Aligned to Flourish
ThemeSystem
Link iconsFavicons on external links
Reading barHow far through a post you are

Accelerating AI Alignment Research (Talk)

The keynote I gave to open the research augmentation hackathon I co-organised with Apart Research. Slides and project suggestions are linked underneath.

Published
Status
notes
Markdown
Download
Topics
Automating alignment researchTalks
Listen, 1 min

Read the transcriptDownload the audio

I gave a keynote talk on how we should be thinking about accelerating AI alignment (safety) research. This is a topic I’ve been thinking about for a while and have given similar talks in the past, but I gave this one to kick off the hackathon I’ve co-organized with Apart Research this weekend.

If you’d like to learn more about how I’ve been thinking about this, here’s the talk:

If you’d like to collaborate on this kind of work, let me know!

Here are the slides for the talk. Here’s some related notes and project suggestions.

Discuss this post below

Reply

Replies are open to anyone on the newsletter. If you get the emails, you can reply here. If you do not, you can subscribe.

Read next

  • Automated alignment research needs a better plan than 'Stop if we catch them scheming'

    My PIBBSS Symposium talk from September 2025, with the deck and the full transcript, on why the dual-use worry is overblown.

    also automating alignment research

  • Fooled by Vast Knowledge

    Models look like they generalise because the training data is vast enough to hide the difference between interpolation and extrapolation. I think that confusion is leading safety research astray.

    also automating alignment research

  • Gaining clarity on Automated Alignment Research

    Four different things get conflated in the phrase automated alignment research, and the crux is whether applying techniques we already have is a short step from inventing a new paradigm.

    also automating alignment research

  • Automating AI Safety: What we can do today

    A concrete list of things to build now so a coding agent can actually run an interpretability experiment, and why the METR slowdown result does not say what people think.

    also automating alignment research

Cite this post

Permalink

https://jacquesthibodeau.com/accelerating-ai-alignment-research-talk/

BibTeX
@misc{thibodeau2024accelerating,
  title = {Accelerating {AI} Alignment Research (Talk)},
  author = {Jacques Thibodeau},
  year = {2024},
  month = jul,
  howpublished = {Aligned to Flourish},
  url = {https://jacquesthibodeau.com/accelerating-ai-alignment-research-talk/},
}