In this post, I describe a mindset that is flawed, and yet helpful for choosing impactful technical AI safety research projects.

The mindset is this: future AI might look very different than AI today, but good ideas are universal. If you want to develop a method that will scale up to powerful future AI systems, your method should also scale down to MNIST. In other words, good ideas omniscale: they work well across all model sizes, domains, and training regimes.

Putting the omniscaling mindset into practice is straightforward. Any time you come across a clever-sounding machine learning idea, ask: "can I apply this to MNIST?" If not, then it's not a good idea. If so, run an experiment to see if it works. If it doesn't, then it's not a good idea. If it does, then it might be a good idea, and you can continue as usual to more realistic experiments or theory.

In this post, I will:

1. Share how MNIST experiments have informed my [...]

---

Outline:

(01:58) Applications to MNIST

(02:42) Gradient routing

(04:43) Distillation robustifies unlearning

(08:39) Subliminal learning

(10:37) Why you should do it on MNIST

(11:30) MNIST is not sufficient (and other tips)

(14:25) The omniscaling assumption is false

(17:09) Code and more ideas

(18:40) Closing thoughts

The original text contained 7 footnotes which were omitted from this narration.

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.