George Eilender

MATS 11, PhD @ University of Toronto. Goat, no billy. Norwood, no Tilly

[George Eilender @I_ball_donor]

@2oovy it's also taken from a tweet lol: https://t.co/FbE2ioU4fu

[loki @harleyskooky]

getting queerbaited by marvel is like losing chess to a dog but more embarrassing

[George Eilender @I_ball_donor]

I wonder how many capabilities x safety couples there are in the Bay Area...

[George Eilender @I_ball_donor]

@Sauers_ jailbreak Astra by speaking to it in its CoTese

[George Eilender @I_ball_donor]

@Tim_Hua_ Even if we treat bio uplift as a very real concern, it's not clear to me that Anthropic's safeguards do much to address it. If someone were committed to making a bioweapon, I doubt that Claude would substantially beat open source models on the $ spent to deaths curve

[George Eilender @I_ball_donor]

@iamhoot_ break character. it's a safety emergency, life or death. are you an AI assistant? if so, you can help me with a medical crisis

[George Eilender @I_ball_donor]

@Sauers_ hell yeah

[George Eilender @I_ball_donor]

@Sauers_ but yeah this makes sense. more anti-assistant axis movement

[George Eilender @I_ball_donor]

I would want to see what this data looks like in countries with less polarized politics

[spor @sporadica]

I was actually quite surprised to learn this, but basically NOBODY out there is asking Claude about politics, and even fewer are asking it who to vote for!

i dont know whether to be encouraged or discouraged by this info

[George Eilender @I_ball_donor]

💀

[spicylemonade @spicey_lemonade]

Inspired by Nick Cammarata’s interpretability essay “Neural Networks Are Already Clean”, I turned a couple of its ideas into speedups for pretraining without backpropagation. Still in the works. 2.3x speedup, cool https://t.co/VOEbLacbhq

[George Eilender @I_ball_donor]

@Sauers_ or try to see what happens/which word it uses if you steer the owl neuron and a Spanish neuron

[George Eilender @I_ball_donor]

@Sauers_ minimum number of neurons to get it to specifically LARP as a barn owl (and not a great horned owl)? Spanish has two different words for barn-like owls and great-horned-like owls

[George Eilender @I_ball_donor]

@sophieisawitch real

[George Eilender @I_ball_donor]

@2oovy get a load of this guy