ICLR 2023 "Tackling Climate Change with Machine Learning" workshop mentorship program

Earlier this year, I participated in the ICLR 2023 “Tackling Climate Change with Machine Learning” workshop mentorship program offered by Climate Change AI as a mentee - and here is my feedback.

Professor James Doss-Gollin was my mentor and we had one session discussing about my draft proposal of an “AI for Climate action” text-based topic modelling and matching tool that I am working on.

Generally the idea of my project is to extract ML and DL methods and their use cases continuously from AI papers, extract specific climate action tasks from countries NDC’s, and connect applicable pairs together in a real-time manner.

For this research project, my scientific supervisor Dmitry Alekseevich is advising me on the technical side, and I got advice from the amazing Saran Selenge on the climate action side.

I had the following set of questions about the proposal, its general idea, and the specific methods and tools being used for it.

  1. What are your expectations from this mentoring process?

  2. What are your initial impressions of this proposal?
    • Does this idea make sense?
    • Does this idea feel worth pursuing / potentially valuable for accelerating or improving efficiency of climate actions?
    • Do you have any suggestions, reading materials, existing tools etc.?
  3. Do this set of data and its collecting criteria make sense to you?

  4. Does this approach (contextualized topic modeling and entity linking) make sense to you?

  5. Extracting topics with human input is taking a lot of time, do you have any feedback on this?

  6. What other questions, feedback, and comments do you have for me?

And I got answers to most of the questions listed, but these following points really got stuck with me.

A) Generally the idea makes sense, but for the initial implementation, the full scope is too big. Starting with only certain domains from the both ends, AI and climate action, might be better.

B) He suggested a tool named Elicit, which is an AI research assistant, and I found this tool to be a good helper when manually reviewing the topics generated by an AI model.

I ended up not submitting my proposal for the workshop this time, but I will definitely go for the next one at this rate!

Now I am reevaluating my approach, refining the data pool and tools used, and starting on the implementation of the project, which is to be the basis of my master’s thesis.

All in all - this short-term mentorship was fruitful, the Climate Change AI community is my new happy place, and I don’t feel alone in trying to use AI for the good of the planet Earth.

If you are also just starting out in this intersection of climate change and artificial intelligence like me, I urge you to check out the CCAI community, seek out help and mentorship, and most importantly, go make that change!