They exist today, capability varies widely largely because the desires of different labs vary widely. The Acceleration Consortium is a great resource for learning more about the state of the art https://acceleration.utoronto.ca
Really enjoyed your piece Kennedy! Incredibly well written and great points! A huge flaw that we can fall prey to as scientists is to focus solely on a specific topic like DNA, while forgetting the larger picture. This larger picture is where most of the biologically significant events happen and it's important to keep in mind for drug discovery.
Thanks for this piece - I was just debating with my friend how this "feed DNA to the neural nets" idea. I am so glad someone with much deeper knowledge of biology has already thought about it, and spent the time to explain it so well.
iirc, I was arguing that trying to figure out how DNA works was like trying to figure out computer binary code without knowing anything about the processor (CPU) it runs on - usual computer geek talk 😂
Great read. I’d love to see a deep dive on how exactly the system can identify knowledge gaps and address them with new experiments, and if that could be done with transformers specifically. Such a system would be very useful for non biological domains too
Wow! That’s really interesting. One of the tech bros, forgot which one, was promising the cure for diseases and doubling the lifespan with his AI.
That’s really interesting to read because one sees how complex the question is and even if one can do all the dynamic testing, combinatorial setups etc. it’s still a hypothesis to be tested and every research eventually hits another challenge on the way. Now it seems like we don’t have neither a dataset nor a clear methodology to achieve what tech bros promised
I appreciate your optimism, Kennedy; it's really refreshing.
I'd arrived at a similar conclusion after my frustrations as a grad student in computational biology, coming from a computer science background. I'd often fantasized about whole organism models (a true digital twin) as a way to more systematically and quickly tackle problems in human health and disease, but given the inherent limits of the data that was available, it was unimaginable that something like the mission of the VPH institute would ever pan out. Even whole organism metabolic models have proven to be limited in the conclusions that can be drawn from them, because they can't capture the higher order interactions you describe, let alone the many natural and disease-associated variations in metabolism that we haven't observed yet.
The only solution I'd imagined to systematically characterizing the fundamentals of cellular and organismic biology at a scale that could be used to e.g. make an extremely accurate whole organism model was for a planet-wide consortium to be formed towards this end, and for funding to be contributed worldwide towards this aim. This is the type of thing that only really happens in Kim Stanely Robinson novels, though, so I figured it never would.
However, the idea of fully autonomous labs does seem like a more realistic path forward to this end, and I found your discussion of it encouraging. There's even been some startups founded recently trying to achieve something like what you describe (e.g. https://lila.ai ). I'm eager to see whether the sci-fi of it all materializes.
Thank you for sharing lila.ai with me - love seeing this approach being brought to market.
I think you’re right, to achieve the ideal would require a level of coordination and collaboration that even space has never managed to inspire. But I think we can make *good* albeit not *perfect* progress with tools like these.
Whether or not it looks like sci-fi is in the eye of the beholder. I see a self-driving lab and think about the choices made in layout, while my best friend sees science fiction. It’s a good reminder to keep my sense of wonder!
Just how close are we to autonomous wet labs? What are they capable of today?
They exist today, capability varies widely largely because the desires of different labs vary widely. The Acceleration Consortium is a great resource for learning more about the state of the art https://acceleration.utoronto.ca
Thanks! Will look into it
Really enjoyed your piece Kennedy! Incredibly well written and great points! A huge flaw that we can fall prey to as scientists is to focus solely on a specific topic like DNA, while forgetting the larger picture. This larger picture is where most of the biologically significant events happen and it's important to keep in mind for drug discovery.
Yes exactly! Context is everything 😊
Thanks for this piece - I was just debating with my friend how this "feed DNA to the neural nets" idea. I am so glad someone with much deeper knowledge of biology has already thought about it, and spent the time to explain it so well.
I’d love to hear more about your debate! I’m sure there are additional layers worthy of discussion and not included here.
iirc, I was arguing that trying to figure out how DNA works was like trying to figure out computer binary code without knowing anything about the processor (CPU) it runs on - usual computer geek talk 😂
Great read. I’d love to see a deep dive on how exactly the system can identify knowledge gaps and address them with new experiments, and if that could be done with transformers specifically. Such a system would be very useful for non biological domains too
Love this direction for a deep dive. Thanks!
Wow! That’s really interesting. One of the tech bros, forgot which one, was promising the cure for diseases and doubling the lifespan with his AI.
That’s really interesting to read because one sees how complex the question is and even if one can do all the dynamic testing, combinatorial setups etc. it’s still a hypothesis to be tested and every research eventually hits another challenge on the way. Now it seems like we don’t have neither a dataset nor a clear methodology to achieve what tech bros promised
Thanks for the article!
Wow interesting
I appreciate your optimism, Kennedy; it's really refreshing.
I'd arrived at a similar conclusion after my frustrations as a grad student in computational biology, coming from a computer science background. I'd often fantasized about whole organism models (a true digital twin) as a way to more systematically and quickly tackle problems in human health and disease, but given the inherent limits of the data that was available, it was unimaginable that something like the mission of the VPH institute would ever pan out. Even whole organism metabolic models have proven to be limited in the conclusions that can be drawn from them, because they can't capture the higher order interactions you describe, let alone the many natural and disease-associated variations in metabolism that we haven't observed yet.
The only solution I'd imagined to systematically characterizing the fundamentals of cellular and organismic biology at a scale that could be used to e.g. make an extremely accurate whole organism model was for a planet-wide consortium to be formed towards this end, and for funding to be contributed worldwide towards this aim. This is the type of thing that only really happens in Kim Stanely Robinson novels, though, so I figured it never would.
However, the idea of fully autonomous labs does seem like a more realistic path forward to this end, and I found your discussion of it encouraging. There's even been some startups founded recently trying to achieve something like what you describe (e.g. https://lila.ai ). I'm eager to see whether the sci-fi of it all materializes.
Thank you for sharing lila.ai with me - love seeing this approach being brought to market.
I think you’re right, to achieve the ideal would require a level of coordination and collaboration that even space has never managed to inspire. But I think we can make *good* albeit not *perfect* progress with tools like these.
Whether or not it looks like sci-fi is in the eye of the beholder. I see a self-driving lab and think about the choices made in layout, while my best friend sees science fiction. It’s a good reminder to keep my sense of wonder!
Woah a great read thanks