I am taking some blogging advice that I’ve read elsewhere: if multiple people ask you similar questions offline, answer those questions in a blog post!
It has been awesome to hear that others are eager to apply for the next cohort of OpenAI Scholars. I wrote a bit about my 2018 summer takeaways in my final “official” blog post for the program, but I would like to expand a bit on:
Note: Much of what I will say is subject to change since I participated in the inaugural Scholar cohort. OpenAI will naturally iterate like any good engineer 😊
UPDATE 10/11/18: Applications for the 2019 Winter Scholars Program are now open! Apply by November 16, 2018!
The OpenAI Scholars program is a really great opportunity for self-motivated newcomers to deep learning. It is important to recognize that Scholars are asked to set aside their regular schedules for 13 weeks of intensive self-study, with a touch of guidance from their assigned mentor.
What you put in is what you get out. It starts with the syllabus you create for your own program (here was mine). I decided to focus on natural language processing; some Scholars had other focuses like reinforcement learning, while others took more of a broad survey of deep learning approaches.
Preparation can only help. While the Scholars program is especially geared towards software engineers with little-to-no deep learning experience, it can’t hurt to become familiar with the subject matter of the program. Prior to the summer, I had taken (a now outdated version of) Andrew Ng’s Machine Learning course and fast.ai’s Practical Deep Learning for Coders. My program mentor Natasha Jaques also found Udacity’s Deep Learning course really useful personally, and it is taught by a research scientist at Google AI. I still had a ton of learning to do over the summer, but it was helpful not having to start at ground zero.
13 weeks can fly by fast. The time pressure this summer was really real. This is all the more reason to try to keep to a schedule. Even so, nearly every Scholar in my cohort created an overambitious syllabus. This turned out to be fine, generally – we could adjust our syllabi and schedules as we went. I even wrote a little bit about the craziness of my schedule in week 5, when I determined to not get discouraged by the pace, as long as I kept working and making progress. I’d really encourage this mindset – my final project turned out fine in the end 😊
Learn by doing. Textbooks are cool, but being able to apply a new skill each week (or week-and-a-half) is really rewarding. Re-implementing tutorials and running experiments in Jupyter notebooks were helpful activities week to week.
Other miscellaneous tips:
When applying, I would suggest highlighting what about your personal story makes you an interesting candidate for this program. Also make sure that your strong technical aptitude comes across clearly throughout the application.
Be aware of both the benefits and requirements of the program.
These reflections are about ways I personally think I could have approached the summer a bit differently.
I hope this provides useful information for anyone considering applying to be a Scholar! Overall, it was definitely a worthwhile summer for me.