I feel like imma be 50 years old and randomly hear about this project again. It will be completed. And I’ll end up checking it out and showing my kids lol
which makes me wonder. In 50 years, won't new buildings be built and won't old ones have been torn down? So what are the builders going to do when large portions of their map are out of date, rebuild it or leave it? Will the map just be a hodgepodge of what the world looked like over the past 50 years?
@@kennarajora6532 Good thought, I wonder if there is a version of google maps that doesn’t update, so they are still using a map from 2021 or whenever this started. If not, kinda cool tho, it would end up a collage of time almost
@@tryingblox Google Earth Pro allows you to view old versions of maps and the online google maps allows you to view old street view data. And despite the "Pro" naming, it's free
@@_pixelicious knowing how much AI is evolving, the amount of info it would have to learn (5 years of human progress) and how this is literally what AIs are for right now (recreating based on existing information) yes, an AI could do it
As a programmer, this makes me really excited. I'd love to see the tech behind it and it gives a lot of room for improvements. If the program also logs your actions and isolates the model that you specified on google maps, you could create a dataset that includes the specific building from a 3d perspective, a score on how good the pregenerated building is compared to what you build manually and also let the program save the finished building after manual edits. This way, the program could interpret the manual adjustments and use it to further enhance pregenerated building quality on future models.
I also kinda feel like it should be able to detect corners of a building to begin with since it can apparently detect mouse collisions. So skipping the first step and just going through the list of building characteristics would improve the build time about 30%. And to be honest this is looking like something for a tablet, that minecraft edit menu looks a lot like a canvas.
@@puppergump4117 Yeah, I feel like you can skip another few steps that were depicted manually here, but this leads to the program automatically transcribing google street view data to minecraft with people just overlooking the results. This most likely needs machine learning reinforcements to learn from manual builds
@@moneyboyreups On the bright side they have many buildings already. On the not so bright side they're gonna have to somehow isolate each and every building they use for training the model. I think it's better to ignore learning from the builds themselves and instead stick to the interface part of it, the selection of the outline and design. I think I remember some tools that can pick out individual buildings from Google Earth, but the design is definitely a Huggingface problem.
@@puppergump4117 Yeah could be. But if you have the user manually select the points of the outline it could be just 2d images of the different sides of the buildings that can be evaluated to the expected output in minecraft. But I agree it gets complicated real fast
@@dread_nought it would speed up other types of city builds with a few tweaks, instead of having to make entirely new skyscrapers you could instead take a prexisting real one and alter the colors, materials and details. building any skyscraper of that size in such speed is ridiculous even outside of 1:1 earth
As a data scientist you should just make an AI to automatically generate the earth for you in Minecraft. You already manually created tons of buildings in Minecraft to train the AI on
@@FHBStudiomaybe not exactly. I think using Google maps and street viewer are enough to make a good enough 3d model for AI to use. Minecraft is a block game, it can't be that detailed to need a detailed 3d model
What blows my mind is that people can construct buildings in Minecraft thousands of times faster than they can construct them in real life, and yet humanity has still made so many buildings in a couple thousand years that the task to replicate them all in Minecraft seems insurmountable. I get it that IRL builders have made buildings over a much longer period of time and are much higher in number, but it's still mind-boggling to imagine the time it would take to create every single building in a city, let alone the planet.
One thing that's probably going ti be a problem though: not every house has a 3D model in google earth, especially in villages. Also it probably will be difficult to get castles to work, even harder to get castle ruins to work, and there are a lot of them all over europe
Good luck building Germany where only the biggest of the biggest cities have 3d models of their buildings and where there is generally very limited street view coverage
hell id join you. it seems like they have the right idea going, they just need to automate the inputs a little bit. I think implementing some sort of algorithm for that might be the best option. They have most of the basics that are needed to have something auto generate it. The next thing few things that need to get done is block pallet type. i am thinking a outline of most used house color combos/color select from google maps and it finds the closest color (this can also be automated) then it builds the house with that block pallet instead of having to change it. They could then add something like building story height for suburbs like 1, 2 or 3 stories and it can auto generate the building height in places that there isn't 3D mode
You could even try to use image recognition to get the correct colors for the color palette. I'd love to help out as well, it seems like a really fun coding project!
Moving from 1 hour per building to one minute per building also opens up the opportunity to create a good set of training data for a machine learning algorithm. By next year you could have an AI good enough that you could click on a building, AI would generate it and show you a preview, you could tweak and click done. From one minute to one second.
Honestly, if they’re able to finish the city centers of like, the 1000 largest/most famous cities, as well as most of the major landmarks, I think that would be more than enough to declare victory.
But with this software, and eventually AI, anyone would be able to log in and find their own house! Which would be crazy! Those small houses might not be as cool as major landmarks - but for that one person who has been to that house in real life, it will be a special moment! 😍
@@segefjord No, you're a hopeless idealist -- in no universe will a Minecraft world become remotely a fair or minimally accurate representation of the Earth. besides multibillion-dollar metaverse investments will take care of the project with FAR more efficient methods. This youtuber is wasting his time
@@dread_noughtCrazier things have happened, I wouldn't count it out. Someone will develop a software at some point to do it probably, might even be a decade from now, but I say it will probably happen at some point. Unless Minecraft falls out of relevance.
Why do we keep moving the goal posts , why cant the community just suck it up and understand this project may take a couple generations , id die alot happier knowing future generations can explore a meticuously crafted 1/1 of the earth than some soul less AI copy and paste
@@segefjord i think being able to walk across thousands of miles and see thousands of towns with realistic identifiable architecture new and old and different enviroment types is better than monuments i wanna fly chicago to detroit lol
hey, if you are interested, we could generate a 3d mesh of earth, then point out each individual color make an algorithm to select a suitable block, followed by the coordinates of their placement. 1) Create a mesh or a grid precise upto seconds or minutes, lets suppose 5 minutes (more precise = better results), then we go to each corner of the grid and the center and capture 180 images 3x each at an 3 degree angle for the same spot, this would roughly be around 21,834,408,000 pictures. 2) Convert these images into 3d models, then convert these 3d models into schematic 3) Use worldedit and paste them accordingly.
To be frank, Minecraft was never a children’s game. It is simply a game. Even ignoring the average age of players, it has the highest sales ever recorded. I guarantee many times over that that’s not just kids
You are essentially generating an enormous labeled data set for various map features. This has to have some sort of monetary value. I suggest that you save/store all of your labels (answers to yes/no questions and builder settings for each building). You never know what companies may be interested in this kind of data. And it could make or break future funding for this project.
@@mxdanger Interesting. Depending on the available info, that seems like a great option to use to auto generate at least an initial structure without toying with the settings / counting windows / etc. Then manual validation after that.
Probably not, companies need incredible scale for training data to be useful, google has been using captcha for training self driving cars and image generation, and 60 million captchas are completed every day
That's only true for individual datasets. When you combine multiple good datasets and draw comparisons, that data becomes much more useful. This dataset is valuable, and I'm positive that the completed Build the Earth has scientific uses. The study and modeling of pandemics, for one thing.
An AI Generative Agent can interact with the entire world programatically, using all the logic and interactions offered in Minecraft. I really think that simulating real life social phenomena for study is a major scientific application of this project. Plague Inc predicted Greenland's resistance to Covid. The Corrupted Blood Plague is studied by the CDC to this day. There's no way this isn't going to push the scientific envelope somewhere, somehow. Good work, lads
@@jennareynolds1403 Interesting points. For this project to be useful for scientific applications outside the world of Minecraft, this project needs to have an intermediate data format that contains all modeling details for each structure, which can then be converted (most likely a lossy transformation) to something that their minecraft engine can understand and automatically generate. Restricting their data set to information that is only relevant to Minecraft modeling will cause the loss of information and hinder its application to more general purposes like the scientific appliations you mention.
I really hope the program will be open source in the end. I'm not a builder, but a dev and I'd be interested to take a look at the program potentially improving it and I'm sure I'm not alone. If the program is developed openly, there are dozens of devs that would love to contribute
One important thing is to make sure you get the UI navigation as fast and as smooth as possible, those little buttons will cause a lot of inconsistency and human error, slowing down time building.
@@slavic_commonwealth i don't even have a mouse bruh 😭 i'd have to get an actual PC then get minecraft then sort out all of the tech, and join the server. Maybe in like 5 years😭😭😭
Data about roof types, window count, building material and so on is actually extremely valuable for a wide variety of Government use cases. Make sure you capture that intermediate information as well as it could quite literally be sold to help fund the team :) Many countries do also have free vector map data - i.e. much more detailed versions of OpenStreetMap - which when combined with the Google Maps 3D API (in cities where 3D coverage exists, it's not very broad at the moment) can collectively fully automate much of this too.
You should be outlining the base of the building, not the roof. The roof will change with perspective but the bass is guaranteed to align with the terrain Google maps and other map programs do have automatic outlines of houses and streets and stuff, as well as street view which shows the front of the house. The mapping could be used to generate a bunch of house positions at once, and you can enter the details manually. Or with a bit of machine learning, you could probably automatically assign specific details using streetview images of the house (though this would be difficult to implement)
That is a good idea to get the basic shape. Most places that have big buildings do have the 3d mode so you could get the size of those buildings but it could honestly come down to how many stories is the house. That takes care of hight for most common houses.
@@owencmykyou have to think on bigger time-frames. Google itself is auto-generating the 3d models satellite images. It's pretty likely, especially with the cost of satellites going down, we'll be able to get even higher quality models of most of earth in a decade or so. This also includes fields, forrests and non-manmade structures.
If we could find a way to have AI scan buildings, replicate them in the UI, and have users make tiny changes (like window placement or structural changes), that would increase the speed tremendously. I’m imagining something like those pixel art websites but it does it in 3D from google maps
Yeah that's what I was thinking. A program that does all the tedious work of clocking corners and selects option for the user, maybe even have it do it with 3 variations so the user can pick the best one. Then all you have to do is put any personal touches on the buildings that you want.
If you were to have almost full automation where instead of 1 person building a house every minute, you have 100 or even 1000 AI bots that scan an area of google maps then creates an outline and fills in most of the details and then places them in the game, then you only need 1 human to go in and fix any details.
As a programmer, there is even more data for maps available then you think, 90% of buildings have a height and outline from the mapping service, so instead of clicking, it already knows, Other data is also available for machine learning ai, which can detect tree type, windows, and roof type, wall colors, and all of that by just using the data with no inputs, that how games like microsoft flight simulator generate the world, this project could generate every building randomly, then allow people to touch them up
The program concept genuinely sounds sick as hell. If something like that is possible, it could have huge implications for creating builds fast. Great video as always!!
As a beginner builder still, this makes me happy. I've always struggled with building and I often have to take multiple attempts at building it for the building to be approved by my team leaders. Most of these errors are to TPLL. TPLL seems like one of the things that might be programmed automatically, which would be sweet.
That "Minecraft Overlay" program (even without the automation features) is something I've always wanted. Like you said it's way too slow to build large stuff in Minecraft, so a what's basically a Minecraft version of Microsoft Paint or Photoshop could go a long way. Surprised no one has come up with this already, or at least made one that's prominent enough to be an industry standard (to my knowledge professional builders still use Worldedit or other in-game Server Plugins to make their work).
I really love how creative this project has been. It’s weird how similar a Minecraft project has been to the development of many real life ambitious inventions/projects. Like, so many people have had doubt and said it was impossible, yet Pippen and his team continue to come up with new ideas and methods to make it work. It’s actually really inspirational to be honest.
okay but hear me out, a ai powered machine that uses image recognition in order to scan through a city and basically automate the process of outlining the buildings as well as the window placement and roof types, any building its not sure about it will make out of red wool that will have a sign placed on it saying what its not sure about so a human can go in and fix any issues so then all a person has to do is add extra details to fill out spaces theoretically increasing productivity from 1000 in 16 hours to closer to 1000 in 6-8 hours which is much more realistic playtimes
This is a lot more likely, plus seems more usefull. You just need a database to work from, and lucky for us we already have one, just get the blocks from minecraft in a way that the computer can read them like idk a list x,y,z block_type orientation . . . Attached with a few images of the building in real life that this is supposed to show, get enough of these to train an AI to spew out a list of blocks coming from a picture of a building and u got it done
this is pretty hard, we often take for granted how ai imaging works but for google maps, and its quality, an ai scanning for the hight window styyle roof style ect is really difficult, things like color and positioning are possible but beyond that, you would need high level developers to code it
In my opinion, the AI should be used to automate the construction process, or rather to carry out an automation in which the builders simply have to verify, comparing it with images, that the AI has carried out a correct reconstruction of x house or building and be able to correct it from the same program. The way in which the AI could be trained would be with the previous constructions of all the builders on the server to recognize how a construction should look based on that of real life, along with the presets for windows, roofs, doors, etc. and so you could simply with images from different perspectives build it as faithfully as possible to what a builder would do, and that with the verification/correction that they perform would literally reduce the process to at most 20 seconds on average
this is crazy! i cant wait to see the program be finished and fully optimized. i cant wait to see the earth be built. i kinda wanna visit my home town on there at some point. it'd be great to do so at some point :D good luck!!!!
There are *so many* buildings that are literally copies of each other in real-life cities and suburbs. For example: NYC brownstones, four-plexes in Minneapolis, apartment highrises across the globe, suburban pre-fab homes, trailer homes. It would probably be pretty simple to find the developers that were putting up the most common ones in any given area, and start with making those and pasting them where they should be. Then you have a *lot* of filled in, true-to-life space, and filling in the rest becomes easier, and also less space for a program to fill in. For buildings that are identical except the paint jobs, it'd be very easy to adapt what you have to randomly generate (within presets) the build palette each time you paste one.
This project seems like exploring stars irl : takes a long time and a lot of ressources, but find some things that benefit us all. What you are getting done benneffit all the builder's community across minecraft. And for this, thank you
Great that you mentioned AI in the end. I was thinking the whole time, that this might be invaluable for the project. E.g., training an AI on 100-200 manually built houses of a certain style (with their outline as a feature), might enable the model to instantly create similar houses for a given house outline. This might enable the automatic creation of houses for each area/city after just a short training period.
I was thinking this exact thing as well, AI detail recognition might enable automatic construction of fully detailed buildings which needless to say would speed everything up quite a bunch.
Hey man, I've been here from the very beginning, and let me tell you; keep pushing, especially when I know your viewer numbers have dropped; all you need to do to keep things on target on the social media side of things is to keep going. I will be with and support BTE, and this channel all the way, Godspeed.
As a software engineer, this is going to be extremely challenging without the aid of massive funding / time. 1) 5:58 - This requires integration of google maps software onto an application , this is not possible unless the google map's software is external. (The BTE program serves as an interface). 2) 6:06 - Google Maps's API does not have an output that returns the length of "ruler". 3) 6:24 - This definitely is not possible , the 3d structures in google maps aren't physical objects rather instead extensions of the ground . This means they do not have a surface. However a program can be created to "harden" the buildings. Additionally this process cannot be automated without the use of machine learning. 4) 6:45 - This is not possible unless a similar Minecraft engine is created. This should be created in such were it can understand the user's inputs exactly as well as output to minecraft. 5) 7:15 - This would be an example where windows are symmetrical , the program can be designed to handle this scenario properly , however if there are even slight fluctuations of window designs, it would be even more tediously complicated. 6) 7:25 - This is true however the roof shapes acts differently for each roof dimension , the roof dimensions may not always be a rectangle , sometimes they could be a triangle , oval , pentagon etc. It's not easy to automate this. 7) 8:31 - Similar to point no. 4 , this would be extremely tedious and would take years to properly implement. 8) 8:48 - Generate -> Boom , This may be possible through single player worlds however multiplayer worlds work differently, this requires higher level action as the user of this program would need direct local access to worlds. Wish you guys the best of luck!
This is an interesting idea! Of course, there will still be plenty of manual work on old and/or unique buildings, but for the general ones... It should work
I have two points of criticism on this: 1. To me, and emphasis on the “to me,” the project was about the process rather the result and although I understand the project was created to complete the earth, I found beauty in the fact that we may never see the end result. Its kind of like a testament to attempting to replicate all of current human achievement in another world Think “he who plants the tree may never have its shade cast upon him.” 2. I’m not sure how much of the earth has 3d mapping, but I know that my area barely even has clear imagery and I can only assume there’s much more of the earth with this issue. So this would really only help with cities which account for a small percentage of establishments, although yes they are more significant. Call me old-fashioned but I’d rather stick to my WorldEdit wand and keep building with in-game commands. Again, this is only my personal opinion on the matter and this critique is not to be taken to heart, simply taken into consideration.
Regardless of all this, I think a program like this could have many uses outside of BTE and would be an impressive feat within itself. Obviously it may have to stray from using minecraft to generate the building but if the source code could be translated to another program which could display irl structures, this would be an amazing architectural tool. Although there probably is already something of sorts.
I'm with you on the first part. I'm someone who's been following this since it was first conceived years ago, and I feel like a lot of the commenters here saying 'oh with ai you could just finish the whole thing automatically' see all of this as a product to finish fast rather than a collaborative project in human creativity like it was first imagined to be. I'd honestly rather have this project never get finished than have it end up with 99% of buildings being placed without a single bit of human input. (However some of the things in this video like roof templates and assistance with building shape could be useful)
I remember seeing someone calculate the amount of work this project would take and it would take actual thousands of years, by calculating the current population and the rate of change of new structures being made in real time. Sure, training models and automation would help tremendously--but They don't even come close to reflecting what the things look like in reality and the funny part is that photogrammetry, satellite data, and more have already produced way more polished results in Microsoft Flight Simulator. This is just a Patreon cash sink like Star Citizen, Chronicles of Elyria, and Dreamworld.
im a developer and this really excites me, doing the calculations, if one person used this method it would take 125 years to finish the earth however with the dozens of developers you have that time can be cut down to like 5 years with enough people, this program is possible to make and i would really like to explore it, i have experimented with procedural generation before and this is very doable
Don't know where you got the calculations from, but it would take 7610 years with one person doing this method with a hundred doing it it's still going to be 70 years
@@informythicalscience7837 100,000,000 = 100million buildings in the world, one minute build time means for one builder it would take 100million minutes, 100million / 60 = 1666666.667, 1666666.667 / 24 = 69444.44446, 69444.44446 / 365 = 190.2587519, i think my original calculation was slightly off cause i used the building number in this video instead of looking it up and google says there are about 100million buildings in the world, there are a bunch of estimates online ranging from 100million to 100billion but it looks like 100billion is a middleground where alot of people say thats the amount, amount of buildings in the us is 5.6million from one source so 100million sounds around right.
This is already insane, but wait till the program is done and we iterate on it with an auto detect tool. Just click auto detect and the options fill themselves with guesses done by an AI. It may not get everything right, but it will be so much faster to fix two things than to input all of them
I have a suggestion that I think can help you guys a lot. The blender 3D application is an open source 3D suite that offers ton of functionality for 0$ one of it's strengths are add-ons, there is an addon that lets you import Google maps 3d models to blender, then there is another addon that lets you turn that model into Minecraft blocks (not just cubes but fully textured to fit the colors closely) and you can set the scale, the powerful part is that you can import that model into Minecraft by using the same addon ans turning it into schematics! I think with some tinkering you can generate entire cities in hours! Hope you read this and let me know if my idea works out for you guys. Nevertheless keep up the amazing job!
As a builder in this project, I believe that this will be very useful. I am very exited for this new program and I do think that this could bring the project to be finished in under a decade.
The only thing is as cool as this sounds, I believe that the beauty of something like Minecraft earth comes from the time and commitment built into it. The seemingly impossible dream is what makes it so cool. With an application you are removing part of that, thus even though you are doing the same thing it just feels different
Very well, let's start development. Shouldn't be that difficult to get this working. Someone else mentioned Google Maps before, and I think Google might actually have some nice datasets you could use when building this tool. Their own 3D-generation tool used in Google Earth and Maps needs to draw their dimensions from some place, after all.
Another way this could be improved if the program also read the OSM tags for how many floors a building has and the roof type. That would decrease the time of needing to manually select the outline points of a building, and you would only need to specify the elevation of the building and possibly correct the height of the floors, if it's incorrect. This of course relays on the accuracy if the OSM outlines, and if the building has these tags. If the program used these tags, this would also incentives people to contribute to the OSM database, by correcting the outlines and adding these tags for buildings in their city. When I first started building I did this process for my selected area and the large surrounding area around. It didn't take me long, and plus since the outlines were now correct, I didn't have to correct them ingame. The editor for OSM itself offers cadastre layers for many countries, that you simply enable, and trace the lines to correct the building outlines, eliminating the need to navigate to your countries GIS service and finding the building coordinates there.
Obviously im no expert in computer algorithms, but as someone who studied it in college: i love your focus on improving efficiency by shifting optimization focus from improving building speed to memo-izing repetitive build patterns. Very clever
One thing to look into, is a shocking number of homes are built by just a few builders and are largely identical with some cosmetic differences, and not just on one suburb but many suburbs across many cities. With a little research you can create prefab and research all their projects and essentially just paste mass quantities of houses. Same concept with chain businesses, most use identical blueprints all across the country, so with a little research, one person could realistically build say every McDonald's in the world in just a few days/ weeks. Same concept with fast food, gas stations, Walmart etc...
Still looks like a huge work, pretty sure at some point you'll be able to just get all the data from google API and automatically build the world with a program. Pretty sure you can train an AI that will generate the building as it looks in real life. You already have a good dataset for training with all the work that's been done.
This is so freaking cool dude! I’m always impressed with the amount of detail that you put into your work. This is extremely impressive that you guys built a whole code for this. I can’t wait to see what you do next.
The fact he recorded multiple videos of him screaming and putting it all in one video to create a beautiful masterpiece just to get the viewer to subscribe it's definitely working
There's a Blender addon called BlockBlender that can convert models made in blender into Minecraft blocks that can be exported into Minecraft. I really hope that this addon can speed up the creation of buildings for Build the Earth.
That's if one person is working on it. Like he said, a few thousand people could bring it down to 1000 days. The few thousand figure is unrealistic, but it still goes to show that it is doable. I could see this being done in about 10 years if this breakthrough works. Especially given that future advancements in building methods are also likely.
@@spiceforspice3461 If 1000 people worked every single day on this, for 12 hours each day, it would still take 15 years. And to have 1000 people building would take a lot of coordination, and at present there aren’t nearly that many builders, and sure as heck none of them work on this for 12 hours a day. Most don’t even build daily. I think if they don’t go mostly ai automated, this would realistically take around 60 years including this most recent breakthrough.
I love how humans innovate not only in things that matter to society in general but in everything that interests them. Building the Earth has no productivity in real life but we're still developing this. Insane.
You can tell this guy is really passionate and struggling to explain computer science. Fun video, glad someone is showing this project to the public and keeping it going.
This would actually be sick. If this tool becomes available for builders on a wider scale, it would be incredibly helpful. I've only been able to finish two buildings on this project so far since it becomes hard to continue doing the tpll thing honestly. Much slower than my usual method of building which is sadly too inaccurate. But something of this sort of program would be excellent. ... I should probably get back into this project either way actually...
What about using some sort of image AI? I feel like with the surge in AI developments recently, its a quite logical step. You could use the Google Earth 3D view of buildings along with an image recognition AI to determine the shape and colours present, then feed that into your hypothetical block generator which would chose the appropriate blocks, colours, and shapes to match. You could even implement yet another AI in this step too, since certain blocks like stairs, slabs, etc. have complicated placement and effects on the overall look of a build. Then you'd just need humans to review the results. It could also train on completed builds, such that it would improve throughout its use. It would require even less human input and be far more versatile!
Incredible, another thing to consider is that with a BTE building program it would make it easier for beginner builders like me, and would allow more people to join BTE
I get the idea, i think it’s great but I also feel like it’s moving away from the point of this project, having a “architecture” program “copy” a map out of satellite images into a new program (Minecraft). For me, what makes this project interesting was actually the “man made” aspect of it. But I digress, good luck with this idea and I’ll keep supporting 😁
I disagree. Worldedit is used regularly to speed up the menial placing, in theory this program would just be doing the same but tailored for remaking buildings. It's still a person going to each individual building, drawing it out, selecting the options and colors, and potentially even making their own designs in creative to be loaded in the program as well. Less time spent making building skeletons and copy pasting windows, more speed and accuracy.
This is a great idea. And I would even like to add something to it. I work in AI, so I think I have quite some insight into the possibilities. We could train a machine learning model to simulate the human actions in the program. The goal is to teach the AI how humans look, click and work at these buildings. The AI tries to simulate the human actions in this program. After quite some training, we could let AI use the program on itself. If done right, people would only have to check, if the result looks alright (~10s per building) or even let it run wild without quality control (
this is the worst misrepresentation of AI I've seen all week 💀. you're more concerned about the specifics of timing for letting people "check the result" than the actual mechanics of simulating "human actions" and letting it "program on itself"? I actually do research in AI and can tell you that you evidently have never touched AI in your life
As a developer, here is my outline of how the bte app could work. This reminds me of a blender plugin where you can import real building data using google maps api and an application called Open Street Map. This plugin could be used to import large blocks of cities. You could then use some AI or just existing applications to turn colors of the building into blocks and get a frame done. From there you can have a list of procedural templates (or upload them as some sort of world edit schematic). One could then use these templates to further construct the building and export that as either a world edit schematic or something similar. That's sort of my general idea implementation-wise. It's a large project but well within the scope of reality and current technology.
There is another way to shave off quite some time East Europe and East Asia are fairly fond of mass produced houses, and there isn't too many of them. Same building can be seen anywahere from Poland to North Korea Building those house once an dthen copy pasting them will let you finish a small town in a couple of hours
I feel like imma be 50 years old and randomly hear about this project again. It will be completed. And I’ll end up checking it out and showing my kids lol
which makes me wonder. In 50 years, won't new buildings be built and won't old ones have been torn down? So what are the builders going to do when large portions of their map are out of date, rebuild it or leave it? Will the map just be a hodgepodge of what the world looked like over the past 50 years?
50 years is just too much
@@kennarajora6532 Good thought, I wonder if there is a version of google maps that doesn’t update, so they are still using a map from 2021 or whenever this started. If not, kinda cool tho, it would end up a collage of time almost
@@tryingbloxthey said it would stay how it is when they built it and how it is on google maps
@@tryingblox Google Earth Pro allows you to view old versions of maps and the online google maps allows you to view old street view data. And despite the "Pro" naming, it's free
the fact that once this is completed, nobody will be able to explore it fully, is absolutely insane
with new technology its very plausible we will be able to experience the earth
@@Dmytro_Podorvan Wow I hope to one day do this in real life once the technology comes available.
Why do you think nobody could explore it fully?
@@Mawyman2316 probably because it will be to big, but idk.
@@Mawyman2316 so basically the entire map would be bigger than a million blocks
I can't wait to visit my childhood neighborhood. This project is so personal to everyone.
i think an "eviction mod" might be needed for people who dont want to leave their old house.
bro you’re going to need a nasa computer to even download the world
@@FardStudios it’ll probably be a server!
I wish I could, but google street views isn't in Germany sadly
it is now, recently just got added, not as in depth as other coutries yet but its a step
@@Dusti_buns
Just wait until 5% of the thing is done in 5 years, and then a AI comes along and completes the entire map in 3 days
AI couldnt get that much detail unfortunately, not now and not anytime soon atleast
@@_pixelicious knowing how much AI is evolving, the amount of info it would have to learn (5 years of human progress) and how this is literally what AIs are for right now (recreating based on existing information) yes, an AI could do it
JAHSJSJSHSHS
As a programmer, this makes me really excited.
I'd love to see the tech behind it and it gives a lot of room for improvements.
If the program also logs your actions and isolates the model that you specified on google maps, you could create a dataset that includes the specific building from a 3d perspective, a score on how good the pregenerated building is compared to what you build manually and also let the program save the finished building after manual edits. This way, the program could interpret the manual adjustments and use it to further enhance pregenerated building quality on future models.
I also kinda feel like it should be able to detect corners of a building to begin with since it can apparently detect mouse collisions. So skipping the first step and just going through the list of building characteristics would improve the build time about 30%. And to be honest this is looking like something for a tablet, that minecraft edit menu looks a lot like a canvas.
@@puppergump4117 Yeah, I feel like you can skip another few steps that were depicted manually here, but this leads to the program automatically transcribing google street view data to minecraft with people just overlooking the results. This most likely needs machine learning reinforcements to learn from manual builds
@@moneyboyreups On the bright side they have many buildings already. On the not so bright side they're gonna have to somehow isolate each and every building they use for training the model. I think it's better to ignore learning from the builds themselves and instead stick to the interface part of it, the selection of the outline and design.
I think I remember some tools that can pick out individual buildings from Google Earth, but the design is definitely a Huggingface problem.
lol ok
@@puppergump4117 Yeah could be. But if you have the user manually select the points of the outline it could be just 2d images of the different sides of the buildings that can be evaluated to the expected output in minecraft. But I agree it gets complicated real fast
Bro goes from a philosopher to a Minecraft builder within 10 seconds
And then back again to a philosopher in the next 5
The truth is that they are one in the same.
@@darkestccino5405 Can't be one and not the other
bro's videos would really be 10-20 minutes shorter if he'd just get to the point haha
@@tabloida would we really want that though??
a program like this wouldn't only be good for the BTE project, but for several other kinds of projects. imagine if it were released to the public
Why do you think it wouldn't be good for BTE? ☺
@@segefjordthe way they said "only" here changes the meaning to "this is good for bte, and also other things"
@@blueninja012 exactly. i think this is a type of innovation that should be available to everyone if/when created, and i'm excited to see this happen
it has no practical application whatsoever
@@dread_nought it would speed up other types of city builds with a few tweaks, instead of having to make entirely new skyscrapers you could instead take a prexisting real one and alter the colors, materials and details. building any skyscraper of that size in such speed is ridiculous even outside of 1:1 earth
As a data scientist you should just make an AI to automatically generate the earth for you in Minecraft. You already manually created tons of buildings in Minecraft to train the AI on
I'm thinking using drone footage around the buildings already built and geo data, and then the minecraft buildings they resulted in as training data.
@@FHBStudiomaybe not exactly. I think using Google maps and street viewer are enough to make a good enough 3d model for AI to use. Minecraft is a block game, it can't be that detailed to need a detailed 3d model
that's why no one trust data scientists anymore lmao what a bunch of bs
No i don't want an ai generated world based on earth i want the actual thing in minecraft
Data scientists don't make A.I models though
What blows my mind is that people can construct buildings in Minecraft thousands of times faster than they can construct them in real life, and yet humanity has still made so many buildings in a couple thousand years that the task to replicate them all in Minecraft seems insurmountable. I get it that IRL builders have made buildings over a much longer period of time and are much higher in number, but it's still mind-boggling to imagine the time it would take to create every single building in a city, let alone the planet.
not really... it can only take a few years to build a city
@@incription Exactly. Such a feat is mind-boggling to think about.
a couple hundreds years bro, not thousands. in 500 years, billions of buildings have been made. Purely insane.
@@airplanemode101 exactly, almost all buildings now are less than 200 years old
turns out there are a lot more people working on buildings irl than in minecraft
Bro just went" screw the humans, GET ME THAT CODE!"
Fr
@@yosefnodeieee Frfr
@@manuelr.7461 frfrfr
@@Sunshinebaby- Ong😐
As a not-robot, I can confirm that robots are better than humans at certain tasks.
5:33 My favorite subscribe plug by far from all the CZcams videos I've watched.
One thing that's probably going ti be a problem though: not every house has a 3D model in google earth, especially in villages. Also it probably will be difficult to get castles to work, even harder to get castle ruins to work, and there are a lot of them all over europe
Obviously theyd only automage standard buildings to begi. With
Good luck building Germany where only the biggest of the biggest cities have 3d models of their buildings and where there is generally very limited street view coverage
As a software engineer, making a program like that is definitely possible. Hell, I'd even volunteer.
If you had the time?
@@Mad-Goose I do have the time. Just probably not the connections to anyone higher up at the big bte teams
hell id join you. it seems like they have the right idea going, they just need to automate the inputs a little bit. I think implementing some sort of algorithm for that might be the best option. They have most of the basics that are needed to have something auto generate it. The next thing few things that need to get done is block pallet type. i am thinking a outline of most used house color combos/color select from google maps and it finds the closest color (this can also be automated) then it builds the house with that block pallet instead of having to change it. They could then add something like building story height for suburbs like 1, 2 or 3 stories and it can auto generate the building height in places that there isn't 3D mode
@@Frozdheck, you could just contact pippen and he could probably get you into the team
You could even try to use image recognition to get the correct colors for the color palette.
I'd love to help out as well, it seems like a really fun coding project!
Moving from 1 hour per building to one minute per building also opens up the opportunity to create a good set of training data for a machine learning algorithm. By next year you could have an AI good enough that you could click on a building, AI would generate it and show you a preview, you could tweak and click done. From one minute to one second.
That’s exactly what I thought, even neighborhoods with correct spacing and stuff.
@@shonewarrior2178 you could have the program run 24 hours a day 7 days a week
Tis is the kind of thing ai would be good for, I'll make my own art thank you very much.
Underrated comment; He needs to see this.
@@PatientMeteor74 I bet this is planned already. It was actually what I expected the video to be about.
Honestly, if they’re able to finish the city centers of like, the 1000 largest/most famous cities, as well as most of the major landmarks, I think that would be more than enough to declare victory.
But with this software, and eventually AI, anyone would be able to log in and find their own house! Which would be crazy! Those small houses might not be as cool as major landmarks - but for that one person who has been to that house in real life, it will be a special moment! 😍
@@segefjord No, you're a hopeless idealist -- in no universe will a Minecraft world become remotely a fair or minimally accurate representation of the Earth. besides multibillion-dollar metaverse investments will take care of the project with FAR more efficient methods. This youtuber is wasting his time
@@dread_noughtCrazier things have happened, I wouldn't count it out. Someone will develop a software at some point to do it probably, might even be a decade from now, but I say it will probably happen at some point. Unless Minecraft falls out of relevance.
Why do we keep moving the goal posts , why cant the community just suck it up and understand this project may take a couple generations , id die alot happier knowing future generations can explore a meticuously crafted 1/1 of the earth than some soul less AI copy and paste
@@segefjord i think being able to walk across thousands of miles and see thousands of towns with realistic identifiable architecture new and old and different enviroment types is better than monuments
i wanna fly chicago to detroit lol
hey, if you are interested, we could generate a 3d mesh of earth, then point out each individual color make an algorithm to select a suitable block, followed by the coordinates of their placement.
1) Create a mesh or a grid precise upto seconds or minutes, lets suppose 5 minutes (more precise = better results), then we go to each corner of the grid and the center and capture 180 images 3x each at an 3 degree angle for the same spot, this would roughly be around 21,834,408,000 pictures.
2) Convert these images into 3d models, then convert these 3d models into schematic
3) Use worldedit and paste them accordingly.
I love how pippen went from hand mining 1,000,000 obsidian to being a philosopher building a planet in the same children’s game
The average minecraft player is 27…
@@realah3001not for bedrock edition
He wenyfrom mining 1,000,0000 obsidian to doing whatever the hell this is 5:34
To be frank, Minecraft was never a children’s game. It is simply a game.
Even ignoring the average age of players, it has the highest sales ever recorded. I guarantee many times over that that’s not just kids
Philosopher?
You are essentially generating an enormous labeled data set for various map features. This has to have some sort of monetary value. I suggest that you save/store all of your labels (answers to yes/no questions and builder settings for each building). You never know what companies may be interested in this kind of data. And it could make or break future funding for this project.
@@mxdanger Interesting. Depending on the available info, that seems like a great option to use to auto generate at least an initial structure without toying with the settings / counting windows / etc. Then manual validation after that.
Probably not, companies need incredible scale for training data to be useful, google has been using captcha for training self driving cars and image generation, and 60 million captchas are completed every day
That's only true for individual datasets. When you combine multiple good datasets and draw comparisons, that data becomes much more useful. This dataset is valuable, and I'm positive that the completed Build the Earth has scientific uses. The study and modeling of pandemics, for one thing.
An AI Generative Agent can interact with the entire world programatically, using all the logic and interactions offered in Minecraft. I really think that simulating real life social phenomena for study is a major scientific application of this project. Plague Inc predicted Greenland's resistance to Covid. The Corrupted Blood Plague is studied by the CDC to this day. There's no way this isn't going to push the scientific envelope somewhere, somehow. Good work, lads
@@jennareynolds1403 Interesting points. For this project to be useful for scientific applications outside the world of Minecraft, this project needs to have an intermediate data format that contains all modeling details for each structure, which can then be converted (most likely a lossy transformation) to something that their minecraft engine can understand and automatically generate. Restricting their data set to information that is only relevant to Minecraft modeling will cause the loss of information and hinder its application to more general purposes like the scientific appliations you mention.
I really hope the program will be open source in the end.
I'm not a builder, but a dev and I'd be interested to take a look at the program potentially improving it and I'm sure I'm not alone. If the program is developed openly, there are dozens of devs that would love to contribute
you're an incompetent nobody
Would be a very simple program to create
Bro just pulled off one of the most beautiful harmonies in existance for a minecraft video
One important thing is to make sure you get the UI navigation as fast and as smooth as possible, those little buttons will cause a lot of inconsistency and human error, slowing down time building.
Needs vim bindings
Yep the UI elements need to be nice and big so they are easy to hit and it absolutely needs key binds easily executable with one hand
Yeah, that and full keyboard control would be most the best.
Yeah they better get a UI designer in on it if they want any substantial gain.
let's not make it an Electron app please
This guy is such a good motivational speaker i love it
yeah that and/or an over-optimistic demanding boss lol
@@richardpike8748 he reminds me of kickstarter projects
but you wouldn't really go and build something in his project after watching this video, would you?
@@slavic_commonwealth i don't even have a mouse bruh 😭 i'd have to get an actual PC then get minecraft then sort out all of the tech, and join the server. Maybe in like 5 years😭😭😭
@@mr.sandman3619 I'm sorry for you bro. get a PC, there's no need to even play Minecraft, PC opens so much opportunities to learn...
I love this. You’ve come a long way and I’m so happy to have supported this and been on staff back at the beginning.
Data about roof types, window count, building material and so on is actually extremely valuable for a wide variety of Government use cases. Make sure you capture that intermediate information as well as it could quite literally be sold to help fund the team :) Many countries do also have free vector map data - i.e. much more detailed versions of OpenStreetMap - which when combined with the Google Maps 3D API (in cities where 3D coverage exists, it's not very broad at the moment) can collectively fully automate much of this too.
You should be outlining the base of the building, not the roof. The roof will change with perspective but the bass is guaranteed to align with the terrain
Google maps and other map programs do have automatic outlines of houses and streets and stuff, as well as street view which shows the front of the house. The mapping could be used to generate a bunch of house positions at once, and you can enter the details manually. Or with a bit of machine learning, you could probably automatically assign specific details using streetview images of the house (though this would be difficult to implement)
He shows himself highlighting the base after the roof because it's 3D aware
@@branthebrave I know, very few cities actually have 3D modelling on Google earth. So for most places that wouldn't be possible
its just a demonstration, not what will actually be probably done
That is a good idea to get the basic shape. Most places that have big buildings do have the 3d mode so you could get the size of those buildings but it could honestly come down to how many stories is the house. That takes care of hight for most common houses.
@@owencmykyou have to think on bigger time-frames. Google itself is auto-generating the 3d models satellite images. It's pretty likely, especially with the cost of satellites going down, we'll be able to get even higher quality models of most of earth in a decade or so. This also includes fields, forrests and non-manmade structures.
If we could find a way to have AI scan buildings, replicate them in the UI, and have users make tiny changes (like window placement or structural changes), that would increase the speed tremendously. I’m imagining something like those pixel art websites but it does it in 3D from google maps
Yeah that's what I was thinking. A program that does all the tedious work of clocking corners and selects option for the user, maybe even have it do it with 3 variations so the user can pick the best one. Then all you have to do is put any personal touches on the buildings that you want.
If you were to have almost full automation where instead of 1 person building a house every minute, you have 100 or even 1000 AI bots that scan an area of google maps then creates an outline and fills in most of the details and then places them in the game, then you only need 1 human to go in and fix any details.
@@therealgoodiegang thats insane
Ok whats the point in replicating the world in minecraft with copy and paste features . I will never explore an ai generated mega build
@@clalam5241 Its Ai assisted, AI is just doing the repetitive menial tasks. It will still allow the builders to make it their own.
As a programmer, there is even more data for maps available then you think, 90% of buildings have a height and outline from the mapping service, so instead of clicking, it already knows, Other data is also available for machine learning ai, which can detect tree type, windows, and roof type, wall colors, and all of that by just using the data with no inputs, that how games like microsoft flight simulator generate the world, this project could generate every building randomly, then allow people to touch them up
you're no programmer LMFAO enough of the run-on, comma-spliced, hopelessly-idealistic, fluffy, unsubstantial, and substanceless word salad
The program concept genuinely sounds sick as hell. If something like that is possible, it could have huge implications for creating builds fast. Great video as always!!
As a beginner builder still, this makes me happy. I've always struggled with building and I often have to take multiple attempts at building it for the building to be approved by my team leaders. Most of these errors are to TPLL. TPLL seems like one of the things that might be programmed automatically, which would be sweet.
i can help
I can help
I wont help.
That "Minecraft Overlay" program (even without the automation features) is something I've always wanted. Like you said it's way too slow to build large stuff in Minecraft, so a what's basically a Minecraft version of Microsoft Paint or Photoshop could go a long way. Surprised no one has come up with this already, or at least made one that's prominent enough to be an industry standard (to my knowledge professional builders still use Worldedit or other in-game Server Plugins to make their work).
There is one sorta, it's called advanced creations.
for landscapes, theres worldpainter. for structures theres mcedit (or a fork of it, mcedit is ancient at this point)
This could be a phenomenal tool for map making as well. Super cool
I really love how creative this project has been. It’s weird how similar a Minecraft project has been to the development of many real life ambitious inventions/projects. Like, so many people have had doubt and said it was impossible, yet Pippen and his team continue to come up with new ideas and methods to make it work. It’s actually really inspirational to be honest.
okay but hear me out, a ai powered machine that uses image recognition in order to scan through a city and basically automate the process of outlining the buildings as well as the window placement and roof types, any building its not sure about it will make out of red wool that will have a sign placed on it saying what its not sure about so a human can go in and fix any issues so then all a person has to do is add extra details to fill out spaces theoretically increasing productivity from 1000 in 16 hours to closer to 1000 in 6-8 hours which is much more realistic playtimes
This is a lot more likely, plus seems more usefull.
You just need a database to work from, and lucky for us we already have one, just get the blocks from minecraft in a way that the computer can read them like idk a list
x,y,z block_type orientation
.
.
.
Attached with a few images of the building in real life that this is supposed to show, get enough of these to train an AI to spew out a list of blocks coming from a picture of a building and u got it done
Nah, AI defeats the whole point. This is Build The Earth, not AI Generate The Earth.
this is pretty hard, we often take for granted how ai imaging works but for google maps, and its quality, an ai scanning for the hight window styyle roof style ect is really difficult, things like color and positioning are possible but beyond that, you would need high level developers to code it
@@betabee1221 the point is to have the earth in minecraft, this isnt a building project, as he said in the video, its a coding problem
In my opinion, the AI should be used to automate the construction process, or rather to carry out an automation in which the builders simply have to verify, comparing it with images, that the AI has carried out a correct reconstruction of x house or building and be able to correct it from the same program. The way in which the AI could be trained would be with the previous constructions of all the builders on the server to recognize how a construction should look based on that of real life, along with the presets for windows, roofs, doors, etc. and so you could simply with images from different perspectives build it as faithfully as possible to what a builder would do, and that with the verification/correction that they perform would literally reduce the process to at most 20 seconds on average
this breakthrough is so crazy i can almost feel the buildings coming inside of me
@@Agvazela_Vega true
wording could have been a bit better
bro is too excited
@@slavic_commonwealthof course he would be
@@SuperiorPelicanthat's the point
this is crazy! i cant wait to see the program be finished and fully optimized. i cant wait to see the earth be built. i kinda wanna visit my home town on there at some point. it'd be great to do so at some point :D good luck!!!!
There are *so many* buildings that are literally copies of each other in real-life cities and suburbs. For example: NYC brownstones, four-plexes in Minneapolis, apartment highrises across the globe, suburban pre-fab homes, trailer homes.
It would probably be pretty simple to find the developers that were putting up the most common ones in any given area, and start with making those and pasting them where they should be. Then you have a *lot* of filled in, true-to-life space, and filling in the rest becomes easier, and also less space for a program to fill in.
For buildings that are identical except the paint jobs, it'd be very easy to adapt what you have to randomly generate (within presets) the build palette each time you paste one.
This project seems like exploring stars irl : takes a long time and a lot of ressources, but find some things that benefit us all. What you are getting done benneffit all the builder's community across minecraft. And for this, thank you
Great that you mentioned AI in the end. I was thinking the whole time, that this might be invaluable for the project.
E.g., training an AI on 100-200 manually built houses of a certain style (with their outline as a feature), might enable the model to instantly create similar houses for a given house outline. This might enable the automatic creation of houses for each area/city after just a short training period.
was thinking the same for the whole video
@@asgdhgsfhrfgfd1170 true! All sorts of different ideas are possible here. Might be a very fun project actually
I was thinking this exact thing as well, AI detail recognition might enable automatic construction of fully detailed buildings which needless to say would speed everything up quite a bunch.
@@Janskill From a programming perspective it seems really fun and I'd love to help out but it's too time consuming
like one bot looks at hundereds of late 20th century american houses and spawns a realistic subdivision
I wish that when I'm older, something like 35 or 40, I can see in my recommendations this project being completed
"But that opens Pandora's Box, which of course, we will."
Ah, human nature.
Am I the only one just blown away from 5:33 ? Pure beauty out of no where
finally I find someone that actually appreciates this part lol
Yeah at first I thought it was a clip from another person but he's actually just an incredible singer. I really want it as a clip now
A stable diffusion like model to convert Map images to minecraft models would probably speed up this process a lot
that's one hell of a developer recruitment pitch. Wishing y'all the best of luck on this terrifying task.
Hey man, I've been here from the very beginning, and let me tell you; keep pushing, especially when I know your viewer numbers have dropped; all you need to do to keep things on target on the social media side of things is to keep going. I will be with and support BTE, and this channel all the way, Godspeed.
it doesn’t matter if you live here, or a dirt hut. Once you place a bed next to the crafting table furnace, it’s a home
As a software engineer, this is going to be extremely challenging without the aid of massive funding / time.
1) 5:58 - This requires integration of google maps software onto an application , this is not possible unless the google map's software is external. (The BTE program serves as an interface).
2) 6:06 - Google Maps's API does not have an output that returns the length of "ruler".
3) 6:24 - This definitely is not possible , the 3d structures in google maps aren't physical objects rather instead extensions of the ground . This means they do not have a surface. However a program can be created to "harden" the buildings. Additionally this process cannot be automated without the use of machine learning.
4) 6:45 - This is not possible unless a similar Minecraft engine is created. This should be created in such were it can understand the user's inputs exactly as well as output to minecraft.
5) 7:15 - This would be an example where windows are symmetrical , the program can be designed to handle this scenario properly , however if there are even slight fluctuations of window designs, it would be even more tediously complicated.
6) 7:25 - This is true however the roof shapes acts differently for each roof dimension , the roof dimensions may not always be a rectangle , sometimes they could be a triangle , oval , pentagon etc. It's not easy to automate this.
7) 8:31 - Similar to point no. 4 , this would be extremely tedious and would take years to properly implement.
8) 8:48 - Generate -> Boom , This may be possible through single player worlds however multiplayer worlds work differently, this requires higher level action as the user of this program would need direct local access to worlds.
Wish you guys the best of luck!
this provides some answers to a lot of my concerns. thanks
This is an interesting idea! Of course, there will still be plenty of manual work on old and/or unique buildings, but for the general ones... It should work
I have two points of criticism on this:
1. To me, and emphasis on the “to me,” the project was about the process rather the result and although I understand the project was created to complete the earth, I found beauty in the fact that we may never see the end result. Its kind of like a testament to attempting to replicate all of current human achievement in another world Think “he who plants the tree may never have its shade cast upon him.”
2. I’m not sure how much of the earth has 3d mapping, but I know that my area barely even has clear imagery and I can only assume there’s much more of the earth with this issue. So this would really only help with cities which account for a small percentage of establishments, although yes they are more significant.
Call me old-fashioned but I’d rather stick to my WorldEdit wand and keep building with in-game commands. Again, this is only my personal opinion on the matter and this critique is not to be taken to heart, simply taken into consideration.
Regardless of all this, I think a program like this could have many uses outside of BTE and would be an impressive feat within itself. Obviously it may have to stray from using minecraft to generate the building but if the source code could be translated to another program which could display irl structures, this would be an amazing architectural tool. Although there probably is already something of sorts.
I'm with you on the first part. I'm someone who's been following this since it was first conceived years ago, and I feel like a lot of the commenters here saying 'oh with ai you could just finish the whole thing automatically' see all of this as a product to finish fast rather than a collaborative project in human creativity like it was first imagined to be. I'd honestly rather have this project never get finished than have it end up with 99% of buildings being placed without a single bit of human input. (However some of the things in this video like roof templates and assistance with building shape could be useful)
@@r0w0bin oh yeah for sure. Roofs are horrible to do by hand
as a musician, that resolve in the subscribe message was so unexpected and amazing.
Cant wait for brazil to be done, i really want to explore this world
I remember seeing someone calculate the amount of work this project would take and it would take actual thousands of years, by calculating the current population and the rate of change of new structures being made in real time. Sure, training models and automation would help tremendously--but They don't even come close to reflecting what the things look like in reality and the funny part is that photogrammetry, satellite data, and more have already produced way more polished results in Microsoft Flight Simulator. This is just a Patreon cash sink like Star Citizen, Chronicles of Elyria, and Dreamworld.
im a developer and this really excites me, doing the calculations, if one person used this method it would take 125 years to finish the earth however with the dozens of developers you have that time can be cut down to like 5 years with enough people, this program is possible to make and i would really like to explore it, i have experimented with procedural generation before and this is very doable
Don't know where you got the calculations from, but it would take 7610 years with one person doing this method with a hundred doing it it's still going to be 70 years
@@informythicalscience7837 100,000,000 = 100million buildings in the world, one minute build time means for one builder it would take 100million minutes, 100million / 60 = 1666666.667, 1666666.667 / 24 = 69444.44446, 69444.44446 / 365 = 190.2587519, i think my original calculation was slightly off cause i used the building number in this video instead of looking it up and google says there are about 100million buildings in the world, there are a bunch of estimates online ranging from 100million to 100billion but it looks like 100billion is a middleground where alot of people say thats the amount, amount of buildings in the us is 5.6million from one source so 100million sounds around right.
open source? 👀
i would like to watch ... and maybe help ... the development
This is already insane, but wait till the program is done and we iterate on it with an auto detect tool. Just click auto detect and the options fill themselves with guesses done by an AI. It may not get everything right, but it will be so much faster to fix two things than to input all of them
I have a suggestion that I think can help you guys a lot.
The blender 3D application is an open source 3D suite that offers ton of functionality for 0$ one of it's strengths are add-ons, there is an addon that lets you import Google maps 3d models to blender, then there is another addon that lets you turn that model into Minecraft blocks (not just cubes but fully textured to fit the colors closely) and you can set the scale, the powerful part is that you can import that model into Minecraft by using the same addon ans turning it into schematics! I think with some tinkering you can generate entire cities in hours!
Hope you read this and let me know if my idea works out for you guys.
Nevertheless keep up the amazing job!
I can see this tool being used in many new ways and could be a whole new way of building things
Imagine what you could do with 10,000 builders building for 8 hours a day
the interiors
@@mikefoster6127They obviously can’t do that bro
@@AustinIsLitbusting down the door of every house on earth so they can replicate it in minecraft
@@AustinIsLit ? why?
@@mikefoster6127 Explain to me how they would make the interiors
As a builder in this project, I believe that this will be very useful. I am very exited for this new program and I do think that this could bring the project to be finished in under a decade.
can i help
blink twice if you are being help hostage
@@hi-db7in blink blink
The only thing is as cool as this sounds, I believe that the beauty of something like Minecraft earth comes from the time and commitment built into it. The seemingly impossible dream is what makes it so cool. With an application you are removing part of that, thus even though you are doing the same thing it just feels different
this is absolutely mind blowing. keep it up, this is awesome!
Very well, let's start development. Shouldn't be that difficult to get this working. Someone else mentioned Google Maps before, and I think Google might actually have some nice datasets you could use when building this tool. Their own 3D-generation tool used in Google Earth and Maps needs to draw their dimensions from some place, after all.
Another way this could be improved if the program also read the OSM tags for how many floors a building has and the roof type. That would decrease the time of needing to manually select the outline points of a building, and you would only need to specify the elevation of the building and possibly correct the height of the floors, if it's incorrect. This of course relays on the accuracy if the OSM outlines, and if the building has these tags.
If the program used these tags, this would also incentives people to contribute to the OSM database, by correcting the outlines and adding these tags for buildings in their city. When I first started building I did this process for my selected area and the large surrounding area around. It didn't take me long, and plus since the outlines were now correct, I didn't have to correct them ingame.
The editor for OSM itself offers cadastre layers for many countries, that you simply enable, and trace the lines to correct the building outlines, eliminating the need to navigate to your countries GIS service and finding the building coordinates there.
Obviously im no expert in computer algorithms, but as someone who studied it in college: i love your focus on improving efficiency by shifting optimization focus from improving building speed to memo-izing repetitive build patterns. Very clever
What an absolutely fantastic "subscribe" harmony. Jacob Collier would be proud
Someone will probably make an AI bot that builds the entire earth in minutes
I loved your videos when I was younger
I keep finding you in the most random of places, lol.
I don't know if minutes, but certainly large chunks of the Earth a day.
That would require a lot of processing power
@@sown3476 minutes wouldn't be too unbelievable tbh.
One thing to look into, is a shocking number of homes are built by just a few builders and are largely identical with some cosmetic differences, and not just on one suburb but many suburbs across many cities. With a little research you can create prefab and research all their projects and essentially just paste mass quantities of houses. Same concept with chain businesses, most use identical blueprints all across the country, so with a little research, one person could realistically build say every McDonald's in the world in just a few days/ weeks. Same concept with fast food, gas stations, Walmart etc...
Eh, too many different designs. It's almost easier adjusting on the fly than preparing so many templates.
Unless the building is slightly rotated...maybe we need a solution to that
Still looks like a huge work, pretty sure at some point you'll be able to just get all the data from google API and automatically build the world with a program. Pretty sure you can train an AI that will generate the building as it looks in real life. You already have a good dataset for training with all the work that's been done.
Honestly, it would be cool if people would try to furnish their empty homes (if possible). Feels insanely cool!
This is so freaking cool dude! I’m always impressed with the amount of detail that you put into your work. This is extremely impressive that you guys built a whole code for this. I can’t wait to see what you do next.
They didn't build a program yet, unfortunately. Everything in this video is theoretical.
That subscribe intermission resonated on a spiritual level tho
0:22
That feeling when you _instantly_ recognize the clip here as being from Kurzgesagt
The fact he recorded multiple videos of him screaming and putting it all in one video to create a beautiful masterpiece just to get the viewer to subscribe
it's definitely working
Man, someone should try building the earth irl, that would be cool
I am lost for words to describe how amazing this is.
There's a Blender addon called BlockBlender that can convert models made in blender into Minecraft blocks that can be exported into Minecraft. I really hope that this addon can speed up the creation of buildings for Build the Earth.
Just subscribed for tier one, keep up the good work! Loving this determination 💯🙏
If every house took 1 minute, then it would still take 7 and a half thousand years…
AI could optimize this process to seconds and run fully automatically. Then all it takes is a massive computer and it’s done.
That's if one person is working on it. Like he said, a few thousand people could bring it down to 1000 days. The few thousand figure is unrealistic, but it still goes to show that it is doable. I could see this being done in about 10 years if this breakthrough works. Especially given that future advancements in building methods are also likely.
@@spiceforspice3461 If 1000 people worked every single day on this, for 12 hours each day, it would still take 15 years. And to have 1000 people building would take a lot of coordination, and at present there aren’t nearly that many builders, and sure as heck none of them work on this for 12 hours a day. Most don’t even build daily. I think if they don’t go mostly ai automated, this would realistically take around 60 years including this most recent breakthrough.
@@enolopanr9820 Yeah that’s basically the only way this can get done. Then they could go over everything and fix imperfections and errors.
i didn't realise the start of a minecraft video would give me existential dread since the mc credits.
I love how humans innovate not only in things that matter to society in general but in everything that interests them. Building the Earth has no productivity in real life but we're still developing this. Insane.
Completing this would legitimately be a huge human achievement. I mean building the entire world to scale even if somewhat automated is insane
PippenFTS: the one guy who can make eating a bag of chips look epic.
Every heard of death note?
God I love you, this is insane, so excited to see the progress
You can tell this guy is really passionate and struggling to explain computer science. Fun video, glad someone is showing this project to the public and keeping it going.
Wow. This is beyond Minecraft.
I'm sorry but noone's talking about the harmonies at 5:33? that was beautiful!
Fr lmao
ive been looking for the source for a while and cant find it. did pippen sing it himself or get it from somewhere?
@PigTranceForm100 nah he sang it himself I'm sure. Just surprised nobody pointed it out cause it's beautiful
@@tundratig damn its impressive but im slightly dissapointed that there isnt a standalone video of it
Dude your vocals on subscribe are well composed awesome work.
For someone that has been to Rochester, that’s surprising!
amazing vid aside, the guy harmonizing subscribe at 5:33 had me in tears 😭
Can’t wait for my suburban stroad hellscape to be recreated
This is amazing I can’t wait to see what this looks like when it’s finished
probably like earth
This would actually be sick. If this tool becomes available for builders on a wider scale, it would be incredibly helpful. I've only been able to finish two buildings on this project so far since it becomes hard to continue doing the tpll thing honestly. Much slower than my usual method of building which is sadly too inaccurate. But something of this sort of program would be excellent.
... I should probably get back into this project either way actually...
Turned into a whole choir, beautiful
I watch too much Hermitcraft. I laughed so hard when I realized you were using the same tune that's Scar's timelapse music here. 12:54
What about using some sort of image AI? I feel like with the surge in AI developments recently, its a quite logical step. You could use the Google Earth 3D view of buildings along with an image recognition AI to determine the shape and colours present, then feed that into your hypothetical block generator which would chose the appropriate blocks, colours, and shapes to match. You could even implement yet another AI in this step too, since certain blocks like stairs, slabs, etc. have complicated placement and effects on the overall look of a build. Then you'd just need humans to review the results. It could also train on completed builds, such that it would improve throughout its use. It would require even less human input and be far more versatile!
yes this sounds doable
Israel is blurred. And also north korea may be censored
@@minecraftify95 some places are required to be made manually.
@@keshi5541 imo the more authoritarian/isolate the country the harder to made with AI to the point it must be done manually
@@scrubfire14 exactly
Good luck to you guys. This is the most important project ever undertaken in Minecraft.
Yes. Our minds are synchronizing in this moment. Incredible video.
Incredible, another thing to consider is that with a BTE building program it would make it easier for beginner builders like me, and would allow more people to join BTE
I get the idea, i think it’s great but I also feel like it’s moving away from the point of this project, having a “architecture” program “copy” a map out of satellite images into a new program (Minecraft). For me, what makes this project interesting was actually the “man made” aspect of it. But I digress, good luck with this idea and I’ll keep supporting 😁
I disagree. Worldedit is used regularly to speed up the menial placing, in theory this program would just be doing the same but tailored for remaking buildings. It's still a person going to each individual building, drawing it out, selecting the options and colors, and potentially even making their own designs in creative to be loaded in the program as well. Less time spent making building skeletons and copy pasting windows, more speed and accuracy.
it would be impossible to do it all like that
I disagree
I feel this could just be WorldEdit on steroids -- a program like this couldn't make anything beautiful without humans
@@ieatpaper Throw some random math equations in and you'll end up with noise and fractals.
This is a great idea. And I would even like to add something to it. I work in AI, so I think I have quite some insight into the possibilities. We could train a machine learning model to simulate the human actions in the program. The goal is to teach the AI how humans look, click and work at these buildings. The AI tries to simulate the human actions in this program. After quite some training, we could let AI use the program on itself. If done right, people would only have to check, if the result looks alright (~10s per building) or even let it run wild without quality control (
this is the worst misrepresentation of AI I've seen all week 💀. you're more concerned about the specifics of timing for letting people "check the result" than the actual mechanics of simulating "human actions" and letting it "program on itself"? I actually do research in AI and can tell you that you evidently have never touched AI in your life
As a developer, here is my outline of how the bte app could work.
This reminds me of a blender plugin where you can import real building data using google maps api and an application called Open Street Map. This plugin could be used to import large blocks of cities. You could then use some AI or just existing applications to turn colors of the building into blocks and get a frame done. From there you can have a list of procedural templates (or upload them as some sort of world edit schematic). One could then use these templates to further construct the building and export that as either a world edit schematic or something similar.
That's sort of my general idea implementation-wise. It's a large project but well within the scope of reality and current technology.
There is another way to shave off quite some time
East Europe and East Asia are fairly fond of mass produced houses, and there isn't too many of them. Same building can be seen anywahere from Poland to North Korea
Building those house once an dthen copy pasting them will let you finish a small town in a couple of hours
A lot of American suburbs are like this too.