Sentient Futures

Transformative AI alignment with non-humans

commitment
10 hrs/week
format
Research and writing
topic
Post-AGI transition/Transformative AI · Longtermism and future suffering
open to mentee proposals
Yes, with the right mentee
mentored by
Jasmine Brazilek
Jasmine BrazilekCompassion Aligned Machine Learning (CaML)
Apply by August 9 →

The project

  1. How mid-training compassion data scales (both data scaling and model scaling projects)
  2. Analysis of open-source Huggingface training data (e.g. The Pile) and their content related to animal welfare
  3. Why does SFT erase mid-trained compassion more than RLAIF? Use mech-interp techniques to find out.

The expected output of all these projects is solid research output that can be turned into a paper given enough time and results

The mentee's role

Mentees will lead the project with guidance and design input from me/the CaML team. Mentees will handle the project implementation and potentially paper write up end-to-end. They will need to keep excellent records of experimental parameters used and data files found so the write-up can be completed and verified.

Who I'm looking for

  • Agentic
  • Coding skills
  • Depending on the project statistical skills may be needed or nice to have

Questions for applicants

  1. Why are you excited for this project in particular?
  2. Do you follow any work in the AI alignment community?

Support offered

  • Technical support
  • Domain expertise
  • The outcome of this project is to write or lay the groundwork for a high quality paper or two on the first 2 projects if I get enough mentees to complete these satisfactorily
Jasmine Brazilek

Jasmine Brazilek

Compassion Aligned Machine Learning (CaML)

I am the co-founder and technical lead of Compassion Aligned Machine Learning (CaML), a research group working to make AI systems more compassionate across animal welfare, digital minds, and moral humility.

  • I lead the engineering behind CaML's evaluation suites (ANIMA, MORU, TAC, and the public compassionbench.com leaderboard) and its research efforts, including the generation specs and hyperstition-based data used to associate the AI role with good outcomes.
  • I have also run the Hyperstition For Good competition and hope to run many more practical intervention projects to mobilize the community towards goals of aligning AGI in the future.
  • I have 6 papers published on Arxiv on my work around AI alignment and preventing mass-suffering.