Love this! The Monomer/Elnora hackathon really opened my eyes to how fast this is coming. I do wonder a little what happens when human knowledge and judgement becomes the bottleneck — even with an entire PhD’s worth of knowledge about growing gram negative clinical bacteria, I couldn’t keep pace mechanistically with Elnora’s suggestions re growing a different bug. I suspect we’d have beat 16% faster growth if I’d spent the previous day or two reading around the problem with the help of elicit or notebook LM. But the promise of being able to run 10,000 tests just for the 2 that matter without me personally being at the bench 24-7 for weeks is clearly going to be a game changer!
I ended up writing a whole post pushing back a little after reading this about what cloud labs can and can't do (for now).
Tl;dr: there's stuff that lab automation really struggles with and I'm not sure cloud labs have the answer to some of these things, at least without some big hardware and robotics innovations.
True! And realistically, we're probably years away from that inflection point — the transition will be genuinely hard for a lot of people. But I'm optimistic that AI × robotics in the lab will be a net positive. More science, faster, with fewer of the soul-crushing repetitive parts that burn out good researchers.
Excellent piece, Carmen! I will ask the high school kids from BioCoderDojo Timisoara to take a bit of time during this Easter holiday and give it a read!
I am confident that it will empower them to explore science even beyond what they are doing now: attending the How To Grow (Almost) Anything course as committed listeners (with midnight watching the lectures or joining the MIT recitation, with homeworks, individual and group projects they work on with their international colleagues in the remote lab nodes in Tokyo, Ottawa, London or San Francisco) or by debating the best project idea for their iGEM team project. And to find ways to use Elnora to build that project!
Love this! The Monomer/Elnora hackathon really opened my eyes to how fast this is coming. I do wonder a little what happens when human knowledge and judgement becomes the bottleneck — even with an entire PhD’s worth of knowledge about growing gram negative clinical bacteria, I couldn’t keep pace mechanistically with Elnora’s suggestions re growing a different bug. I suspect we’d have beat 16% faster growth if I’d spent the previous day or two reading around the problem with the help of elicit or notebook LM. But the promise of being able to run 10,000 tests just for the 2 that matter without me personally being at the bench 24-7 for weeks is clearly going to be a game changer!
Nice article!
I ended up writing a whole post pushing back a little after reading this about what cloud labs can and can't do (for now).
Tl;dr: there's stuff that lab automation really struggles with and I'm not sure cloud labs have the answer to some of these things, at least without some big hardware and robotics innovations.
https://partialagonism.substack.com/p/cloud-labs-and-the-floodlight-effect?utm_source=share&utm_medium=android&r=58mdpo
Even when this technology appears the bench scientists will still be winning for a while.
True! And realistically, we're probably years away from that inflection point — the transition will be genuinely hard for a lot of people. But I'm optimistic that AI × robotics in the lab will be a net positive. More science, faster, with fewer of the soul-crushing repetitive parts that burn out good researchers.
“Kill the bench” is uncomfortable, but directionally right.
The bottleneck is no longer tools.
It’s human execution.
Same thing already happened in software.
Now it’s happening everywhere else.
The people who win won’t be the ones doing more manual work.
They’ll be the ones who design better systems around tools like Gemini or NotebookLM.
Thinking > doing.
That transition is already visible if you look closely.
Tried to map this shift into real workflows here: https://shorturl.at/59Jmd
Excellent piece, Carmen! I will ask the high school kids from BioCoderDojo Timisoara to take a bit of time during this Easter holiday and give it a read!
I am confident that it will empower them to explore science even beyond what they are doing now: attending the How To Grow (Almost) Anything course as committed listeners (with midnight watching the lectures or joining the MIT recitation, with homeworks, individual and group projects they work on with their international colleagues in the remote lab nodes in Tokyo, Ottawa, London or San Francisco) or by debating the best project idea for their iGEM team project. And to find ways to use Elnora to build that project!