A typical day for researchers on Google’s Brain Team

What do you and researchers on Google’s Brain Team have most in common? You both probably spend a lot of time triaging email. In a Reddit AMA, 11 Google AI researchers took time to share the activities that consume the greatest chunks of their days. Email was a frequent topic of conversation, in addition to less banal activities like skimming academic papers and brainstorming with colleagues.

Google’s Brain Team is one of the company’s research groups tasked with performing research into artificial intelligence. The group is led by Jeff Dean, a Senior Fellow at Google and one of the key people behind MapReduce, among many other things.

Dean noted that he splits his day between sending emails, commenting on technical documents, meeting with researchers, reviewing code, writing code, planning and giving talks. As leader of the team, this is just about what we would expect — higher-level leadership and some Brain Team evangelization.

Everyone universally spends a lot of time reading papers on arXiv that are relevant to their research and collaborating with colleagues. Sara Hooker, a resident on the team, likes to chat with colleagues over breakfast — and lunch and dinner — to keep up to date on what other researchers are doing to solve similar problems.

Members of the team seem to like exposing themselves to applications of AI that they haven’t considered before — Hooker noted space discovery as one such application of interest.

A number of the researchers on the team retain some responsibilities at their home universities, and some even take the extra step of involving themselves in planning and management activities for key industry conferences like NIPS, the conference on Neural Information Processing Systems.

Most folks like getting their hands dirty, running experiments on huge, massive, gigantic, cosmic, colossal GPU clusters. Some engineers like Jasmine Hsu are even lucky enough to get to work with robots. She spends her days prototyping research ideas in simulations and testing models on physical bots. Generally speaking, few people like working at their desks all day.

We could all probably learn a thing or two from the time that researchers spend just thinking. Daniel Tarlow and Nicolas Le Roux, both research scientists, contemplate current projects and plan priorities away from other distractions.