Lots of interesting topics intersecting in this essay.
One thing that stands out to me is how much we get in our own way by setting up institutions to add external structure to science – the journals are one great example of this, so are government agencies like the NIH and FDA. At the outset of when these institutions are founded the good tends to outweigh the bad, but over time, like any organization of humans, they get bloated and inefficient, and start to make bad decisions. But it also becomes harder to get rid of them because things depend on them, people are used to the status quo, and they have political power.
On the more optimistic side, it's clear that the huge advances in software and especially these large ML models will open up new doors for researchers. Coupled with automation (I've been wanting to get an Opentrons for a while...) it's possible to design experiments in a way that just wasn't feasible before.
One of my favorite applications of this is directed evolution – several groups have developed closed-loop, continuous systems that improve enzyme activity by automating the process of selecting the best mutants, eg: https://pubmed.ncbi.nlm.nih.gov/32374988/.
Hey Shasta, I really like this comment. Thanks for the thoughtful reply. Institutions of science is a topic that I think about a lot, especially when I was at New Science editing pieces on the NIH!
De-convoluting incentives, so that researchers rely less on looming agencies for validation of their work, should be a 21st century priority in science.
Inspiring essay Niko! Awesome read to start the week.
Wonderful article!
Lots of interesting topics intersecting in this essay.
One thing that stands out to me is how much we get in our own way by setting up institutions to add external structure to science – the journals are one great example of this, so are government agencies like the NIH and FDA. At the outset of when these institutions are founded the good tends to outweigh the bad, but over time, like any organization of humans, they get bloated and inefficient, and start to make bad decisions. But it also becomes harder to get rid of them because things depend on them, people are used to the status quo, and they have political power.
On the more optimistic side, it's clear that the huge advances in software and especially these large ML models will open up new doors for researchers. Coupled with automation (I've been wanting to get an Opentrons for a while...) it's possible to design experiments in a way that just wasn't feasible before.
One of my favorite applications of this is directed evolution – several groups have developed closed-loop, continuous systems that improve enzyme activity by automating the process of selecting the best mutants, eg: https://pubmed.ncbi.nlm.nih.gov/32374988/.
Hey Shasta, I really like this comment. Thanks for the thoughtful reply. Institutions of science is a topic that I think about a lot, especially when I was at New Science editing pieces on the NIH!
https://newscience.org/nih/
De-convoluting incentives, so that researchers rely less on looming agencies for validation of their work, should be a 21st century priority in science.
Really nice article, thank you
Let’s share BioData at www.getarius.com :)