Humanization is our main goal
Hello guys! Do you know that during the first 2 months of summer, we released 9 updates. Relocated to a new place. Brought a dog to our…
Hello guys! Do you know that during the first 2 months of summer, we released 9 updates. Relocated to a new place. Brought a dog to our office. Released an epic update with Sandbox mode. We had almost completed working on even more epic update with Reinforcement Learning and Self Driving cars and were able to introduce it to a private friend players group in Discord. We were displayed at the front page in Steam. Made a few special kitties for our friends. Met thousands of new awesome players. And a lot more. But let’s speak about future, shall we?
We’ve already told you that while True: learn() is a very unusual game. It drew in an amazingly positive and thoughtful community (we ❤ you), who we are always eager to speak to. Do you know what it means to us? It means that there are no doubts left about what we are going to do next. Of course, we will not have enough time to create everything we want to even if we stop sleeping, but it is a poor excuse to stop trying.
Therefore, our main strategic priority for today is humanization of our game. Let me explain what I mean by using an example. How does it work now: we receive a letter with a task. In this task, a school asks us to divide students into classes based on their data. This is a classification task. The game shows you a funny issue of the Deep News magazine about how machine learning was arising in the 1960s. We open the editor and build a scheme. Launch it. Red and green cubes move. If we have built the scheme correctly, the cubes arrive at the correct exits. Our task is completed, we send the system to the school, which ordered it, the customer pays us the promised amount of money and we spend it on clothes for our beloved kitty, knowing that we have not spent the day in vain.
We have some questions to this order of things. We do not like that the player has not learned anything about the insides of the node he used to solve the classification task. What is even more, he may have not even understood that he built the classification yet! He hasn’t learned what happened to the school after he has created the system. And what could have happened to the school if he has sent a bad system over there? Also, these green and red cubes that are moving… it is a hard task for one’s imagination to realize that those are the pupil’s data you need to sort out by class. And, if you have already forgotten, we will remind you that the main hero’s goal is to create a machine learning system which will be able to decipher his kitty’s speech. And how does the player eventually understand he is getting closer to reaching his goal? It remains uncertain. And this is what bothers us. Too many abstractions and lack of explanation.
You may rightfully ask us why we have been making something we do not like in the end. To be honest, we like what we have created, but there are some things that can definitely be improved, and we will take care of them. Thanks to the abstractive task structure, we are able to move forward fast. We surely do realize that there is a part of players who do not treat Early Access projects well or may get disappointed in the high level of abstraction of the game. But there are many people, as it turned out, who find joy in solving puzzles, building systems and playing games about machine learning in general. If there had been many players who did not like the approach we took, we would have definitely changed our tactic.
You can also say that a well-designed system does not need explanations. I cannot but agree, but I won’t make any comments about this. Next time I’ll just take a picture of a list of options that we throw out every time we discuss new game updates. If you have played through some part of the game, make sure to believe that we have tried to come out with a bunch of other options and make them work, but the final result turned out to be the best one to leave.
Our goal is to make the game reach the stage when all the tasks one has completed will leave an understanding in one’s head of what and how exactly he has done, and what are the consequences of the implementation of the system built by him and how he can use the obtained knowledge in other areas. We admire the Human Resource Machine game. It’s a wonderful idea to show the Turing machine so clearly, through the work of little humans. Moreover, what a great idea they came up with: to build the development of the game around the career of a character who is corporation worker. In general, we are very inspired by this game and we hope that we will approach its level of visibility and clarity over time in our game.
Speaking of tactics, right after we end up working on Self Driving Cars, we will begin making the game more humanized. Some of us have already begun to do so. If you want to see what exactly we are up to — just take a look at our public roadmap. We plan to get back to non-supervised learning and convolutional networks a bit later, maybe even after we will release the game on iOS and Android. In August, many of us will finally go on long-desired vacations for a week or two, so that after having a good rest we will all be able to work with more power and enthusiasm. Do not forget that we are writing about what is going to be changed/added in the Friday updates on Steam every Monday. And on Fridays, you are always able to play the new versions of the game. Do not forget to leave your comments under our posts: each new comment increases the amount of good stuff in our game by 1%.
Thank you for staying with us. Together, we will reach this huge goal: to make the machine learning process as interesting and entertaining as possible. Who if not the game developers should make learning fascinating? An old Indian man sitting on the hill told that this is the reason why they are needed in this world.
Useful links
Our Friends Club (free Pre Alpha and unique kitty promo code are included)