What is the adventure?
The machine learning scholar adventure (MLSA) is a self-directed learning journey designed to ramp up my knowledge in machine learning rapidly. It’s inspired by the structure of the OpenAI’s Scholars program, which works by providing a stipend for 4 months for full time study, pairing you with a mentor and also giving cloud credits for machine learning experiments! Even though, I’m not able to pursue it (e.g. one limitation is that only those in the US can apply), I believe that the most important aspects can be replicated!
Consequently, I will follow a curriculum, blog regularly (at least weekly) and produce an open-source project. Hopefully, I’ll be able to present at an ML meetup too. To help me succeed, I will be obsessively applying the framework from Ultralearning by Scott Young. By using the optimised principles, I’ll make the most of my time, focus my energy effectively and update the project sensibly to keep on track. Although I don’t have all the pieces of the puzzle figured out (noted below), I’m confident solutions exist and the universe will assist in proportion to my efforts to elevate my skills and share understanding. Right now, I’ve got to dare, be bold and move forward with my vision.
Why did I create this initiative?
At a strategic level, machine learning technology is advancing at an incredible rate and transforming more of our lives with each passing day! However, these innovations don’t arise spontaneously but are driven by the passion, energy and talent of human beings. At the same time, humans are fallible…if unchecked, our biases and irrational world view negatively affect everything we do. This is concerning because technology never exists in isolation. Whilst tech provides massive leverage, the direction of change is not automatically guaranteed.
Whether it maximises the potential and freedom of people or entraps humanity by limiting options and manipulating their behaviour depends on how consciously it’s created. The world’s top AI labs realise this so there exist several initiatives to include greater diversity of thought from different perspectives and backgrounds. I wholeheartedly support these endeavours, machine learning and artificial intelligence must be as positive as possible for all beings!
On a more personal level, I’ve always been fascinated with the human experience, with our minds, our vessels, emotions, behaviour and everything that entails. What is consciousness? What is intelligence? How do I know I’m aware and how exactly does reality arise? Are we really in a simulation? Overtime, this has fueled a passion for artificial intelligence. How? Well, to explore the aforementioned questions, one way forward is to try and re-create intelligence. At the moment, it doesn’t seem like we’re close to answering these mysteries but we have progressed and machine learning today incorporates the latest realisations humans have made about learning and creativity.
With my pursuit of the MLSA, I hope to advance my knowledge to a point where I can make significant contributions to the field! At the same time, I wish to inspire those who are underrepresented in the field. Without positive role models and examples, it can be hard to find the self-belief necessary to pursue a path less travelled. This would be to the detriment of society since it’s on the unexplored paths that all discoveries lay hidden. Who knows how many inventions have gone uninvented? How much art has gone unexpressed? How much value remains unrealised? How many problems remain unsolved? I want to do my part to address this issue! How am I beginning? By starting with myself, elevating my capability so that I can chart a path forward for those who are unaware of this possibility!
“Be the change that you wish to see in the world.”Mahatma Gandhi
What is my technical background?
My current technical experience stems mostly from the excellent introduction to computer science and programming and using python for research courses on edX. For machine learning, I’ve explored the helpful fundamentals of machine learning tutorial, the fun make your own neural network book by Tariq Rashid and completed Kaggle’s intro to machine learning microcourse. I’m currently playing through the insightful grokking deep learning by Andrew Trask. A while ago, I also completed version control with git and shell workshop on Udacity plus reviewed my mostly forgotten linear algebra using this recommended article!
I have never programmed professionally but I’m confident I can build on skills I’ve developed so far. The two other courses I have found extremely valuable are learning how to learn and mindshift on Coursera which help you to learn faster. How? They teach you about metalearning which is all about looking at how you learn and how to make it more efficient and enjoyable! I think that’s a good foundation and whatever else I’m missing, I’ll pick up on the way.
How much time will I dedicate to this initiative?
The OpenAI Scholarship programme is full-time for 4 months but in my case, I anticipate it will take more time. Why? I’ve got no stipend so I’ll need to also work to support myself. Luckily, I am able to commit the first two months (April and May) to the MLSA exclusively so I’ll be front loading as much study as possible in this period. From June, I’ll be working part-time so will have less time to devote to machine learning. Hence, I think it will take 5-6 months to complete the MLSA so I’m placing a light deadline of August 31st and a hard deadline of September 30th for the project.
What excites me most about the MLSA?
I’m excited to play through my MLSA and especially deep learning because of the incredible positive applications of the technology. From image recognition to self driving cars to world class performance in games like Go and even in fundamental scientific research, it’s been phenomenal to see how the combination of more computing power, availability of data and neural networks combined with intellectual ingenuity has ignited a resurgence into the quest for artificial intelligence!
I remember my past video game play (faves: Zelda series and World of Warcraft) and it’s super cool how we’re now using video games as environments to train ML agents. Having looked at some past projects from previous cohorts of OpenAI Scholars and also briefly at the directions of research from OpenAI and Deepmind, I’m most interested in deep self reinforcement learning as it’s the field of machine learning which seems (from my current understanding) closest to strong AI. Will also be cool playing with video games from a totally novel angle. With this in mind, I’m taking advantage of Udacity’s 30 days free access to delve into their deep reinforcement learning nanodegree. It’s been two days since I started and it’s really engaging so far.
What are the missing puzzle pieces?
For my vision for this initiative to be realised, I need to figure out the following:
- Find a mentor to help with:
- Verifying the MLSA project design is optimal
- Guidance with learning e.g. asking questions/feedback
- Accelerate my progress e.g. helpful resources
- Workout how to get free credits for training
- Get a part-time hustle which frees up maximal learning time
My belief is that by learning obsessively, creating consistently and sharing value, those who see the realised passion behind this project will be inspired to help. The first two items are the most important for me as the last one is something that I have greater control over.
What are the next quests on the adventure?
- Study more of grokking deep learning + deep learning reinforcement nanodegree
- Create a curriculum page to share resources with others ✅
- Write a blog post to document learning ✅
Moving forward, I’m grateful for all the failures I’ve experienced in previous learning endeavours. Reflection and analysis from those experiences have made me stronger and brought me to this ultralearning opportunity. This is going to be a wild adventure!
On that note, it’s time to play, let’s go 🚀