It seems increasingly clear that parts of the scientific system need to change. At the same time, AI may give younger scientists new ways to navigate it more intelligently.
But I also worry that large-scale use of LLMs could push scientific visibility even further toward prestige and narrative, and away from merit and reality.
I wrote a short essay (on my substack) on reproducibility, prestige, and preclinical research, partly from first-hand experience.
I read your essay; it’s nice! I completely agree. Reminds me of the literature on causes of obesity. So much ink spilled, but no (non-pharmacological) clinical solution. The driver(s) of obesity seem to lie outside of the scope of mainstream methods. It’s an area I’m working on.
It seems increasingly clear that parts of the scientific system need to change. At the same time, AI may give younger scientists new ways to navigate it more intelligently.
But I also worry that large-scale use of LLMs could push scientific visibility even further toward prestige and narrative, and away from merit and reality.
I wrote a short essay (on my substack) on reproducibility, prestige, and preclinical research, partly from first-hand experience.
I read your essay; it’s nice! I completely agree. Reminds me of the literature on causes of obesity. So much ink spilled, but no (non-pharmacological) clinical solution. The driver(s) of obesity seem to lie outside of the scope of mainstream methods. It’s an area I’m working on.
Maybe this will finally force "science" to overhaul the process. I think it's been broken for quite a while, even if I couldn't tell you exactly when.
Just posted one minor 'fix': https://jasonmfletcher.substack.com/p/can-ai-fix-journal-reviewinga-little?r=i2hui&utm_campaign=post&utm_medium=web
Thanks! Nice post.