Building the Dot Matrix

I've taught myself to enjoy learning. This was both a conscious and accidental decision that I made in high school. My first two years of high school, I was failing a handful of classes. I just wasn't interested in them. However, I was acing all of my engineering courses. Around the end of my sophomore year, I realized that if I wanted to live the life of an engineer, I needed to raise my grades. This is where my love of learning began.

Two things happened in parallel. First, I put the effort to reread and revise all of the boring subjects. There was nothing fun about this, it was just grueling work. Second, I began to watch science YouTubers. Science YouTubers had a common technique in those days; it was to ask a question and then explain why your intuition was wrong. Through this process, they would help me see the science problem with more clarity. This clarity was the genesis of drive to understand the details.

Fast forward just a little bit and these two facets combined. I started to absorb all kinds of information. In the process, I would filter them down into smaller and smaller pieces that I could keep in my mind. This way I could reconnect theses pieces when I needed to explain it to myself or others. I visualize these bits of information as dots.

These dots aren't just small trivia facts. They are short stories, steps, a passage from one thing to another. You can almost see them as the blueprint for a stream of consciousness. Whenever a dot connects with another dot, the dots grow in size. When dots are frequently used together, their connection thickens.

Let's take the word steam for example. My mind immediately breaks that into vaporized water and STEAM (science, technology, engineering, arts, and math). From vaporized water, I think about how most forms of energy are used to boil water in order to generate electricity. From there, I think about the hydrogen bonds that gives water its life-giving properties. From there, I think about carbon... and so on. Let's go down the cultural movement STEAM. I think about how it started as STEM in the early 2010s. Later on, society added art into the mix and it became STEAM. From there I think about my own relationship with art. I think about how much I appreciate art these days than I did in my early 20s. I even see some of my engineering decisions as artistic in nature.

I hope you can see that each sentence here, is its own dot. Some lead to a conclusion and some connect to other dots. At one point, this dot matrix was purely for academic understanding, but it was so useful that I unconsciously began using it for everything. This transition was sometime in college.

These dots aren't very well organized. There aren't topic headings, there aren't circles around sections. They live in a scattered mess while acting a bit like an organism. It's constantly shifting and adjusting based on new stimuli. It reorganizes itself to bring dots that are frequently used together, closer together.

Recently, I was learning about how the brain triggers a muscle to move. The neurons send an electrochemical signal to the muscles. These signals are short pulses of electricity. Neuroscientists' call the pulse an action potential. Electrical engineers use the exact same technique for running motors, they call it PWM (pulse width modulation). To make muscle move faster, the brain sends more pulses. To make a motor move faster, the computer sends more pulses. It's the exact same problem with the exact same solution.

I've spent a lot of my life working on robots and machines that use PWM. I've seen situations where there is too much noise in the wiring and that causes the motor to move in weird ways. This is analogous to Alzheimer's disease[^1]. I've seen situations where the computer is communicating significantly faster than the motor can react which causes it to move sporadically. This is analogous to a seizure [^2].

By connecting this small dot about neural signaling to the large dot of PWM, I'm able to access a whole host of additional information and experiences. If at some point in the future, I decide to learn more about neuroscience, I already have a framework to connect into.

Coming back to this picture of the dots in my head; neuroscience is an example of a huge gap in my knowledge. This can be visualized as blank space. Just to be clear, I have huge sections of blank space. If you ask me anything that isn't post-1850s science or engineering, I'm absolutely clueless.

Now, we get into the complex part. There is blank space between the dots and blank space around the dots. For the space in between the dots, I find I'm often able to interpolate and get pretty close to the right answer. To the areas outside of the dot cloud, I'm completely clueless.

There is something interesting about this process of interpolation. When I'm right, there might be a new dot that's introduced or it might just strengthen the existing connection.

When I'm wrong, a lot of change occurs. Being wrong, is an oversimplified way of saying I didn't understand the full complexity of the ongoing situation. My brain immediately zooms into this gap. The gap feels gigantic. I'll artificially add in a bunch of dots and systematically test each one until I find the fault. The artificial dot technique is a bit like building a staircase while standing on the staircase. I have to build a step and stand on it before I can build the next step. If I'm unable to put my weight on that step, then there's the fault.

However, in the real world, I rarely get the luxury of going through each step. I often have to bandaid the problem to keep the project moving. My mind won't let the bandaid be the true solution. My mind will run simulation after simulation to try to understand the true fault. The mental simulations will usually narrow the problem down to a few dots. But, I need real world testing to understand which one is actually the problem dot. I will often turn to research at this point.

This research phase, critically, is not only pinpointed at this specific fault. Instead, it is focused on the general area as to build many more dots and adjust the existing dots. Also, generalized problems are significantly easier research. This builds a robustness that prevents future versions of this problem.

Some faults take days to months to understand and others take years. The multi-year ones are questions that I have to put down in the short-term and if I'm lucky, it'll come to me in the future. It is deeply satisfying when I do find the truth.

One of the last major projects I did at Apple was calibrating and testing a sensor on the AirPods Pro. With the help of some amazing mentors, I had designed and manufactured a machine that could calibrate and test the AirPods both quickly and accurately. Most machines are either one or the other. However, it turned out our machine was very inaccurate. This was a shock to the team. I troubleshooted the machine, ran multiple tests and made multiple modifications. But I still wasn't able to improve the machine's accuracy.

At this point, I decided to leave the company. I dealt with this kind of problem countless times and was just tired. The associated politics don't make it any easier.

Just before I left, I met with the sensor team and starting talking to them. I found out that they were using an equation, Chi-squared, that assumed two variables were totally independent. But they weren't, they were dependent and that changes the underlying equation. I told them this in the meeting and after a short discussion they said they would get back to me.

Months pass by and I meetup with one of my former mentors. He told me that I was right. The equation was wrong and the product had a manufacturing issue. The equation made the system more sensitive to small variations and the product to product variation was huge. The issues weren't in my machine, it was in the product.

After learning this, I didn't walk away with hate towards the sensor team. It instead reinforced the dot that questions everything. If that dot had spoken up earlier, I could have ran different tests to prove that my machine was valid and the issue was with the product.

Using the Dot Matrix

When I'm faced with a new problem, my mind will begin path planning. The dot matrix is dense and I need to simplify it. You can imagine me looking at a screen of dots and my job is to tap all of the dots I want to use to solve this problem. I'm looking at these dots and I'm choosing the largest dots and thickest connections. The path with the largest area of dots will have the highest confidence. This becomes my primary path. But I don't stop there. I'll plan out a couple more that include completely changing the core set of dots. These are alternatives that I will keep in the forefront of my mind.

This path planning process is the most rigorous of all of them. Second most rigorous is working backwards. Surprisingly, there are a handful of situations where working backwards is significantly better.

A few friends and I were hiking in a national park and on the way back, we saw a family with a broken down car. First we gave the car a jump start but the engine only turned a few times and then stopped. We tried it again but had the same result. That's when I realized this was going to be more difficult so I began to apply the matrix. I quickly searched for the dots related to gasoline cars and tracked their connections. The starting dot is fuel and the ending dot is rotating wheels. In between are the dots related to the engine, transmission, starter motor, and electronics. All other dots fade into the background.

We spoke with the dad and asked if he noticed decreased power or any weird noises and he said no. There also weren't any error lights on the dashboard. So it was likely isolated to just the engine. For an engine to create an explosion, you need a chamber, fuel, heat, and oxygen. For the heat, engines use a spark plug. We verified the battery had power and checked the fuses. Since the car had been driven here and didn't have a decrease in performance, it was unlikely there was a major issue with the cylinders or pistons. Next was fuel, I spoke with the dad and verified that gasoline was used and it was not filled with diesel by accident. The last thing was to check was the air intake. However, we didn't have the tools to open up the intake. So we had to give up. We helped them get a taxi and a few days later the dad texted me saying that there was a buildup of leaves in the intake. If only my Swiss Army Knife included a socket wrench could I have solved it for myself.

You can see why I love escape rooms, The Martian, Macgyver and others. It's the ultimate test of the connections in my mind. There are a finite set of objects in this space and you have to use them to solve a difficult challenge. This is also why I love learning new skills. There are so many connections that a new skill can bring. Fairly often, I will focus only on the basics. The basics are where you make the most number of dots and connections. Intermediate and expert levels are dialing in the skills you learned in the basics. Yes, there is still more to learn in the higher skill levels but they aren’t as fruitful as picking up another new skill.

The car example above was of a pretty ideal scenario where I had a clear mind and great people around me. However, in most cases my view of the dot matrix is skewed. Here are some of the variables that affects my judgement: budget, novel vs. known tool, similarity to previous project, who I'm working with, timeline, criticalness of success, outreach if successful, access to equipment, success/failure of the most recent project, global emotions, and global energy level.

These biases can be viewed as a stack of lenses that hover above this dot matrix. Sometimes the lenses distorts the view to make the dots and lines look larger. Other times it makes them smaller. These lenses aren't uniform, the same lens can amplify one area and shrink another. They can also amplify a loop of dots which can force me into an infinite loop.

Now you should have an image of the dot matrix and the lenses above. There is one more layer and it's the problem solving layer. It lives below the rest of the dots. This layer attempts to super generalize the dots with the goal of finding other similar dots. The goal is to take the problem solving from one space and apply it to another space. This should be visualized as switchboard and I'm the operator who is connecting the dots together. I sit there and study the dots. If there are two that have a similar story then I connect them and see if anything useful comes from it.

Most of the time, they are garbage connections and they don't leave the confines of my head. On the rare occasion that I do make a great connection, it's absolutely wonderful. This is also why I focus on inputing more dots into my brain. You can connect more dots if you have more dots.

Therapy has worked on each layer. The problem solving switch board, the dot matrix itself and the lenses above. Therapy has helped me see and understand the uniqueness of the wiring I've created. It has also added dots and connections to each level. These are more than just gaps in my knowledge, they are a side effect of my upbringing and exposure.

The lenses stated earlier are fairly commonplace. In therapy, I've learned about so many other lenses that I was completely blind to. In hindsight, it's a little funny at just how much these lenses skewed my view. It took a very skilled hand to help me see these lenses. It also took me a long time to accept and understand the lenses. But now, I get to work with the lenses rather than fight them.

The Mountain of Dots

Once a path is chosen for a given problem. I then visualize the path as going up a mountain. The dots with weaker connections represent steeper sections of the mountain and stronger connections are shallower. The goal is to roll a large boulder up this mountain.

While rolling the ball up, a wedge can be used to prevent the ball from rolling down. This wedge is great in that if I come back to it in a few days the wedge wouldn't have budged. But if I come back after multiple weeks or longer, the wedge would have slipped a little bit. This doesn't represent work being undone, but it represents the time to re-familiarize myself with the given situation.

The mountain that I've drawn with these dots is just my best guess at how the project will go. Only when I'm actually rolling the ball do I experience the incline. For sections that are steeper than I imagined, I can choose to push through or divert. Choosing to push or divert is a skill I'm still working on.

If I'm forced to divert, I have to step away and let my mind work. Remember that this path had the most amount of confidence, so which dot(s) is wrong? Using the dot cloud, I can plot out a detour. It then becomes a process of testing the various detours. From experience, if I have to check more than a few detours, I'm likely on the wrong side of the mountain.

In my life, there isn't only a single mountain that I'm rolling a ball up; there are many mountains. Each mountain is a thing in life that I'm working on. It's a vast sea of mountains that each have their own boulder. These mountains aren't static, they are moving with the organism of the dots below. This is why making more connections is so important. Every dot and connection affects the sea of mountains. They create more detours, they causes mountains to overlap, and it raises the floor.

I don't see rolling boulders up these mountains as a chore or something I'm forced to do it like Sisyphus does. I don't feel the weight of the world on my shoulder's like Atlas does. I see the boulders as a personal challenge for me to take on. I see them as what makes life so interesting.

Sharing the Dot Matrix

I will often reference the dots in my mind with others but I don't share the map very often. I find that when I'm teaching someone, I will first figure out which dots they know of and build off of them. If I don't find their footing, then it turns into a "A becomes B which becomes C" kind of statement. They are following the words but have no understanding as to why 'A' turns into 'B'.

It's a skill that I'm still working on. I've made all sorts of mistakes, assuming the person know too much or assuming they know too little; both lead to disaster. I've learned to be blunt and ask about what they know.

Since finishing college, I've not stopped learning. I've taken many classes and now I've moved into hiring tutors for any topic that I want to learn about. When I'm learning, I'm placing and connecting the dots. There is the predetermined path the teacher has and then there is the path I'm drawing. Questions and comments are my way of testing if the path is correct. Some teachers are great at following this path and guiding me. Other teachers don't do this and I never actually understand what's happening.

A teaching style that I've been experimenting with is, share the question, then tell the answer, and then present the explanation. The question puts the first dot down, the answer puts the last dot down. The explanation places the in between dots. This way the teacher's path doesn't feel random. The student should be able to see the buildup to the answer.

The other benefit of this style is that I can build up the emotion for how big the problem was and just how successful the solution is. This can create a stronger level of drive to understand the dots in between.

Ateev Gupta
Initial Thought: 6/23/2022

[^1]: Alzheimer's disease is actually where the 'insulation' between the neurons starts to fade away which causes interference between neurons. The problem I'm referring to with PWM noise is if the power cables are too close to the PWM cables. Power cables are high voltage and PWM cables are low voltage. The high voltage cable will often interfere and create noise in the low voltage cabling which means the motor won't receive the correct signal.

[^2]: Technically speaking a seizure is when the brain get's into a positive feedback loop with itself. This means that it constantly triggers itself before it has chance to respond to the previous trigger. The body shakes that we are familiar with are just a side effect of the internal brain issue. The computer-motor model I refer to doesn't talk about an additional device in the middle called a motor controller. The motor controller and the computer are the ones that actually need to communicate at the correct speed, this is often called the baud rate. If the baud rate doesn't match, the motor will move sporadically.