Fermat's Library, Open Science, and the Human Backstory of Scientific Discovery

This podcast episode features a conversation with Luiz and Joao Batalha, co-founders of Fermat's Library, a platform designed for annotating academic papers to make them more accessible. The discussion begins by emphasizing the importance of understanding the human backstory behind scientific discoveries, arguing against the romanticized "eureka moment" narrative. Examples like Richard Feynman's QED paper, Ian Goodfellow's GANs paper, and Dijkstra's shortest path algorithm illustrate how personal experiences, random interactions, and even seemingly mundane observations can spark profound breakthroughs. The hosts suggest that including a "backstory section" in papers could provide crucial context, make discoveries more memorable, and reveal the often non-linear, collaborative nature of scientific progress, contrasting it with the solitary grind often imagined.

A significant portion of the conversation critiques the current state of scientific publishing, particularly the prevalence of paywalls that restrict public access to research funded by government or industry. The hosts highlight that journals profit from unpaid peer review labor while gatekeeping knowledge, which they argue goes against the fundamental principle of free dissemination of ideas essential for scientific advancement. They advocate for open science, where all articles are freely accessible, suggesting that costs could be covered through donations, advertisements, or public funding. The discussion also touches upon the technical barriers to collaboration, such as the divide between communities using LaTeX for typesetting and those using word processors like Microsoft Word, hindering interdisciplinary work.

The episode delves into the role of preprints and platforms like arXiv and BioArchive in accelerating scientific communication, especially in fast-evolving fields like machine learning. They note the incredibly long publication times for traditional journals, making preprints vital for sharing new work quickly. However, they also identify challenges within preprint systems, such as the need for mechanisms to distinguish good work from bad without formal peer review. The historical context of the 1960s Information Exchange Group in biology, which was shut down due to pressure from journals, serves as a cautionary tale about the resistance to open scientific sharing and its long-term impact on fields.

Ultimately, the Batalha brothers and Lex Fridman envision a revolutionized scientific ecosystem. They propose a system where papers are not static, restricted documents but rather living, collaborative entities that can be continuously updated and annotated. Fermat's Library, with its journal club, annotation tools (Margins), and browser extension (Librarian) for arXiv, aims to foster this open, collaborative environment. The broader goal is to move science in the "right direction" by democratizing access to knowledge, enhancing clarity, and facilitating collective problem-solving, thereby freeing the "potential energy" trapped within dense, restricted academic papers.

I believe that all scientific articles should be freely accessible to the public.

the papers are you know chapters of a way more complex uh story

it wasn't necessarily the grind of like andrew wiles of the females last theorem for example it was more like a a brilliant moment of insight

as fields mature like those types of moments i think naturally uh are less frequent

knowing that backstory helps sediment that algorithm in your mind so that you don't forget about it as easily

the cool thing about alchemy... it almost like to me symbolizes dreaming of the impossible

he always responded saying i'm not here to answer your questions i just want to open it more questions

academic papers can be one of the densest forms of content out there and generally pretty hard to understand at times

papers are not optimized for clarity and for a proper explanation of of that content because there are so many restrictions

this is literally the longest upload period on the internet

you really need a way for instance to maybe it's not peer review but distinguish what is good work from bad work on archive