First and foremost I felt I need to do this with a purpose, an end goal. I would love to be able to diagnose profanity from uploaded pictures at the end of this, but it feels somewhat limiting to start with this one so I decided to explore goals that I can set at the end of my first week into this within an industry or cause. Hopefully I will have some more clarity on how I can contribute towards bringing “The Rising Billion” (Reference from “Abundance”) closer to western living standards.
I chose to review my Machine Learning Resources to start with and expose them here with some comments on their usefulness towards my goal.
What is #100DaysOfMLCode?
To start off, let me save you a few keystrokes:
- twitter tag #100DaysOfMLCode or facebook #100DaysOfMLCode
- 100 Days of ML Code Challenge
- wiki page for machine learning
- youtube is apparently full with information, not just for beginners Machine Learning on YouTube
Here are a few entries I went through from people already taking on the challenge:
- WhiteChills trying Support Vector Regression
- Lenilson Nascimento going through his learning live
- “The info grid” introduction of this tag
- mc.ai introductory post, he’s got some more Machine Learning articles on the blog, I really enjoyed going through Serving Pytorch NLP models on AWS Lambda
- A rather funny introduction by Trackster
So basically there are 2 main rules:
- dedicate at least One Hour a Day for 100 Days to learning or coding Machine Learning
- document your journey in order to make your dedication public
Why take on this challenge ?
- My belief is that learning something new adjacent to your daily work is tricky, a challenge like Raj’s can just about bring out some inherent motivation to the surface
- If you are considering a career in the field, taking on a challenge like this would help build something to showcase your skills and talents while learning, take a look at Engineered Thruth’s perspective on How To Get A Job in Machine Learning
- This is FUN!
- Also as a software developer reading What machine learning means for software development convinced me I need to start and extend my skills into this domain. By the way O’Reilly AI is an amazing source of information for beginners in this field.
- TRAIN AI 2018 - Building the Software 2.0 Stack convinced me of the paradigm shift being eminent. New paradigm: instead of engineered code, make use of ai to find code that does what you need, while paying in compute power
Where to find my code
Resources for study
My resources for trying to identify an end goal
reinforcement learning excels at problems that fall outside the realm of unsupervised and supervised machine learning
Despite these challenges, reinforcement learning is beginning to see real-world usage in areas like industrial automation
Bloomberg’s The Rise of AI Series
A few golden nuggets:
If you want to understand a brain, build one first! (engineering approach)
“This is not going to work” vs “I am not going to make this work”
if we are to build intelligent life it is likely to understand our values and how to make the best use of us, rather than a “Terminator” scenario
danger comes from misuse of AI, eg mass manipulation
Singularity University Summits’ Youtube playlist of Exponential Manufacturing 2017. This one is just packed with amazing ideas being explored, Additive Manufacturing and Collaborating Robots being my 2 favourites.
Rob May’s Inside AI
Rob is a founder and VC in this space with amazing insight and interest in the current AI ecosystems.
Resources explored on day one
Quick Note, I run by “A BAD Plan is BETTER Than NO Plan!”.
- Machine Learning APIs by Example (Google I/O ‘17)
- The Race to Quantum AI
- Machine Learning for Humans🤖👶
- Tensorflow.js Explained by Siraj Raval
- Backpropagation Explained by Siraj Raval
- A Feature Selection Tool for Machine Learning in Python
Unfortunately I didn’t get too much code on day one, this needs to change quick, looking forward for day 2.