My journey from DeepMind intern to mentor

Former intern turned intern supervisor, Richard Everett, describes his journey to DeepMind, sharing suggestions and recommendation for aspiring DeepMinders. The 2023 internship functions will open on the sixteenth September, please go to https://dpmd.ai/internshipsatdeepmind for extra info.

What was your path to DeepMind?

Like many individuals, I cherished enjoying multiplayer video video games rising up. The interactions between human gamers and seemingly clever computer-controlled gamers fascinated me, and I dreamed a few profession in AI. This dream led me to pursue an undergraduate diploma in laptop science; a standard (however not unique!) pathway into the trade. Nonetheless, after engaged on a number of analysis tasks with my professors, I developed a style for analysis and determined to proceed on in the direction of a PhD.

Across the time I began my PhD, a small startup known as DeepMind was acquired by Google. As I regarded nearer at their analysis, I shortly discovered it inspiring my very own analysis, and so in 2016 I made a decision to use for an internship. After a handful of interviews with engineers, researchers, and program managers, I didn’t get a suggestion. Nonetheless, having met a bunch of nice researchers I made a decision to reapply the next yr and acquired the internship. That have led to a full-time provide and I’ve been right here since, engaged on AI and serving to interns who’re going via the identical expertise.

Are you able to describe the internship interview course of?

The interview course of was thorough, nevertheless it’s developed since I utilized. Immediately’s interns can count on all the course of to final just some months, which features a technical and a crew interview. In my utility, I listed the researchers that I used to be notably excited by working with, and was fortunate sufficient to talk with them after my technical interview. I used to be so excited. This was a novel alternative to speak about my previous work and brainstorm potential internship tasks with world-class researchers I had adopted for years, and ask them questions on DeepMind.

My recruiters had been extremely useful in guiding me via the method and offering sources to assist put together for the interviews. For the technical interview, I ready by revisiting my first-year undergraduate programs on arithmetic, statistics, and laptop science. For instance, reviewing linear algebra, calculus, likelihood, algorithms, and information constructions. I additionally practised some coding workout routines the place I attempted to speak via what I used to be doing.

For the crew interviews, I reviewed the crew’s latest work (e.g. papers, weblog posts, articles, talks), and thought of how my work may relate to it. I additionally got here up with a brief record of questions I wished to know extra about, just like the collaboration model of the crew and the way previous internships had labored out.

What was it like whenever you joined full-time?

It took me a very long time to search out my footing! With so many thrilling tasks happening and sensible folks to speak to, working at DeepMind typically appears like being a child on the earth’s biggest sweet retailer. For interns, creating and specializing in only one challenge out of so many is difficult, particularly inside a restricted timespan. This was a problem I discovered in my very own internship, and these days I get pleasure from supporting new starters via this course of who’re experiencing the identical pleasure for the primary time.

Why did you get entangled with the internship programme as a full-time worker?

Having gone via the internship expertise myself, I can relate to what our aspiring and present interns undergo. It may be nerve wracking, thrilling, complicated, and galvanizing, all on the identical time. After receiving a lot help throughout my internship, I wished to provide that very same help to future interns. Because of this, I now coordinate my crew’s internship programme and am in a number of teams which constantly search to enhance the programme throughout DeepMind. I additionally interview, mentor, and handle interns, in addition to spend time reaching out and speaking to potential candidates (e.g. at GraceHopper, NeurIPS, and analysis talks).

What sort of labor do interns do?

It’s all the time thrilling to see what interns determine to pursue throughout their time with us. On my crew (Sport Principle and Multi-Agent), we work carefully with interns to co-develop tasks which they’ll make their very own, and this has led to an unbelievable vary of tasks through the years.

To focus on just some public examples, interns have designed new multi-agent environments (e.g. impressed by the social deduction game Among Us and assembly lines), developed infrastructure for studying human-agent interaction, used cooperative sport concept for language models and negotiating team formation, labored on multi-agent inverse reinforcement learning, uncovered adversarial examples for reinforcement learning, mastered the game of Stratego, and utilized evolutionary game theory to online learning.

How would you describe the tradition at DeepMind? And your crew?

In brief – variety and collaborative. Over time, I’ve heard dozens of interns and new starters make the identical comment: “I can’t consider how pleasant and supportive everyone seems to be!”. The period of time, vitality, and help that DeepMinders give one another is exceptional, and this extends all the way in which from firm veterans to first-day new starters. Everyone seems to be all the time joyful to seize a espresso to speak, focus on their work, share suggestions, and accomplice collectively on tasks.

For example, one in every of my favorite tasks at DeepMind (Learning Robust Real-Time Cultural Transmission without Human Data), got here from the shut collaboration between artists, designers, ethicists, program managers, QA testers, scientists, software program engineers, analysis engineers, and extra over two years. This various and collaborative tradition additionally extends to our internships, with intern tasks usually involving a number of collaborators and advisors from throughout the corporate (spanning roles, groups, and even workplaces!). For instance, a number of of our interns on the Sport Principle and Multi-Agent crew work carefully with DeepMinders from each the London and Paris workplaces.

From left to proper, a subset of the challenge’s authors: Ashley Edwards (RS, London), Miruna Pislar (RE, Paris), Kory Mathewson (RS, Montreal), Alexander Zacherl (Designer, London), Richard Everett (RS; London), Edward Hughes (RE, London), Avishkar Bhoopchand (RE, London).

Any suggestions for aspiring DeepMind interns?

For college students who’re getting excited by AI, there are a great deal of simply accessible resources accessible so that you can independently study extra in regards to the trade and DeepMind: from papers, blog posts, and talks to open-source code, demos, and tutorials. It’s simpler than ever to get caught in! You can too get entangled in workshops and conferences, a lot of which provide pupil reductions and mentorship alternatives (e.g. Deep Learning Indaba, Cooperative AI). For me, I discovered my love of AI analysis by speaking to professors about their analysis in-between lessons, engaged on tasks with them, after which reaching out to different researchers within the areas that excited me.

DeepMind is made up of variety, collaborative, and pushed folks from all walks of life, and our internship programme displays that. Whether or not you might be an undergraduate or PhD pupil, finding out a technical, bodily, or social sciences topic, and have AI/ML expertise or not, there may be probably an internship alternative for you. We provide internships throughout numerous groups in Analysis, Engineering, Science, Ethics & Society, and Operations.

Having gone via the method myself (twice even), I can fully perceive and relate to how intimidating it may be to use. I’ve spoken to so many extremely gifted college students who incorrectly consider that DeepMind is out of attain or that their expertise are inadequate, and subsequently don’t even apply. If you’re contemplating making use of for an internship, my real recommendation for you is to only do it. You don’t have anything to lose, and maybe each you and DeepMind have rather a lot to realize.