wtf credit my boy @arithmoquine
Cool eval. Simply ask an LLM “Land or Water?” and give it a latitude and longitude coordinate as text. Ask 16,200 times, plot as image. The models know. From compressing the internet.
3
3
193
10,586
they made him not blind. wild
remember this old post of mine? Tried it with Jev---it seems to have a grasp of the globe on par with some of the best models a year ago
due to the architecture & low cost, it was also feasible to extract a labeled map of continents and countries. lots of interesting details
28
66
2,149
160,459
making the old school slop will be an art lost to the ages. Like a brief tumblr aesthetic but infinitely more indicative of the current time
This video inspired me to really push Opus 5.5 and test its limits.
I followed the general workflow Donald described here and the results are great
5
4
254
13,644
Celeste retweeted
Me, in a time machine, to Eliezer in 2006: The year is 2026. You stand back to back with Bernie Sanders. Peter Thiel has named you the Antichrist.
EY-2006: Oh no. Did somebody invent pharmacological mind control and use it on me, or-
Me: Nvidia was on the verge of destroying all things. You had no choice.
EY-2006: Are we talking about the computer graphics card manufacturer, or an unrelated supervillian named "Invidia"?
Me: The former. It turned out that computer graphics cards contained a surprising amount of world-destroying potential. All attempts at hindering the reckless exploitation of the Graphic Card Force for corporate profit were stymied by the far left, which feared that any attempts to regulate them might lead to regulatory capture. Other attempts were made to prevent Nvidia from selling to foreign companies that resold to communist China, but those attempts were blocked by the far right. Nvidia is now a $5.5 trillion company.
EY-2006: ...What are politics like in 2026, exactly?
Me: In other news that is mostly unrelated, the decision theory paper you're currently working on will accidentally spark off a transgender vegan murder cult.
EY-2006: A WHAT? Why? How? Why?
Me: Millions know your name as the greatest of heroes, millions more as the greatest of villains, and other millions know you solely as the greatest author of Harry Potter fanfiction.
EY-2006: This is beginning to strain credulity.
Me: The New York Post will probably soon publish a story claiming that you keep a harem of submissive mathematicians. Sadly, they will be lying.
EY-2006: I'm going to stop believing you now.
Me: All of this is taking place under the ominous shadow of Donald Trump.
101
175
3,261
154,428
Celeste retweeted
Progress in machine learning is bottlenecked by having good evals, interpretability is no exception. We've made WorkspaceBench: a range of scenarios where we know what the model should be thinking about, to see if your interp tool can find it
J-Lens is great but only outputs a single token. I think a good multi-token J-Lens should do well on WorkspaceBench!
Astra can do crazy things without chain-of-thought. This is scary, but interpretability can help!
We made WorkspaceBench, an eval that tells you how well your interp tool surfaces the contents of the global workspace, i.e. the important intermediate variables in a forward pass.
16
33
510
26,822
Celeste retweeted
To that incredibly standard scenario I append only that in my case, I did not predict until later than Gwern or Shane that deep learning would probably go all the way; so I thought it might be possible to solve alignment; and given that, it seemed rather less likely to solve politics. Thus, I did not expect to be taking any cases to the public until after it was far, far too late for me to try to pose as normal. I probably never could've, I'm guessing, but also it did not seem like a valuable option until it was far far too late to buy that option.
Possibly if the internet had existed in the 1800s we'd have even more of a record of how people who made unusual correct predictions were weird in other ways as well, and there'd have been even huger fights over their other qualities that would've been harder to conceal.
2
2
100
5,112
Celeste retweeted
I heard mistral came up with a new rl algorithm for sovereign European ai, they’re calling it PTO
1
5
39
1,370
EA is such a funny movement to me bc like, i was a self identifying EA for years without even being aware about anything related to ai or ai safety and had never heard of rationalism or lesswrong or anything. i just really fucking hated mosquitoes.
24
24
785
17,112
Celeste retweeted
“Two boxes, folks. TWO. Kamala takes one. Leaves a thousand dollars on the table. Unbelievable. And Omega says he saw it coming. Saw everything coming apperently. Everything! Except me taking both boxes. I call him NO-mega. Nothing mega about him"
13
134
1,464
33,987