Related: I am ok with people using LLMs to write grants and propose new research. If it’s good stuff, it’s good stuff. Think it’s gonna be a while before an llm can propose competitive projects tho. As the tech bros would say, they lack ‘taste’

the issue I take with it is honestly that these models aren’t there, there’s no sign they can *get* there, but they’ll still tank every power grid and pollute every waterway and air system they can get their greedy little hands on in the process of pretending it’s possible

it's so bleak. their idea of success is also very dark. feels so naive to be like, wait, can we try to do good things instead?

Yyyyep

Hi guys, as one of the people mentioned in the article, I'm happy to clarify any questions about the workflow. There is constant human verification along the process of finding research and researchers. I would not trust a fully automated LLM system and that is mentioned at the end of the article

hey César, do you have a manuscript detailing your methods, or would you be willing to conduct simulations or real-world experiments to evaluate the methods? is this approach to finding grant recipients more effective than alternatives? more broadly, what is your goal in using LLMs for this task?

i will say, i'm personally not *that* worried about bias from LLMs compared to bias inherent to any granting process. and i like the proactive funding model as a complement to traditional calls. i'm not sure if an LLM could identify the best science/grants/recipients, but..

evaluating funding effectiveness is not easy, and it's totally possible that LLMs could have positive, negative, or neutral effects on the whole process

I agree. This is really tricky, especially when we didn't have an open call. Biases remain even with human evaluation. Contacted researchers were asked to write an application and the final selection came from there. But the LL+ML models helped us pinpoint a selected group of people

The purpose of using LLMs is to scan through loads of papers and label them according to previously human defined labels on a smaller set of relevant papers. Then other ML models are used on top to 'predict' and have an indication of what is promising and what fits our mission.

Hey Jase, I have a high level blog I wrote a couple of months ago if that's useful. We are still working on the validation of this process in the real-world as we just awarded our first cohort. Can't say this is more effective as our remit as our mission is specific to finding climate innovation

To expand on this, the purpose of the group is to find climate innovation in scientific papers. And being proactive at contacting researchers saying 'hey! we think your paper tackles climate change due to X and has potential to be commercialisable'