Hi, Associate Professor in Software Engineering here.

I had hoped I could use this game to teach my children about programming, and while I will definitely recommend the game for its entertaining puzzles, I fail to see much connection to any of the traditional C-syntax programming languages, including Python, and even less in the way of teaching/showing the fundamental aspects of Artificial Intelligence, Machine Learning and Neural Networks, though it is advertised. As example, the “while True: learn()” title hints to the traditional C-syntax concepts of program iteration, boolean evaluation, and call to a function, but none of these are really present in the game and definitely not in a way, that you would be able to identify them unless you know them already.

When it comes to Machine Learning I do not see any relation to actual data processing, training and classification. A real-world predictive system would have no problems classifying a red circle from a green square when it has been trained with those exact shapes and colors. The real “fun” would have happened if it had to classify a purple star instead.

Furthermore, processing time and transmission speed are major elements in the game, but are not given the credit to their importance in real world applications maybe because they are not tied directly to Machine Learning. I am really lacking a way to track transmission speed in the designs I make in the game, especially as it seems to be very important for the “startups” puzzles.

In short, I will recommend the game for its puzzles, but the talk about demonstrating programming and Machine Learning principles I classify as “hot air marketing speech”.

I love programming games. Sadly, this is not a programming game. I struggle to even call this a puzzle game when so many of the levels are putting a single piece in the middle of the board and connecting its ins-and-outs to the board.

Buying better hardware means you'll run the same solution you had on an earlier level and upgrade from a silver star to a gold star - not because you made anything better, but just because "your computer is now faster".

Levels randomly restrict which nodes you can use so while they added the ability to reuse your solutions from previous levels, odds are that level used a node you're not allowed to use and you'll just have to do everything from scratch. Except for that one level where it forces you to use a previous level's solution.

Late game pieces are not fun to play with - because the game grades you on "processing time" you get faster pieces with small error chances. The game expects you to briefly hook these up to your system with an extra "gradient descent" node to "teach it", but once you do this, just copy that node to all the places you need it to avoid "retraining" them.

The worst offender is, sadly, the driving simulations levels. They claim to use a genetic algorithm to evolve the AI. The solution to all of them is to just click the evolve button once and everything's done. Each level adds more sliders which tune the way in which your AI evolves - but you just ignore them and click the evolve button. Later levels add side-facing LIDAR so your car's AI has more information about when you pass another vehicle! Except this only confuses the poor AI and removing the side LIDAR solves everything.

I played it to the end to try and see what it had to offer. I'm sad to say that past the first hour or two, you'll have seen everything of value in the game.

Pros:
- The graphics and sound are clean. The production quality feels high.
- The visual editor for solving puzzles is easy to use. It's also easy to use solutions from past puzzles as nodes in your current solution.
- The game gives useful links to those interested in learning more about machine learning.

Cons:
- After a short time, the puzzles become very uninteresting. Most of them fall under the "load balancing" category, which means that you spend most of your time figuring out how to allocate output symbols to output streams to match the volume dictated by the puzzle.
- The test run does not always terminate when the test is unsolvable, for example if all inputs have been sent to output streams, but the current accuracy and / or volume in the streams is not adequate. In these cases, if you want to see what went wrong, you will have to wait for the test to time out until you can view the stats for each output stream. This is annoying in combination with the prior point, when you are trying to debug which stream is getting too many nodes so that you can balance the output.
- Most of the machine learning concepts introduced in the game end up being irrelevant to the puzzles. In particular, the reinforcement learning puzzles and RNN puzzles don't let you use the ideas to come up with clever solutions. The RL puzzles give you a lot of knobs to play with, but the puzzle doesn't require anything other than sending some inputs to the output stream. The RNN puzzles give you the idea of RNN, but don't allow you to connect the nodes in any interesting or meaningful way: each node has memory, but you are forbidden from doing anything with it other than feeding it back into the same node that produced it.

Overall I don't recommend this game. The puzzles stop being interesting very early, and those that attempt to introduce more complicated machine learning topics avoid the hard task of introducing it in a way that the user can learn about the topic via the puzzles.

TL;DR It won't make a Data Scientist out of you, but it's still pretty fun :)

Let's start by talking about what this game won't do (despite promises by the developers on the contrary):
- It won't teach you new coding concepts.
- It won't "keep you brain busy". At least not more than any other game that requires minimal planning
- It won't improve your problem-solving skills
- It won't teach you machine learning, neural networks, big data and AI (at least not directly)

Now let's talk about what the game can do:
- It can give you some starting point from which to start reading and learning about machine learning and AI.
- It can teach you a bit about logical thinking (as can any similar game, for example any Zachtronics game).
- It's fun

I would say this game teaches you about machine learning, as much as a Civilization game teaches you about world history.
It may motivate you to learn on your own.
You may learn some by reading the text description and expecially following the "further read" links shown.
But the game itself (solving puzzles) won't do anything in that regard.

All in all, as fun as any Zachtronics game, and as educational.

I spent seven hours getting to the end of this game, and I'm of mixed feelings about it.

At its core, While True Learn is a puzzle game: given bins of input objects (red/green/blue circles/squares/triangles), move them to the output bins that want them (like, one bin might want red circles and blue squares). You do this by building a flowchart made of conditional blocks: a "Decision Tree: Color" can send one color to the top output and another color to the bottom output (the third color is sent randomly to either output), a "SIFT" sends one shape to the top and the other two shapes to the bottom, a "MARK-1 Perceptron" can divide the colors across its three outputs but requires training (and misidentifies some objects), and so forth. The challenge is to send the inputs to the correct outputs using a minimum of blocks and within a time limit, so you may have to spend some time optimizing your approach.

At the same time, though this game is all about machine learning. "Learn how machine learning really works!" says the game's blurb, and the puzzle descriptions sound right up this alley. Predict a presidential election! Sniff out industrial espionage! Optimize roadway traffic! Write a chess bot! Recognize UFOs! The blocks you use all are named after real-world machine learning algorithms and have similar strengths/drawbacks. As you progress through the game and are introduced to new kinds of blocks, the game provides you with links to Wikipedia articles and YouTube videos explaining them and their history.

Problem is ... none of that actually has any bearing on the puzzles, which are still only about moving red/green/blue circles/squares/triangles from input bins to output bins. Whether you're identifying viruses or writing a movie rating system, you wouldn't know it from the puzzles themselves, which all look the same. Some of the puzzles (even in later levels) are trivially simple; others are too complicated to be fun. The only puzzles that really stand out are the ones about teaching a car to drive by writing a solution for it and adding LIDAR to the front and sides, but these still confused me - the car would accelerate until it rear-ended someone and then it would tailgate that car until the test was up, at which point I'd usually get a passing grade because the car's average speed was above the requirement.

On top of this are a strange array of features. You can buy hardware upgrades (after which simply re-running your earlier solutions can make them finish faster and get you more points), decorations for your office (the menu screen, which you don't spend much time on and you can only display a few decorations at a time anyway), and different kinds of cats (I didn't understand that at all). You can invest in startups, which means writing a solution for them and then clicking "Next Day" over and over again until the company stops making money. You can unlock "DLLs," which appear to be a way to re-use your previous solutions (I never used these because it was easier to write each solution from scratch), and "CatHub repos," which are just a notepad where you can try different approaches. There appears to be a well-developed "secret code" system where you can enter codes and unlock cosmetic items in the game.

All in all, it's really uneven. This could have been a much more fun puzzle game if it had focused on the puzzles, or it could have been a much more educational game about machine learning if it had made the puzzles resemble real-world problems. (Think of Kerbal Space Program.) It's enjoyable, but I don't think it stands out. This game teaches machine learning as well as Civilization teaches politics - which is not at all; you only get to use the end results. There are better puzzle games out there, and there are better simulator games.

Started as a simple game to kill time, 3 hours later I'm diving into articles on how to improve ANN.

Great game mechanics, simple to get you started, hard to get it mastered, easy to get hooked on Machine Learning!

Also, great graphics. Smooth 60fps on any PC =p

Funny thing, that as a developer, I took a break from coding to play this game, which is basically coding.

The programming in this game is only present in its theme. You will not learn anything about machine-learning, programming, or algorithms. Watching the shapes fly past can be satisfying, but this game heavily boasts that it teaches machine-learning. If it didn't pretend it was more than a shape-sorting game, I probably wouldn't be leaving a negative review.

On first glance this is just another program algorithm" game. However, once you get into it, the solutions become increasingly complex and you will accidentally learn things. What I like most are the links to YouTube videos and websites to learn more about the math behind it all. I haven't finished the game yet, so I can only hope it will get increasingly more complex and the goal for the game (teaching people machine learning) will be fully achieved. The cheeky story behind it all is a cherry on top of it all.

Sad to give this a negative review, but after 6 hours I just can't make myself continue.

1. It's not a programming game. And there's no machine learning other than names and links to Wikipedia articles. In fact, it's not even a puzzle game. It has things like node upgrades which means solutions are not deterministic: running them twice can yield different results. You can also unlock nodes in the shop to use in levels, which means solutions are even less deterministic and players can't compare.
2. I don't know how they did the RNG, but it's not right for a puzzle game. You can run the model, see where the inputs go, and then make a minor tweak only to find the distribution is totally skewed. What makes this worse is that the input sizes are absolutely tiny. So while a node may have a 50/50 of output, you may actually end up with 70/30. So any plan you had that works at statistically significant scale is ruined because it's 1 off. I've had 2 or 3 levels where my solution worked, but I had to move the ordering of nodes around to get the RNG to work right.
3. Side features. Training the "AI" (neither the manual training or the node training), startups,... they're not good. The mechanics aren't interesting or, worse yet, are poorly explained. What's fun about having to put down a node, hook it up to the input, then run "Train" and wait 20 seconds TWICE to get it to go down to 20%? The developers even seem aware of this because the sandbox area explicitly turns this feature off!
4. How the game presents it's levels is just... boring. In games like Shenzen or Infinifactory there's topology and each level has a unique characteristic to it. This game takes puzzles about filtering video game categories or movie ratings or pizza toppings and turns every single one into a completely unskinned "sort by 3 shapes and/or 3 colors".
5. There are quality of life issues. You can't adjust the input/output nodes on the board and they're unaffected by zoom - given the node limit, this doesn't constrain your solutions at all it just means you have to pack any complex solution together in a way that makes it unreadable. No orthographic lines or cross-through visuals like real diagrams use, so sometimes you can get wires crossed. Some nodes require an aforementioned annoying training process. The hitboxes for selection aren't constrained to the visual node, so you'll often select more boxes than you mean to for copy & paste or dragging. Puzzles can only speed up 3x, so you can waste 40 seconds testing solutions.