Welcome to science! Everything is so complicated that either you devote your life to your scientific specialty and the necessary math or you devote your life to understanding code. Few can squeeze two lives out of one, so something must give.
-
-
throwing the compute power at the problem of simulating and modeling the world system: going ok. not close to investment in the AI domain, but not terrible. throwing the equivalent IQ of programming at the problem: not close. programming scientist driven, needs engineering love.
-
There is so much great high quality open source software out there. Yet, AFAIK, very little open source science projects. Surely plenty of engineers would be willing to put time into a climate change model? Where's the gap? Why isn't this a thing yet?
- 4 more replies
New conversation -
-
-
we actually really really need next gen software toolchains for science, especially these big simulations, with GPU compute as the background abstraction, but also adaptive support for high precision where necessary. this situation supports using toy models for decisions; simple
-
You are absolutely correct. The bad news is that we programmers are seemingly unable even to supply ourselves with a next-gen software toolchain. Until we can meet our own needs, which we ought to understand the best, it is unlikely we can meet anyone else’s.
End of conversation
New conversation -
-
-
I also think university labs struggle to recruit hire and retain high performing engineers, since they can make more money working on ads at Google or whatever.
@broadinstitute is an example of an institute that actually built an engineering culture -
Whenever I think about Larry and Sergey diverting brilliant PhDs into a life of load-balancing ad servers and bathing in caviar, my mind goes to this passage.pic.twitter.com/uDFbj6D9am
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
we actually need probably >>100x the computational resources and 10x the programming resources code-complexity on this, when we are looking to models to predict, in detail, the specific trajectories and outcomes we have to mitigate, vs just sending out "warning flares"