I don’t know what research is (yet).
Anything we seriously pursue in life is foggy at the beginning. We don’t really know what we’re doing, whether it is made for us, or whether we are made for it. Sometimes it even seems like we are doing the worst error in our existence, but at the end it is what will make our success. The story of Steve Jobs is something I love to remember about this, read it here.
In this small essay I will be trying to present the complex concept of ‘research’ as seen by the following researchers: Jason Wei (Meta Superintelligence Labs), Tom Silver (Assistant Professor, Princeton University), John Schulman (OpenAI cofounder, worked at Anthropic, chief scinetist at Thinking Machines), Michael Nielsen (Research Fellow at the Astera Institute), Andrej Karpathy (OpenAI cofounder, founder of Eureka Labs), Richard Hamming (American Mathematician) and Fei-Fei Li (Professor in Computer Science at Stanford University).
Truth won’t come to us easily (and will always come partially), we must dig …
Before research
As Jason Wei said in his blog post about Practicing AI Research: “doing research is a skill that can be learned through practice, much like sports or music”. Many of us have this early feeling that we “have a taste for research”, like an athlete or a musician who just knows they love it , even before they fully understand it. We start without really knowing its essence, its implications, what comes after it, or which mistakes to avoid in it. This shows, in my opinion, that research is not an innate talent reserved for a selected elite, but a direction to pursue for everyone who has a genuine curiosity and that is ready to put the dedication and the work on it.
But we cannot work on anything ! In the same blog post, Jason accentuate in the necessity of working in something that is fulfilling to us, it may be something impactful to community, fun to us, or something really crazy (machine learning was a crazy idea in its beginning!)
Another predisposition to research in my opinion is the amazing acknowledgement of knowing literally nothing in front of the immense amount of things we don’t know. This is a liberating idea, and a strong brick to effectively incorporating feedback on our work. I think if we don’t have this we will be wasting a lot of time dealing with our own biases, instead of doing the actual work! One way to detect this is to seek underconfidence traits in a person, overconfident people need to read more about this and watch this video from Veritasium about “Why you shouldn’t trust confident people”.
Moreover, Karpathy mentions another criteria to know if your are predisposed to research in his survival guide to a PhD, which is finding a (almost romantic) satisfaction about diving deeply into a topic, “There’s something beautiful about that and if you disagree, it could be a sign that PhD is not for you.”
Lastly, every researcher I cited here agrees to the need of having a high emotional intelligence: you can work 3 months on something and not having any results, you can have low sleeping nights before a paper deadline, you can work so hard then receive feedback that you did not expect from reviewers, you can have so many projects/papers at once, your supervisor could possibly ignore you for more important project he is leading … heavy is the sword that lead to (a partial) truth!
Okkay your a disposed to do research, now how we even do it ???