Never repeating the same day twice
2021 Year in Review
There’s so much power in the ability to say I’ve never repeated the same day twice. It gives me a reason to get excited to learn/build/experience something new every day and allows me to conclude by reflecting on a new piece of knowledge/insight that I learned that day.
This idea is tied to claiming extreme ownership over my time and freedom. Despite constraints (like school), that sometimes trap us into a redundancy loop, it was up to me to make sure every day was unique in the value I was receiving.
2021 was freaking amazing, the amount of unconventional experiences, new relationships, knowledge gained, interesting conversations and “developing a great relationship with myself” (h/t Zayn).
Every single entity throughout the year (meeting, paper read, experience, etc.) was like a puzzle piece, contributing to the overall picture of who I am. Contributing in things like specific knowledge to hobbies to mental models to relationships and so many more.
Exactly like when solving a real puzzle, sometimes the pieces don’t fit. Other times, finding that perfect piece can uncover a major part of the final image. For example, let’s look at it from the perspective of what problems and emerging technologies I am interested in. For the first part of the year, I thought nanotechnology was it, but the deeper I dove, the more I realized it wasn’t for me. In the problems and technologies puzzle, this piece didn’t fit. However in the month of August with the kickoff of my Machine Learning Internship, I fell in love with Machine Learning, the theory and implementation. Ah, perfect puzzle piece *click*.
Progress as a Funnel
In the beginning of the year as I was just gaining exposure and knowledge, my progress was in a lot of different areas, but fairly slow in each of them. As the year went on, I started to narrow down my biggest priorities and progress increased. Hence, the funnel analogy.
The intention is not to make it seem like I had a perfect year, because that’s not true. It’s more just to reflect on how the events of this year shaped my trajectory forward.
January - June
A very wide but shallow entrance into the funnel. Progress in lots of different areas, but very slow in each area specifically, as I was excited to start gaining technical depth in nanotech and explore other technologies.
Hack the North - learning web development and JS
Presenting our “White-Glove Shopper Service” as consultants to executives at Instacart [Deck]
DNA Nanotechnology Simulation [Article | Video | Presentation]
Non-Invasive Deep Brain Stimulation Proposal [Video | One-Pager]
Consulting with the United Nations to Increase Female Employment in the Digital Economy in South Africa - One month sprint [Report | Deck]
Using Phage Therapy to Tackle Antimicrobial Resistance - One month sprint [Article | One-Pager]
Review paper on Non-invasive Nanoparticle Interfaces fro Brain Stimulation [Paper]
July
The funnel is starting to get a bit more narrow and deeper as a bunch of the focus areas from before are replaced. Main areas were academics, school and closing any loops from the previous year and prepping for the future.
Quantiphi internship interviews and acceptance
Summer school (reaching ahead)
Completing next year’s math and science curriculum
August
AI Month! The funnel transitions into its bottleneck with the only priority for the month being learning AI and ML, catalyzed by my internship at Quantiphi.
Quantiphi Internship full time
3 AI Courses
7+ AI projects
One major computer vision project at Quantiphi
September - October
The funnel bottleneck in AI continues as this month is focused on the Patrick Poirier: Solving the Impossible Award.
Researching into AGI Approaches
Presenting my vision to judges [Presentation]
Winning $250k Crowdfunding award [Info]
Quantiphi Internship Computer Vision Project
October - November
Narrows down further as I explore my passion for improving AI.
Quantiphi Internship - Transitioned to a client project
Read the first part of ‘Neural Networks and Deep Learning’ by Michael Neilson
Completed ‘Essence of Linear Algebra’ Youtube course [All Notes]
Read ML papers + explored different concepts to improve AI
What I’m up to now
In early December, I was first able to articulate what I am working on:
Applying neuroscientific principles to improve current day AI
I’m currently evaluating the best way to tackle this goal, with a focus on learning through building > reading.
That is how I discovered Capsule Neural Networks, which is the state of the art way to accomplish my goal, specifically when it comes to mimicking the human visual system.
Capsule Neural Networks were developed by Geoffrey Hinton in an attempt to perform inverse graphics, which is taking a visual entity and breaking it down into its parts and each part’s pose parameters. It works by using capsules instead of neurons and an algorithm that sends lower level capsules output to higher level ones that agree with its prediction.
Throughout December, I have been working on developing a super in depth understanding by watching the presentations (insert one more thing) and reading both the papers. The papers have been a major learning curve because I have never gone super deep to understand everything (especially the math), but it’s been super fun!
The main goal for this understanding is to then build my own implementation of Capsule Neural Network by hard-coding all of the functions, which can then serve as a sandbox for experimenting with my own ideas such as different algorithms and architectures.
🔑 s to Personal Health
I love the phrase “developing a good relationship with yourself” because that captures exactly what it is: a relationship. It takes time, effort and consistency to maintain. Below are some quick thoughts on what were the most successful for me this year and what I want to double-down on in 2022.
Focus on what you can control, accept what you can’t
On a podcast, the guest recommended that his number one exercise is to write down everything you are worried/stressed about. Cross out what you can’t control, and put all your energy into fixing the things that are left.
Exercise
Maintaining a strict exercise schedule was first achieved by understanding that exercise can be fun, specifically cardio. Instead of forcing myself to run on the treadmill or do a bike workout, I used to play tennis several times a week. In cases where that is not possible, refer to the Discipline → Will to Power section below.
FORMULA 1!
I absolutely love F1, and finding this productive hobby was so important to me throughout the year because it was a proper way to relax and enjoy as opposed to scrolling through social media.
Below are different analogies of how lessons from F1 can apply to real life:
I watched almost every single race this year and what stood out to me the most was the number of variables every team had to take into consideration, as opposed to just the skill of the driver and the performance of the car. Tuning the downforce of the car, the race strategy, the tyre wear, teammates and more all contribute to the ultimate successes and failures.
Things go wrong in F1, and sometimes there’s nothing you can do about it. Tires pop, someone else crashes into you, the weather, etc. After the race, they identify what went wrong and come up with a plan to prevent it from happening again (Car vs. Plane Crash section below). This same principal can be applied to accepting what you can control and not taking stress for the rest.
What I found very appealing is that the goal of each driver is very clear: overtake the next car. When you are trying to overtake a car ahead of you which has a similar top speed, all the driver’s energy goes into trying out different approaches until you’re successful. But, it’s at 300 km/h, so one wrong move could be very dangerous.
I think the above point can be applied to goal setting. One goal, all your energy towards achieving it, but you still have to balance other priorities (tire degradation), evaluate the risks (the driver’s safety) and remember that it’s a sprint but still part of a longer marathon (pushing too hard and crashing the car → negatively impacts points in entire championship). What I mean by the last point is sometimes it’s worth it to give up a position to be able to perform again next time.
Car Crash vs. Plane Crash
I’ve had a lot of failures this year. A lot of them maybe not public, but I know that I’ve failed and learned a lot. It’s almost common knowledge that learning comes from failure, but I disagree. I think learning can come from failure, but it isn’t guaranteed, which is the trap I fell into a couple times this year.
As outlined in the tweet above, the two approaches for closing the loop post-reflection are car crash and plane crash. Essentially, car crash is adopting the “it’s ok I will fix it next time” mentality whereas plane crash is to immediately create a plan of action so it never repeats.
I failed at this multiple times, which led to me repeating the same mistakes and leaving more unclosed loops. This year I started doing proper daily, weekly, monthly and now even yearly reflections, with the goal of capturing and eliminating all the pitfalls, using the plane crash mindset.
The area where I saw the most growth was working in teams after a big project. Doing a proper post-mortem after every project allowed us to reflect on the experience, identify what went wrong, why it went wrong and how to avoid it in the future, in that exact order.
Serendipity
By writing a random article on something I was researching, it was read by the right people and then led to my full time summer internship.
By deciding to apply to something I saw online on a random Sunday, I ended up as a finalist to win $1m.
The thing in common: serendipity. The art of expanding the surface area for unplanned good things to happen. Serendipity has changed my and so many others’ trajectory for the better, hence the use of the word “unplanned”.
Although my aim for 2022 is to continue progress in the few specific areas of highest priority, I am intentionally going to seek out unconventional experiences and serendipitous opportunities in various other areas by continuing to build in public plus attending conferences and programs.
4 very simple things you can do to create some serendipity right now:
Build in public - Post about what you are working on right now
Make a piece of content - Content is timeless and can lead to opportunities in the future
Coffee chat someone - serendipity is always activated by people
Cold outreach - expand your network to meet more people
Hence why maximizing serendipity is one of the major drivers of my 2022 goal to double down on networking and relationships.
Discipline
I view discipline as prioritizing long-term goals over instant gratification. Eating a piece of cake now versus eating a fruit to maintain your overall health. Like any form of conditioning, discipline takes time to train.
I’m no expert on this, but I have found good ways to effectively get the hard things done. My entire view on procrastination shifted after hearing this quote: “Distractions are caused by a desire from the inside to escape negative emotions.” Every time I wan’t to procrastinate or get distracted from the work I am doing: the simple question of “why am I avoiding this task?” “What negative emotions am I avoiding” This is often enough to force a quick reflection and come to the conclusion that delaying the work doesn’t solve the emotion.
Another method is the one developed through a conversation with Mikael and Zayn, derived from Fredrich Nietzche’s Will to Power philosophy. Essentially, the takeaway is to zoom out and view each activity as a step towards a larger goal, and if we truly care about achieving the end goal, we will do the hard thing. A great example of this is school. It’s boring, its redundant and it doesn’t always feel like a good use of our time. However whenever I think about why school is important to my journey, I understand that post-secondary is crucial and school is one of the steps I need to take to get there.
Uncovering a cool insight
In early 2021, I was super excited about nanotechnology and its applications and the problems it would be able to solve. However, after February I was massively demotivated and my progress flatlined, until I decided to finish my review paper (A Review on Non-Invasive Nanoparticle Interfaces for Deep Brain Stimulation), and final project (Designing the Future of Medicine: An Artificial Red Blood Cell), before which I moved on to AI.
The activation energy required to break out of the loop of not wanting to research nanotech was much higher than the actual energy required to complete the paper, once I just got started. I had heard others share this same insight but it finally clicked when I was able to experience it.
This leads to my next point about engineered interest versus genuine interest. When I decided to commit to the field of nanotech, my only datapoint was some articles and Youtube videos describing all its potential applications from space elevators to curing cancer, but I hadn’t taken the time to experiment and gain experiential learning (h/t Asif, who highlighted the importance of this point in our first meeting). Therefore, once I had dove deep enough past all the buzzwords and hype, I wasn’t interested in building with the technology.
For this reason precisely, when I was learning AI for my internship, I made sure to spend time learning, building, failing and debugging in a “trail period” of sorts, focused on experiential learning > researching.
My period of maximum growth
August 2021 was by far my period of highest growth, just as I was entering my first full-time summer internship @ Quantiphi and transitioning from the field of nanotech into AI.
Fueled by the excitement of trying something radically different and surrounded by incredible smart mentors (h/t Darien, Asif, Biplab and Priya) I was able to complete an entire course and start building full neural networks, coming from absolutely 0 background, within a week. (Scroll up to August to see the results for the entire month).
TLDR of why: I had day one mentality, and one of my daily goals is to employ that attitude every single day this year.
What’s even more interesting to me is that despite the extremely rapid timeline, I still almost remember everything I learned from those initial courses. Why?
applying the initial concepts learned right after
continuously revisiting (in the current ML work I’m doing)
breaking everything down to first principles when I was learning (ex. a filter → a cnn)
Priorities for 2022
Building something novel in ML x Neuroscience
After discovering about the field of applying neuroscientific principles to improve AI and getting my feet wet with Capsule Neural Networks, my goal is to completely 10x my growth in this area.
Instead of replicating architectures (which is great for learning), in 2022 I want to be able to design my own models and/or algorithms to implement my vision of improving current day AI.
What this looks like specifically, I don’t know. And thats intentional.
Exponential growth happens in periods of ambiguity. As Sara Blakely once said “disruption is not having any idea how it’s supposed to be done.” It’s when we don’t know what to do next when major progress and growth occurs.
Learning the Fundamentals
In order to accomplish my previous goal, I realized I have been approaching this wrong. While I have some knowledge about the fundamentals of ML, majority of my learning has been in through the top-down approach.
Essentially, when I start with a research paper and understand the math or implementation for that paper only. This approach works for specific projects, but if my goal is to come up with unique ideas, I need to have an incredibly solid knowledge of the fundamentals, combined with knowledge of state-of-the-art models. A combination of top-down and bottom-up approaches.
Top-Down: Focusing learning and building on a specific architecture and understanding the math only relevant for that architecture
Bottom-Up: Developing transferable knowledge of the math and other fundamentals, making it easier to apply when learning different topics.
The three main fundamentals I am focusing on:
Math
Math is the building block of all machine learning, and currently my knowledge is very limited. Specifically linear algebra, analytic geometry, calculus and probability and distributions.
As part of a study group, I am going through the Mathematics for Machine Learning textbook which covers all the concepts and gives practice questions.
Programming
As pointed out by a coding test I had to take as part of an interview, my programming skills and logic can definitely be improved (basic machine learning projects are actually very easy to code).
The current plan is do daily DMOJ/Leetcode problems to practice the skill. I will also be taking part in the Canadian Computing Competition in February.
Neuroscience
A large portion of the research I need to do for my projects are in neuroscience to understand how the different parts of the brain work. Once again, my knowledge is limited and a combination of reading books, papers and talking to experts is my plan of action.
Independent Thinking
I’ve spent this year surrounded by incredibly smart people and consuming content by some of the world’s smartest thinkers. However, reflecting on the discussions, conversations and insights from consuming content I realized that I need to be more of an independent thinker. Especially in a group setting, it is very easy to agree with what someone is saying without questioning and developing your own insights, questions and perspective on the topic.
Independent thinking is the ability to understand various perspectives, mental models and to come up with unique insights and judgements based on their knowledge. One of my major goals for 2022 is to become an independent thinker and a couple ways to kickstart it:
Passive listening → Active listening. Question what is being said and how it works
Seek out other perspectives
Getting rid of all preconceptions
Developing my Network and Relationships
As I have learned as I venture deeper into my journey, people are everything. People are responsible for making things happen, unlocking hidden doors and even contributing to providing knowledge, mindsets and skillsets.
One of my major goals is to double down on prioritizing network and developing high quality relationships. Although I had several meetings this year both technical and coffee chats, I want to prioritize more meetings, and better meetings.
If you know anyone in the field of machine learning or working in an AGI company (OpenAI, Deepmind, Sanctuary AI, etc.) or working in a machine learning lab, please reach out to me!
Shoutouts:
Mikael, Sri, Zayn, Dev, Anush, Kabeer, Arnav for pushing my growth but for allowing us to all to grow collectively.
Harrison, Michael, Patrick, Brandon and Kim for being amazing mentors who are genuinely interested in helping me level myself up. Each of you have helped me grow in so many ways, and I thank you for all the time you’ve spent.
Darien, Asif, Biplab, Priya, Vishal and everyone else part of the Quantiphi team that welcomed me and taught me so much.
All the meetings I’ve had that provided a unique perspective, insight or some new knowledge to contribute to my projects.
Thank you so much for reading this letter and being a part of my journey. I am super grateful for all of you who have supported me this year.
2021 was amazing, but now it’s just the benchmark. I am going to make 2022 exponentially better.