Moore’s law of knowledge
Most people will have heard of Moore's law. It describes the exponential increase in computing power of microchips: every two year, you can get computations done more or less twice as fast. A similar law has been proposed for general human knowledge. Architect Buckminster Fuller postulates that human knowledge doubles every 12 months, in his book 'Critical path'. That's not all, the rate at which it doubles is also increasing.
A similar law will likely hold for scientific/engineering knowledge. The world population and GDP is increasing, meaning more money is spent on more people to do research. As a rough example, if knowledge doubles every 10 years or so, this means a postdoc will have had access to half the knowledge of the undergraduate he is supervising at the time he was an undergraduate. His professor would have access to only ¼ the knowledge when she was in undergraduate.
The ideal of the Renaissance man, a polymath specialized in a wide range of scientific disciplines, becomes ever harder to pursue. Nowadays people need to specialize to some degree (sometimes referred to as a I-shaped skill set) or work interdisciplinary and relinquish in-depth knowledge of any particular subject (sometimes referred to as a Y-shaped skill set).
The increase in knowledge also means it is hard to keep up, in particular in a bustling field such as AI. It is impossible to read everything in depth, especially for Y-shaped people who work in interdisciplinary topics. A key challenge is deciding what to work on and what to read, this means getting a high level gist without going into detail first. Some things I do to try to manage:
Focus on pre-selection by respected peers
Follow a few people on twitter, LinkedIn, Google Scholar or some other media that are in line with what you think is interesting. Other sources that present information in a simple manner (such as distill.pub for machine learning or good blog posts on Medium) are also great.
All this does mean you will be slightly behind the curve though.
Trust on ‘gut feelings’ (or simply subconscious reasoning)
This sounds strange and irrational, but often works. Most decision making is subconscious, if you have a good feeling about a topic, dive deeper. There can be a reason for it that is simply difficult to verbalize, but has a representation in some area of the brain.
Skim and iterate
Do not read a paper or source in detail. Skim the abstract, figures and introduction. After that, the first and last sentence of relevant paragraphs. If you are still interested, you can go into more detail. If information is presented well, this should be enough to get a gist of the idea, if it is not presented in a way that skimming works, the source may not be worth reading anyway.
Lectures instead of books/papers
Books are great if you are on vacation, but usually not a very efficient way to get knowledge. Typically, the author has a few main ideas that can also be explained in a 50 min lecture, that you can watch while cleaning or doing laundry.
The reason books are so long is that publishers need to justify asking 15 dollars for it. Also, do you really remember 400 pages of text half a year after you read it, or just the main ideas and reasoning that may have led to it?
