If you're interested in helping to run the ARENA program, note that we're currently hiring for an Operations Lead! For more details, and to apply, see here. Summary The purpose of this report is to ev
This piece is crossposted on my blog . Good news! This post is shorter than it looks. There’s a bunch of supplemental info in endnotes if you want the extended edition director’s cut. But it’s much sh
Like apparently every other startup in San Francisco, Theorem works in an open plan office. Not by choice. It was long ago that the barbarian hordes swept across Silicon Valley with battle cries of “less square footage per head!” and “increased collaboration!”, demolishing every interior wall in…
Once upon a time (true story), I was on my way to a hotel in a new city. I knew the hotel was many miles down this long, branchless road. So I drove for a long while. After a while, I began to worry I
People who have received and are considering an offer from an AI lab are in a uniquely good spot to influence the actions of that lab. People who care about AI safety and alignment often have things t
Since a lot of interest in AI alignment has started to build, I’m getting a lot more emails of the form “Hey, how can I get into this hot new field?”. This is great. In the past I was getting so few m
Posted on Jan 9, 2021 This article is a rough note. Writing rough notes allows me share more content, since polishing takes lots of time. While I hope it's useful, it's likely lower quality and less c
This is actually a revised copy of something i wrote on March 7th 2021, before i started my PhD. It had two purposes at the time: 1) it clarified my ideas about pursuing a PhD and 2) acted as a manife
Among my friends interested in rationality, effective altruism, and existential risk reduction, I often hear: “If you want to have a real positive impact on the world, grad school is a waste of time.
Someone recently asked me if I had any updates on my 2014 post on job hunting. I haven’t done a systematic job search since writing that post (I decided to join Wave without considering any other offe
Consider reading How to pursue a career in technical AI alignment. It covers more topics and has more details, and I endorse most if not all of the advice. To quote Andrew Critch: I get a lot of email
February 2009 I finally realized today why politics and religion yield such uniquely useless discussions. As a rule, any mention of religion on an online forum degenerates into a religious argument. W
I've been leaning towards a career in academia for >3 years, and recently got a tenure track role at Cambridge. This post sketches out my reasoning for preferring academia over industry. Thoughts on I
[Editor’s note: At nearly 7,000 words, you probably don’t want to try reading this on an iDevice. Bookmark it and come back later.] Imagine something a wee bit outside your comfort zone. Nothing scand
We recently talked about a mystery: All large language models (LLMs) are terrible at chess. All, that is, except for gpt-3.5-turbo-instruct, which for some reason can play at an advanced amateur level
The gibbous moon riding higher in the cloudless sky, the stars and wash of the Milky Way visible in all their majesty within the darkness: All these illuminated thirty-seven skull masks gleaming above
Even as Harry had raised the gun, he'd known he was making a mistake, his forebrain saw it and tried to stop his hand, but somehow the sick certainty didn't propagate fast enough to prevent his finger
The Dark Lord was laughing. From the empty air came the voice of the Defense Professor laughing wildly, so high and terrible his laughter; it was Voldemort's laughter now, the Dark Lord's laughter bey
The grimness on Albus Dumbledore's face lasted only an instant before giving way to bewilderment. "Quirinus? What -" And then there was a pause. "Well," said Albus Dumbledore. "I do feel stupid." "I s
Even the greatest artifact can be defeated by a counter-artifact that is lesser, but specialized. That was what the Defense Professor had told Harry, after dropping the True Cloak of Invisibility to p
People say "I think" a lot. Here are some examples: I think you brought me the wrong order.I think the numbers in the report are wrong.I think you need to turn left at the light.I think we need to rep
The Defense Professor had set up a cauldron, floating it into place with a wave of his wand, another wave starting a fire beneath it. A brief circling of the Defense Professor's finger had set in moti
Can we keep powerful AI under control, using AI? The Paper: https://arxiv.org/abs/2312.06942 AI Lab Watch: https://ailabwatch.org/ Thanks to my wonderful patrons: https://www.patreon.com/robertskmiles
Gwern's blog: https://gwern.net/ Gwern is a pseudonymous researcher and writer. After the episode, I convinced Gwern to create a donation page where people can help sustain what he's up to. Please go
This guide is patterned after my “Doing well in your courses”, a post I wrote a long time ago on some of the tips/tricks I’ve developed during my undergrad. I’ve received nice comments about that guid
The spiraling leaves of the gigantic dieffenbachia felt like forest loam beneath Harry's shoes, not as unyielding as concrete, but supporting his weight. Harry kept a wary eye on the tendrils, but the
After a single step into Dumbledore's forbidden chamber, Harry shrieked and jumped back and collided with Professor Snape, sending the two of them down in a heap. Professor Snape picked himself up and
Tom Riddle. The words seemed to echo inside Harry's head, sparking resonances that as quickly died away, broken patterns trying to complete themselves and failing. Tom Riddle is a Tom Riddle was the R
June 13th, 1992. It was the last week of school in Hogwarts, and Professor Quirrell was still alive, barely. The Defense Professor himself would be in a healer's bed, this day, as he'd been for almost
Have you ever noticed that reality has some properties that are quite annoying? For example, have you noticed that some people do bad things? And yet those same people sometimes have interesting ideas
A year ago, there was a lot of talk about large language models (LLMs) playing chess. Word was that if you trained a big enough model on enough text, then you could send it a partially played game, as
The stereotyped image of AI catastrophe is a powerful, malicious AI system that takes its creators by surprise and quickly achieves a decisive advantage over the rest of humanity. I think this is prob
College, 2012—Internship recruiting season. “What are you looking for in your internship?” the recruiter asks. “I’d like to solve hard technical problems,” I reply. I end up at Jane Street writing sof
June 4th, 1992. Daphne Greengrass was in the Slytherin common room, writing a letter to her Lady Mother (who was surprisingly intransigent about power-sharing, despite not even being in Hogwarts to ex
June 3rd, 1992. Professor Quirrell was very sick. He'd seemed better for a while, after drinking his unicorn's blood in May, but the air of intense power which had surrounded him afterward hadn't last
Harry stood, panting, in the midst of a brief wasted circle amid the forest, more destruction than a first-year should have been able to reach, by himself. The Severing Charm wouldn't bring down a tre
May 13th, 1992. Argus Filch's face appeared twisted in the light of the oil lamp he held, shadows dancing over his face. Behind them the doors of Hogwarts quickly receded, and the dark grounds moved c
[EDIT: Many people who read this post were very confused about some things, which I later explained in What’s General-Purpose Search, And Why Might We Expect To See It In Trained ML Systems? You might
Opening positions ryan_greenblatt I'm somewhat skeptical about mech interp (bottom-up or substantial reverse engineering style interp): Current work seems very far from being useful (it isn't currentl
bilal