What do you think about regularly reading papers outside one's own field, even those which are not related to one's area? My intuition is that it can give a totally different perspective or technique that may help someone in solving his or her own problems. But at the same time I am a bit skeptical about it as it will be a waste of time if the new knowledge doesn't help him in his own work.

What are your thoughts on this? Is there any other way one can look for results unrelated to his own work? What about TCS+ series?

• To clarify: by "outside one's own field", do you mean (1) outside one's subarea of theory, (2) outside theory (e.g., systems, programming languages), (3) outside computer science (e.g., mathematics, physics, economics, etc.)? Aug 20 '18 at 18:12
• Outside one's subarea of theory. Also, outside one's subarea of complexity theory.
– VRS
Aug 20 '18 at 18:40
• Is it beneficial to regularly read books that are not textbooks? Sep 26 '20 at 23:39

Understanding papers outside your field can be surprisingly difficult --- every field develops its own specialized terminology and shared background knowledge. A much better advice I once received from a very productive researcher was: Try to attend research seminars outside your field. These are typically more accessible than "cold" paper reads, and you have the advantage of being able to interactively ask questions. Just talking to people about their research can be extremely useful as well. You'll occasionally hear stories about how someone from field A happened to join a lunch table with people from field B and realized that he has the tools to solve their long-standing open problem (or, conversely, they have a tool to solve his).

• Here's an idea for a soft question: something along the lines of, which long-standing open problems in one field were solved with relative ease by importing techniques from another field? Aug 21 '18 at 0:35
• One reason it’s so hard to understand papers outside of your subfield is that not only do subfield a have distinct terminology and background knowledge, they often have very proof techniques. Someone who leverages combinatorial group theory do complexity theory isn’t necessarily going to be aware of why Fourier coefficients are so important in additive combinatorics. Combinatorial game theory looks almost nothing like infinite game theory (which is largely focused on computable properties of ordinals). TCS has a really broad set of skills and techniques in a relatively small space. Aug 21 '18 at 11:39

You mention TCS+, and as Aryeh's answer suggests talks and seminars may be a much easier and accessible way to learn about other results.

From my point of view (inherently subjective, and not too experienced, so take with a grain of salt) reading papers outside your subarea/subfield may be a good idea*, but going to as many talks and seminars as possible is a necessity. Below are a few things I'd suggest; in short, "going to talks, and absorbing information by reading abstracts."

• In your institution, subscribe to as many mailing lists as you can -- combinatorics, machine learning, theory, whatever your department or adjacent departments (math, electrical engineering for instance) have. You won't go to all talks, of course, but at least you'll receive the announcements, and specifically the titles and abstracts. (This sounds somewhat silly, but this is important -- reading abstracts and succinctly describing what is new and what is going on in the other fields.)

• Speaking of which, subscribe to arXiv notifications for several tags, or follow the TCS Blog aggregator and blogs like the Property Testing Review Blog. You'll get notifications about new papers, and sometimes a bit more info from a blog. (Disclosure: I am one of the writers for the PTR review blog.) Again, just reading abstracts and short blurbs about the new papers doesn't seem much, but it helps -- and if a paper actually sounds relevamnt, you can read it in full.

• when in conferences, spend some time going to "random" talks not really applicable to your research at first glance. It depends how much time you have, and how overwhelming conferences are to you (they can be pretty tiring), but attending a few of these can be helpful.

• You mentioned TCS+; this is indeed a good idea (full disclosure: I am one of the TCS+ organizers). Keep track of those talks, look at the slides, etc. There are other online seminars or recorded talks you may want to browse: e.g., the Shannon Channel, the Simons Institute videos (from their programs and workshops), the videos from the IAS lectures...

$^*$ if only to learn from the style and exposition of great papers.

As a grad student make use of the opportunity to take classes outside of your area and outside of CS. You can audit classes or attend a course partially. This is different from reading papers.