can be placed down. This means that potentially, with the right combination of stacking items on top of each other, I could build a ladder for my arena with bows and arrows. So I set up my boost rig to try to perfect my technique. But after many minutes, I couldn't even get that to work. So I tried something different. Since you could now grab blocks by simply breaking them, no matter how high up I was, I could grab any block below me and not have to worry about picking it up.

Maybe with that I could climb higher. But even this discovery did not help progress. Arrows and bows could not float on their own, they functioned like sand and cacti, where they had to be placed on the ground. And after experimenting more, I realized I just couldn't do it. There weren't enough blocks to reach 28 blocks. Was over. He had lost against the seeds, he had lost against the quartz, he had lost against the snow and now he had also lost against the cacti. The AI ​​won. In a last ditch effort, I thought maybe I could trick the game into stacking cacti on non-sand blocks if I threw them instead of placing them.

Nothing, it just landed on the block in a sort of strange state of limbo, with no additional lift. But when I tried to see if I could step on it, something strange happened. They hit me in a random direction. This limbus cactus had a strange property that made me feel driven. So I ran into that again and again, and then this happened. I was gaining height. I didn't know how it happened and I had no idea how far it could push me. All he knew was that he had accidentally discovered a quirk in the game's code.

A failure that was now my only hope. And so, for 20 minutes, for the sake of humanity, I encountered the cactus, over and over again. Different setups, different positions, different throws. I continued running towards the cactus. Trying to push the limits. No matter how much damage I took, no matter how much time I spent, no matter how many times I had to run into that blind cactus, praying it worked, I was going to win. , because I have something that AI will never have: human stubbornness. Never give up, no matter how stupid the odds. Representing humanity, there was no way he could quit.

Death wouldn't stop me, nothing would stop me. And when I saw that stupidly high, gravity-defying throw, I knew. Had won. The score was 4-3. Humanity. But he wasn't out of the woods yet, he still had to beat the next three escape rooms if he wanted to win. And I planned it. In the next room I had to climb 25 blocks, with only 20, but as you can probably see, this was not a challenge. I could fire impulse with flint and steel and just use kelp, but since that's difficult in single player and, frankly, not fun, I had to, of course, go the other way around.

TNT boost, and this was probably the first time my solution was aligned with chatGPT. Using the TNT to blow me up, one at a time, off camera. 2 more to go. Room 9 was somewhat similar and nothing too fascinating either. I imagine GPT must have thought lava was as valuable to crops as water since I didn't get any water, but sadly that's not how farming works. Anyway, this couldn't stop me from actually using the one thing I needed. Lava and some dirt. Every time you disconnect from a world, you gain 3 seconds of invulnerability in which you take no damage from anything.

Then, by continually disconnecting from my world again, I was able to slowly but steadily crawl through the lava, until I reached the exit. The score was now 6-3. The battle between humans and robots would be decided right here, in this final escape room. If he lost, everything would be in vain. 30 blocks with only 20 clay and 3 stairs. This looked absolutely terrifying. My first idea was to abuse the stairs. Since you can climb a ladder at any point along its height, you can use a ladder to climb two blocks. But even with that, and all my clay, I would still be 4 blocks away.

Just out of reach. So, discouraged, I dug deeper. Fear was already sweeping. I didn't want to give up, but this really seemed impossible. My only crafting recipe was to turn my clay balls back into blocks; Without a kiln, there was very little I could literally do with clay alone. The chickens were definitely not going to be of any help, nor was the shovel. Even GPT generated this escape room ADMITTING that it might not even be possible. I was completely devastated. It seemed like everything was over. I fired up Chrome one last time to do some research.

To check past versions of everything I had to see if there was anything that could help me. Obviously Clay wouldn't have anything interesting, so I skipped it. After digging, the chickens offered no help and the stairs have been functionally exactly the same since they were added to the game. I was hoping for a second that maybe eating raw chicken would poison me and could harm the ticks in every clay. But it just made me hungry. I was going to give up. AI had truly defeated humanity. And then, in my really final moments before closing the game, on a whim, I decided to go with the flow and look into the history of clay.

The most boring block in the game. And I saw it. A mechanic that existed in the game for only 3 months in 2015 before it was removed. A mechanic so ridiculous you wouldn't believe it if I told you. What if I told you that if you built a vine face out of clay in 15w31a, you could summon the ender dragon out of thin air? For the first time I had hope. All I had to do was transfer my world back to 1.20 and wait. No matter how many deaths, how many times I have to restart this escape room over and over again.

He was going to be hit by this dragon and fly out of this room. Not only to win for this video, but also for humans and everything we represent. Was it bad luck that ChatGPT had chosen clay, bad luck that clay was the only block used to summon the dragon, bad luck that the dragon hit me at the perfect angle to get out? Or was it just luck that chatGPT originally gave me water for this escape room and I said no? Was it pure luck that I chose to regenerate the item, not because I wanted it to be easy, but because I wanted it to be difficult?

You tell me. But it is not luck that when we humans set out to do something, when we decide to aim for the stars, there is no obstacle, obstacle, barrier that we cannot fly over.