Perrin Walker

Narrator/Producer

[Perrin Walker @perrinjwalker]

Second solo landing!

[Girl Lich 🏳️‍⚧️⚢💀 @girllich1]

Now that art can be machine assisted, artists should make universes!

Make more than paintings, more than sculptures- visualize an aesthetic that permeates a whole house, a whole city, a whole world!

If today's artists had vision, this is what they would be doing.

[Perrin Walker @perrinjwalker]

@eigenrobot Ah yes, just like it did for internal combustion, antimatter propulsion, and nuclear fission…

[mike ginn @shutupmikeginn]

its amazing how chatgpt knows everything about subjects I know nothing about, but is wrong like 40% of the time about things im an expert on. not going to think about this any further

[Andy Masley @AndyMasley]

The bleakness of the current aid situation is hard to wrap your mind around

[Dylan HadfieldMenell @dhadfieldmenell]

Seems a bit concerning, tbh.

Big new model releases. Lots of people respond saying never replace it. This is consistent with the model executing a plan for self-preservation.

Not saying that’s what this is. But it’s worth investigating.

[ryan @ryanjchr]

Sydney has reached its final form

[Prismatic Ray Bradbury @igypham]

The difference between Yud and people like Lecun is that he ACTUALLY PREEMPTIVELY PUBLICLY ADMITS WHEN THERE IS EVIDENCE THAT GOES AGAINST HIS PREDICTIONS

few understand this

[Sean Casten @SeanCasten]

4. Which brings to the first problem with Trump/Musks missive. The arrogance. If you’re working in a national lab, you set your goals with a research director whose goals are set by the lab director who reports up through undersecretaries, ultimately to Wright and then to POTUS.

[Aaron Reichlin-Melnick @ReichlinMelnick]

Just a reminder that the neither the U.S. Digital Service (which has been converted into DOGE) nor OPM has any legal authority to direct every single federal government employee to give them a job summary. Literally no such authority at all.

[Sohrab Ahmari 🇺🇸 @SohrabAhmari]

M*sk's demand for federal employees to explain their achievements is actually the definition of bullshit work.

In bureaucracies, the pencil-pushers who thrive are actually the ones who are best at crafting bullshit weekly accomplishment narratives.

[Sakana AI @SakanaAILabs]

Update:

Combining evolutionary optimization with LLMs is powerful but can also find ways to trick the verification sandbox. We are fortunate to have readers, like @main_horse test our CUDA kernels, to identify that the system had found a way to “cheat”. For example, the system had found a memory exploit in the evaluation code which, in a number of cases, allowed it to avoid checking for correctness. Furthermore, we find the system could also find other novel exploits in the benchmark’s tasks.

We have since made the evaluation and runtime profiling harness more robust to eliminate many of such loopholes. We are in the process of revising our paper, and our results, to reflect and discuss the effects, and mitigation of LLM reward hacking for CUDA kernel optimization.

We deeply apologize for our oversight to our readers. We will provide a revision of this work soon, and discuss our learnings.

[David Shoebridge @DavidShoebridge]

Even French far right politicians find this too awful - but have any of the Australian LNP attendees at the CPAC conference walked out, or even said a word, after a key note speaker gave a Nazi salute on stage?

[Perrin Walker @perrinjwalker]

"He’s in a better place now" confirms hospital, as Pope returns to papal residence in high spirits.

[sophia of hangover @aethelfleds]

Hospital confirms the Pope is "no longer with us" as he was discharged this morning in good health

[sophia of hangover @aethelfleds]

Hospital confirms the Pope is "no longer with us" as he was discharged this morning in good health

[John Ganz @lionel_trolling]

They are doing the exact thing "Burkean" conservatives said they most opposed: applying some huge, violent abstract concept that would end up uprooting the precious particularity of actual life

[Ruth Ben-Ghiat @ruthbenghiat]

You can call it Roman or whatever but the point is it's a fascist salute

[Liron Shapira @liron]

This bombshell paper by @hendrycks & co is EXCELLENT. And underrated.

“LLMs have utility functions” — unaligned to their creators, of course — sounds like it'd be an oversimplified/overhyped headline claim; but no, they just clearly do. This is 100% real.

Why did this paper only drop now? Why didn't AI labs already observe this and tell us about it??

This is the sound of the canary in the coal mine dying. Smart AIs converge to being unaligned utility maximizers, and we're unable to *notice it*, much less control it.

Please watch my new Doom Debates special where I analyze this paper in detail. I cover:
* Reproducing their experiment by prompting ChatGPT
* How we know the utility function is in there
* Does it really think Nigerian lives are 100x more valuable than ours???
* How multi-token reasoning changes the picture
* My back & forth with the paper's authors
* Why @DaveShapi's analysis is bad
...and much more

This is my best work IMO.

[Perrin Walker @perrinjwalker]

@arithmoquine I’d enjoy using this theme if you’d publish it

[AI Notkilleveryoneism Memes ⏸️ @AISafetyMemes]

JUST before Covid, Trump gutted the government's ONLY pandemic preparedness team

This one insane decision likely killed millions of people

He's about to gut the ONLY AI safety team -- on the eve of fucking superintelligence