Pat, I understand where your estimates are coming from, and I’m sure that your advice is truly meant to be helpful to me. But I also see that advice as an expression of a kind of anxiety which is not at all like the things I need to actually think about in order to produce good fiction. It’s a wasted motion, a thought which predictably will not have helped in retrospect if I succeed. How good I am relative to other people is just not something I should spend lots of time obsessing about.
I'd like to work meaningfully on helping people live longer and healthier. I keep falling into moods of "I'm not very smart, someone's probably already tried all the ideas I could come up with". Trying to figure out how efficient different parts of the world are is hard but important when deciding what to work on. But I don't need to pay attention to this hard-to-read signal; I can just look at what people have tried!
Keeping up with current research
There are lots of papers published everyday. What does keeping up with this look like? Here are some ideas:
- read all articles on bioarxiv
- replicate all articles on bioarxiv
- update by gears-level models
- read old important papers in the field
- think about core ideas in each paper
The amount of reading that seems required is just really high. Reading fast isn't my comparative advantage; can I avoid reading a lot? There are two things I could do here.
Reduce number of papers I read
This means I need some way to filter papers. What properties should this filter have?
- likely to falsify my model
- likely to be replicable
- likely to impact other things I'm paying attention to
I could also use tools like Elicit and Google Scholar to decide what to read.
Reduce amount of time per paper
Some ideas:
- an assistant to summarize papers
- an assistant to read the paper and then explain just specific sections in context of the whole paper
- get good at parsing "methods" section to find red flags
- can I find some "methodology reviews" or something to teach me what red flags to watch out for?
- get good at looking at graphs to find where stuff smells of p-hacking or error-bar
hacking
- I need to understand these two things in more depth then. That doesn't seem too hard.
Switching between breadth and depth
I shouldn't commit to just one of these as my "learning style". Depth-first is nice when doing a project, but it's a terrible idea when mapping out a field. This might not be a high-order bit though.
Making lots of beliefs that pay rent
A superpower here is understanding the epistemic state of the field:
- reading history of biology, what were the biggest mistakes and what orientations prevent me from making the same mistakes if I were in that person's shoes?
- read some blogs of people who point out epistemic problems in papers
- find researchers with really high epistemic standards and figure out why the papers I'm reading are skipping some steps. How could skipping those steps prove fatal?