
Buddi Bot is an isometric puzzle game where you must (re)train Buddi Bot, an advanced AI with neural network technology. With just a click, Buddi will complete tasks like washing dishes or cooking a steak, by learning COLORS and SHAPES. Are you prepared for the future? Play and find out.
Revenue comes from in-game items and microtransactions, which our price-based model doesn't estimate. Copies below reflect the player base, not paid sales.
Media
About
Genres
Stats
Languages: English
Engagement
Reviews
Buddi Bot is a cute and engaging game. Figuring out how this little AI helper works is a bit confusing at times, but the interactions that occur because of this are hilarious. Buddi Bot's interactions can be broken up into destructive and nondestructive. If you cause a destructive action to occur, and the destroyed item is one required to complete an objective, you can choose to undo the interaction, but are unable to retrain Buddi Bot to do something else with that input. In this way, the game acts like a classic "river crossing puzzle", where you have to figure out an action that won't destroy the thing you need for your next goal, even if that action is the wrong one just so that you can use it to retrain Buddi Bot. This puzzle is fun! Now technically, Budi Bot is in my understanding very far from a neural network. Budi Bot undeniably learns, but that machine learning is akin to a goldfish. The silliness of the game brackets it outside of reality, and so although it might use neural networks as a backdrop, the player should expect the representation of machine learning to be just as silly, and it is. Buddi Bot does not need to be implemented as an actual neural network, this is a game first and foremost. Budi Bot's memory is very limited. It remembers the training you have given it, but when it is told to interact with something it looks through that list of previous training (most recent to least recent), and picks the first match it finds for the features (color and shape) of the object you told it to interact with. This is far from the statistical nature of neural networks, but it does demonstrate one downfall of them. I think you could say that any iteration of Buddi Bot is over-fit to its most recent past experience. The accuracy of a neural network is a function of the completeness of its training set. Without enough data, we might label it as biased. If a human recognition algorithm is only trained on white males, it will likely only consider white males to be human. Buddi Bot's very short term memory exemplifies this issue as well, and it touches on the ethical dilemma. In a robot form, a bad AI can burn down a park, injure animals, or cause issues with heavy machinery. Of course, this happens in the silly context of Buddi Bot, and it doesn't go as far as to explicitly draw connections with real world issues. And I think that is to some degree to its benefit. It's not heavy handed. I personally have not finished the game. The factory level requires Buddi Bot to be trained for its puzzle before reaching it, and going back to retrain without being able to see the content of the factory level at the same time is a pain. It's a nice idea for a final level, but poorly executed. Completionists may be happy to finish, but the monotony dragged me down. For a free game, Buddi Bot is wonderful. Although dragging on near the end, it is well worth an hour of your time!
Related
Related
Trends
Concurrent players on Steam, sampled daily since tracking started.
Hours played at review time, across 24 reviews with recorded playtime.
⠀
lit the park on fire, goose gang
Bought it in the POC devs sale because it looked like a cool idea. It's full of spelling errors, it doesnt even work half the time, and the dev put more thought into shoehorning black supremacist propaganda into the game than the actual mechanics
虽然我没搞过人工智能,虽然我知道做游戏肯定需要在一定程度上扭曲现实,但按照游戏里这个逻辑做出来的八成是人工智障,而且家庭清扫机器人和工厂组装机器人强行用一套参数还不能调整?组装过程具体有哪些环节需要用图里哪些装置都不介绍一下的?然后每次都要把整个命令列表从头走到尾(不能后退一步或几步,前面看了若干遍的动画也不能快进)。 Although I know nothing professional about AI and I admit that the conversion of reality into computer game is necessarily distorting, the way how the AI is implemented is far from satisfactory in my opinion. The AI recognizes only two parameters, the color and the shape, each being divided into only a handful of categories, and sometimes the AI makes some migration on its own (like use a green bottle instead of a blue bottle if it cannot find one), which makes its behavior quite confusing and unpredictable. In the fourth level, you will be asked to used the AI you trained in the first three levels (in a household situation) to engage tasks in a toy car assembly room, without the chance to re-train or adjust its logic by any means, i. e. you are using exactly the same parameters for a house-cleaning robot and an assembly robot without adjustment. That does not make much sense. What makes it more confusing is that the mechanism and workflow of the assembly is never properly introduced to the player, like what does the furnace/chute do? One more frustrating factor is that you must rewind to the beginning of this level if your plan goes wrong, which means you may have the first three steps right and the later steps wrong, but you have to watch the AI do the first three steps over and over again when trying to figure out the remaining solution. And you cannot fast forward the animation. So even if the game does reflect some aspects of AI development, it is not good enough for my recommendation.
A free game that will not teach you a lot about machine learning but is cute and fun for an hour or so. You are retraining an AI to associate objects with the object you want them to interact with. The trick is you cannot always retrain the object you want to without sending off to do something unwanted first, so you have to indirectly train about an object that shares a trait, so the learning will carry over to a new object. Sound tricky? It kind of is, though if you get it right on the first time it makes sense. The main downside is you can't save your progress so if you want to come back to the later levels on a different sessions you have to replay the earlier ones. Ok if you can play a longer session at once but less convenient if you have to quit partway through. Anyway, it costs nothing, is pretty funny sometimes and has a light atmosphere. Worth a try if you like small games and don't mind a bit of trial and error.
I have a hunch that actual AI and machine learning doesn't work like this. Seems like someone got an idea to be witty and have an AI shatter a plate in your face without worrying about the logic behind it, in order to make some philosophical points. The gameplay is frustrating and it should be easier to reprogram the bot without having to rewind time. There's also some out of place political stuff thrown into the levels. Keep the blatant agendas out of video games. There's a clear line between including something and it not being a big deal, and including it in such a way that feels unnatural and shoved in someone's face. The former subtly promotes equality and unity as it normalizes things that should've already been normal; the latter only serves to alienate a large portion of your audience and push them in the other direction. But considering that this was a college project, all of this makes sense.
Buddi Bot, sinir ağı teknolojisine sahip gelişmiş bir yapay zeka olan Buddi Bot'u (yeniden) eğitmeniz gereken izometrik bir bulmaca oyunudur. Buddi, RENKLERİ ve ŞEKİLLERİ öğrenerek bulaşık yıkamak veya biftek pişirmek gibi görevleri tek bir tıklamayla tamamlayacak. Geleceğe hazır mısınız? Oynayın ve öğrenin.
good game
Игра хорошая, она помогает лучше понять механизмы машинного обучения и нейросетей.
Momentum
Pace estimated from the game's recent review growth (our daily snapshots) at its own sales-per-review ratio — a momentum signal, not booked sales.
Audience
Estimated from the language of this game's reviews — a proxy for where players are, not official Steam demographics.