Voozh

Leon Lang

6,150 posts

Leon Lang

@Lang__Leon

Head of curriculum development at Iliad. Previously PhD in AI safety and multivariate information theory @AmlabUva. Trying to prevent x-risks from AI.

London

langleon.github.io

Joined September 2013

Pinned
👁 user avatar
Leon Lang
@Lang__Leon
May 12
We just released the full course materials of the Iliad Intensive — a month-long, full-time AI alignment course for mathematicians, physicists, and theoretical computer scientists. ~20 contributors, 19 modules, at a depth that doesn't exist elsewhere for most of these topics. 🧵
👁 Image
👁 user avatar
Leon Lang
@Lang__Leon
Nov 28, 2024
From a Gwern comment.
👁 Image
👁 user avatar
Leon Lang
@Lang__Leon
Sep 5, 2025
Replying to @MatznerJon
That’s a bad explanation. They could do math in the past, too. So the counterfactual explanation for them renting more can’t be their math ability.
👁 user avatar
Leon Lang
@Lang__Leon
Aug 18, 2025
Replying to @koltregaskes
I got this wonderful response once:
👁 Image
👁 user avatar
Leon Lang
@Lang__Leon
Jul 13, 2025
Reward Learning is just supervised learning, and so should be equally safe, right? Wrong! Our paper “The Perils of Optimizing Learned Reward Functions: Low Training Error Does Not Guarantee Low Regret” shows that policy optimization causes issues. It was accepted to ICML! 🧵
👁 Image
👁 user avatar
Leon Lang
@Lang__Leon
Aug 18, 2025
Replying to @dundereloise
On the one hand, this does surprise me. On the other hand: salting water seems easier than salting the spaghetti afterward since it makes it easier to distribute the salt evenly. So I will continue to salt the water.
👁 user avatar
Leon Lang
@Lang__Leon
Jun 21, 2025
With 10 seconds per short, and assuming zero counterfactual value, and a human lifespan of 3 billion seconds, this means 666 evaporated lifetimes each day.
👁 user avatar
YouTube Creator Insider
👁 YouTube Creators
@YouTubeInsider
Jun 20, 2025
Now averaging over 200 billion Shorts daily views!
👁 Image
👁 user avatar
Leon Lang
@Lang__Leon
Oct 19, 2024
Shane Legg, cofounder of DeepMind, in 2009. SIAI was the Singularity Institute for Artificial Intelligence, today called MIRI -- the Machine Intelligence Research Institute.
👁 Image
👁 user avatar
Leon Lang
@Lang__Leon
Jul 26, 2023
Replying to @coryfromphilly and @thechosenberg
Emotions are beyond reason. I find your take low in compassion/empathy.
👁 user avatar
Leon Lang
@Lang__Leon
Dec 14, 2024
Replying to @NeurIPSConf
What was the cultural generalization?
👁 user avatar
Leon Lang
@Lang__Leon
Sep 19, 2024
Replying to @yacineMTB
At the risk of being pedantic by providing one counterexample: no non-mathematician can within 1-2 weeks understand what a perfectoid space is or why it is important.
👁 user avatar
Leon Lang
@Lang__Leon
Nov 22, 2024
Replying to @TheZvi
I never flew anything except economy. I think you have seen an unusual sample.
👁 user avatar
Leon Lang
@Lang__Leon
Nov 10, 2024
Replying to @Aella_Girl
Without any net worth that’s an easy question :P
👁 user avatar
Leon Lang
@Lang__Leon
Aug 19, 2024
Replying to @robertskmiles
Future version: "LLMs aren't geniuses, they just memorized some thinking patterns from reading Terrence Tao's papers and now apply those patterns to solve the biggest mysteries of our time."

URL: https://x.com/Lang__Leon

⇱ Leon Lang (@Lang__Leon) / X