My journey from DeepMind intern to mentor
Former intern turned intern supervisor, Richard Everett, describes his journey to DeepMind, sharing ideas 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 liked enjoying multiplayer video video games rising up. The interactions between human gamers and seemingly clever computer-controlled gamers fascinated me, and I dreamed a couple of profession in AI. This dream led me to pursue an undergraduate diploma in pc science; a standard (however not unique!) pathway into the business. Nevertheless, 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 rapidly 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 proposal. Nevertheless, having met a bunch of nice researchers I made a decision to reapply the next 12 months and received 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 by way of the identical expertise.
Are you able to describe the internship interview course of?
The interview course of was thorough, nevertheless it’s advanced since I utilized. Right this moment’s interns can count on the complete course of to final only a few months, which features a technical and a crew interview. In my software, I listed the researchers that I used to be notably fascinated by working with, and was fortunate sufficient to talk with them after my technical interview. I used to be so excited. This was a singular 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 by way of 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 pc science. For instance, reviewing linear algebra, calculus, likelihood, algorithms, and knowledge buildings. I additionally practised some coding workouts the place I attempted to speak by way of what I used to be doing.Â
For the crew interviews, I reviewed the crew’s latest work (e.g. papers, weblog posts, articles, talks), and considered how my work might relate to it. I additionally got here up with a brief record of questions I wished to know extra about, just like the collaboration fashion of the crew and the way previous internships had labored out.

What was it like whenever you joined full-time? Â
It took me a very long time to seek out my footing! With so many thrilling tasks happening and good individuals to speak to, working at DeepMind usually seems like being a child on this planet’s biggest sweet retailer. For interns, growing and specializing in only one mission 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 take pleasure in supporting new starters by way of this course of who’re experiencing the identical pleasure for the primary time.
Why did you become involved with the internship programme as a full-time worker?Â
Having gone by way of 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 assist throughout my internship, I wished to provide that very same assist to future interns. Because of this, I now coordinate my crew’s internship programme and am in a number of teams which repeatedly 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 kind of labor do interns do?
It’s at all times thrilling to see what interns determine to pursue throughout their time with us. On my crew (Recreation Idea 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.Â
To spotlight only a few 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 idea 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?
Briefly – type and collaborative. Over time, I’ve heard dozens of interns and new starters make the identical comment: “I can’t imagine how pleasant and supportive everyone seems to be!”. The period of time, power, and assist that DeepMinders give one another is exceptional, and this extends all the best way from firm veterans to first-day new starters. Everyone seems to be at all times glad to seize a espresso to speak, focus on their work, share suggestions, and companion collectively on tasks.
For instance, considered one 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 sometimes 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 Recreation Idea and Multi-Agent crew work carefully with DeepMinders from each the London and Paris workplaces.

From left to proper, a subset of the mission’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 ideas for aspiring DeepMind interns?
For college students who’re getting fascinated by AI, there are a great deal of simply accessible resources accessible so that you can independently be taught extra concerning the business and DeepMind: from papers, blog posts, and talks to open-source code, demos, and tutorials. It’s simpler than ever to get caught in! It’s also possible to become involved in workshops and conferences, a lot of which supply 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 courses, engaged on tasks with them, after which reaching out to different researchers within the areas that excited me.
DeepMind is made up of type, collaborative, and pushed individuals from all walks of life, and our internship programme displays that. Whether or not you’re an undergraduate or PhD pupil, finding out a technical, bodily, or social sciences topic, and have AI/ML expertise or not, there may be doubtless an internship alternative for you. We provide internships throughout numerous groups in Analysis, Engineering, Science, Ethics & Society, and Operations.
Having gone by way of 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 proficient college students who incorrectly imagine that DeepMind is out of attain or that their abilities are inadequate, and due to this fact 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 loads to achieve.Â