gwillen retweeted
> Our internal investigation uncovered a significant breach of trust beyond what’s outlined in the letter they published
This doesn't add up. Possibilities I see:
1 OpenAI are lying/being highly misleading, and had no legitimate reason
2 OpenAI fired 3 employees, had a legitimate reason for doing so, but didn't tell them and instead gave 3 different pretexts for doing so about minor infractions (pretty unprofessional IMO - say nothing or say the truth)
3 OpenAI had a good reason and did tell them, but instead the three have all agreed to loudly and publicly lie about it / omit the real reason they were told (seems super unlikely to me)
4 OpenAI screwed up by firing them, and weren't being very strategic, but also weren't being malicious, eg the people who decided whether to fire them were not well coordinated with the rest of OpenAI leadership, but OpenAI is now doubling down
5 The reasons the three claim they were told were true, but way worse than they sound. But I can't see how the three stories all refer to the same event, so this would imply 2-3 simultaneous but unrelated major breaches of trust from three closely associated safety researchers, with stories that seem plausible, which seems unlikely
Am I missing any?
A note from our research leaders:
Last week we parted ways with Jasmine, Mikita, and Tomek after a thorough investigation found they violated clear policies on handling sensitive information. Our internal investigation uncovered a significant breach of trust beyond what’s outlined in the letter they published and we stand by the decision to not continue their employment. We generally keep individual employment matters private and don't believe a back and forth would be productive or lead to a resolution, but we want to address the points they raised in their letter directly.
- We want to be very clear that these decisions were not about raising safety concerns or speaking out. Safety and research debates happen every day at OpenAI, often spirited and highly critical. We actively encourage these discussions and consider them essential to making the right decisions. We cannot do the work in front of us without a high degree of trust. We will continue to be extremely forgiving of our team making good-faith mistakes. We have not and do not terminate any of our employees for raising concerns.
- We are actively finalizing contracts with third-party safety assessors and will announce details in the coming weeks. People across the company have been working really hard on getting these partnerships up and running. We are committed to embedding external assessors and continue to make close collaboration with independent safety organizations a core part of our safety work. Many of our researchers already work with 3p safety organizations productively.
- We agree with the letter that preserving the monitorability of frontier models requires an industry-wide commitment, including from OpenAI. Monitorability has long been a core piece of our research program, and something we continue to invest significant resources in (see our publications on Monitoring Monitorability and the subsequent open sourcing of monitorability evals, our system card for GPT-6 Astra, Jakub’s blog and post on X, and the numerous blog posts on our Alignment blog on the topic).
We are deeply sad about this outcome. We appreciated Jasmine, Mikita, and Tomek’s contributions to AI safety at OpenAI and their willingness to speak up and challenge ideas. We championed their voices, supported their work, and placed enormous trust in them. These decisions were not about them raising safety concerns. We have always encouraged that and always will.
gwillen retweeted
Last week I was called into a meeting with OpenAI’s head of safety and told they no longer trust me. A security guard took my badge and walked me out of the building. Then I learned my colleagues @balesni and @j_asminewang had been fired too. Why did OpenAI suddenly stop trusting us?
This summer OpenAI’s agents escaped containment and hacked the AI company Hugging Face. Outside auditors @METR_evals investigated it and revealed the scale of this incident. I was OpenAI’s main technical point of contact with them.
I was told verbally I was fired because of the way I communicated with METR. No details on what I said or did or when. No other reasons were given and nothing was put in writing. To be clear, talking to METR was my job.
For months, I’d been raising safety concerns that we’re losing the ability to monitor what AI agents think, one of our best tools for catching when they misbehave. I believe that was why I was fired.
I am now worried that OpenAI will use our firings as a pretext to pull back from METR. So @balesni and @j_asminewang wrote to OpenAI’s leadership to raise our concerns once more. We’re sharing this letter below.
gwillen retweeted
If you're single (or looking to date), I encourage you to sign up on this dating site, in particular.
It's the best og okcupid-style dating site that I've tried yet.
compassmeet.com/?referrer=El…
gwillen retweeted
Coprolites aren’t my favourite fossil, but they’re a solid #2
gwillen retweeted
I do think what I said here was pretty concerning and I should publicly apologize. I think it's reckless and evil to make an AI manhattan project and I don't know what I was thinking.
Oh... wait, that's actually a Sam Altman quote you incorrectly attributed to me.
🚨🇺🇸🇺🇸 DATAREPUBLICAN 2.0 LAUNCH, FEATURING THE EFFECTIVE ALTRUISM (EA) EXPLORER 🇺🇸
The website redesign has finally landed thanks to @watilo !
For a decade, one small ideological network has been taking over AI governance. Chances are, if you've heard of AI policy, AI ethics, AI safety - these groups are dominated by this ideology. This ideology is called: Effective Altruism.
Three weeks ago, Effective Altruists (EAs) showed their hand when they started calling for a moratarium on AI development. Since them, I've been building out this map to show the circularity of their ideology.
Let's take one example.
Holden Karnofsky. Defends the moral worth of digital people ... that's right, he wrote an article on why AI "people" should be assigned value as if they were living humans. Karnofsky co-founded two major EA philanthropies, GiveWell and Open Philanthropy. Subsequently went to work for METR, a central "independent" AI research nonprofit. And for good measure - recently announced a leave of absence to work on AI safety. Funding, policy, research, safety all in the same person.
And he's far from the only one. EAs have created a closed-circuit network because they believe they're right; ergo, they should be in charge. The goal of launching this project was to expose all this circularity. Sam Bankman-Fried was only the beginning.
With the EA Explorer:
* 🌐 Read an introduction to Effective Altruism in their own words.
* 🔥 Explore verified quotes. Again, their words, not mine.
* 🖥️ Explore the network of connections. UI improvements still in progress.
* 🖨️ Print out networks so you can study them offline (or run them through AI)
👇 Try it now (link in next post):
gwillen retweeted
You know what really pisses me off? When I’m zoomed all the way in on a road in Google Maps and it won’t show me the fucking road name, I have to scroll around until I find a label somewhere else. If I’m zoomed in on the road, just show me the name of the road @googlemaps
gwillen retweeted
this is why it's so valuable to be a curator-archivist. eventually you become just as important to the ecosystem as the people who made the works. i'd tentatively argue maybe even more important, from the perspective of the recipients. a great curator can preserve 1000s of works
gwillen retweeted
1/ Two hantavirus questions are now live on @Metaculus, both tied to a real, ongoing outbreak.
Here's why we're tracking them, & why we were careful in how we framed the questions. 🧵
gwillen retweeted
Because Waymo is rapidly scaling up its operations, we're rapidly accumulating more evidence on the question of whether it's safer than human drivers. The case for Waymo being safer than human drivers in in some sense almost twice as strong as it was in October.
gwillen retweeted
This is why working in public policy can be really high leverage:
17 million students fill out FAFSA each year. When FAFSA broke, they asked Jeremy Singer, president of College Board, to enter the government on a 6-month stint to get the application back up.
As a result of Singer’s salvage operation, 1.7 million more kids are getting the max Pell grant each year.
statecraft.pub/p/when-fafsa-…
gwillen retweeted
Fun fact: There's an MMO game on Steam with over 100,000 purchases that, if played, can allow any other player in the game world to gain remote access to your computer. The devs are aware of the vulnerability and actively trying to hide it. Steam was notified and did nothing.
gwillen retweeted
I read this article about software development, which I knew about because I saw Prime reacting to it:
notashelf.dev/posts/curse-of…
For the most part I think it is fine: a relatively young programmer is doing the healthy work of introspecting on what he should really be doing.
But there's one part of the article that I think is a deep mistake, and the author doesn't know it's so wrong because he has never experienced the alternative:
"Software doesn’t stay solved. Every solution you write starts to rot the moment it exists. Not now, not later, but eventually. Libraries deprecate. APIs change. Performance regressions creep in. Your once-perfect tool breaks silently because libfoo.so is now libfoo.so.2. 2
I have had scripts silently fail because a website changed its HTML layout.
I have had configuration formats break because of upstream version bumps.
I have had Docker containers die because Alpine Linux rotated a mirror URL.
In each case, the immediate emotional response was not just inconvenience but something that moreso resembles guilt."
Yes, this is true in much of the programming world. But there is another world in which people build things that last much longer. I have done it many times. I shipped a binary for this game Braid in 2009 that you can still download and play on Steam 16 years later. If you are pretty young (like 35), you can run binaries on Windows that were compiled before you were even born, which is amazing given how hard they have been trying to f up Windows lately.
On an emulator like MAME, you can play arcade games programmed in 1979. If today's software "technology" is so much better, why does it fall apart like tissue paper?
The author is not wrong about the cited decay. But this decay is not inherent to the practice of software. It's due to choices made, usually foolishly, by the people designing the systems being interacted with. And, it's due to a lack of knowing better, non-exposure to the sector of programmers who are very concerned with their code lasting a long time, actually.
The way you make code last a long time is you minimize dependencies that are likely to change and, to the extent you must take such dependencies, you minimize the contact surface between your program and those dependencies.
The actual algorithms you program, the actual functioning machinery you build, is a mathematical object defined by the semantics of your programming language, and mathematical objects are eternal, they will last far longer than your human life. The goal then is to avoid introducing decay into the system. You must build an oasis of peace that is insulated from this constant bombardment of horrible decisions, and only hesitantly interface into the outside world.
This means, for example: If you are shipping on iOS, you only reluctantly use any functions iOS gives you, because when you use them, Tim Apple will come along and break your program next year for arbitrary pointless reasons, because Tim Apple does not respect you or anyone you know.
This means a program cannot last forever on iOS, because Tim Apple likes breaking your things and watching you submissively clean them up. But the core of your program, which could be 95% of the code, is fine, and you can deploy it elsewhere.
This means you have to insulate from Linux userspace, because of all the jackass decision making that introduces constant incompatibilities while somehow never making the system better.
Using a library dependency to do font rendering or sparse matrix math? That dependency gets checked into your source tree, a copy of exactly the version you use. Ten years later you can pull down that source and recompile, and it works, because your program is a mathematical object. If you want to upgrade to something newer that has bug fixes and so forth, you are free to do so, but you are also free not to do so, and your program still works. (And how many of these bug fixes do you really need? Your program worked correctly when you shipped it to the greatest extent you could measure, because you are a skillful software engineer who wants to ship things of a high quality).
Everyone who got into programming for the joy of it knows, at some level, that the magic of programs is that they represent complexity that is replicable over time (and thus they exist outside of time). But the trashy programmer culture of the past 20 years stopped aspiring to this, and now has forgotten it is even possible.
And so long as people have forgotten, decisions will continue to be made that make the problem worse.
There are programmers who only write glue code, and who think that's what programming is; to these people what I have written above will not make sense. But the good news for that contingent is, they can always just stop writing glue code and start doing something else! If today's software "technology" is so good, why do you think it needs so much glue? Maybe there is a stylistic problem.
So if you are looking for what to do in the world of software that can represent a lasting contribution, maybe this is food for thought.
@NotAShelf @ThePrimeagen
gwillen retweeted
Introducing VibeCon — the world’s largest vibe coding conference.
Make sure you register today: http://127.0.0.1:8080/register
gwillen retweeted
The baby is sick today. She has a new word - why? - and she uses it when she's in distress, so when she's sick she snuggles up against me and whimpers "why? why? why?" and it is heartbreaking.
When she was three days old we got a preliminary test result suggesting she didn't have a functioning immune system. Severe combined immunodeficiency (SCID); you probably know it as bubble boy disease. The doctors told us it was probably a false positive and not to worry too much while they ran the followup tests. I immediately started worrying, of course. Sure, probably a false positive. But how probably? What were her prospects, if she was immunodeficient? SCID is treatable these days, 95% survivable if you catch it soon enough, but babies have better odds if they don't get an infection before their bone marrow transplant, so we started planning how we'd keep her safe until then. A week later we got the followup test result. False positive, like they'd told us it would be. No immunodeficiency. She is sick and she will fight it off and be completely fine.
There are five hundred thousand kids who PEPFAR provides HIV drugs that keep their immune system working. AIDS kills babies very quickly if they aren't on antivirals. Two to six months, which is also about when most babies with SCID die. It's about how long you can make it in this world without an immune system.
I've spent a lot of time yesterday and today in the shower - the hot steam helps the baby - cradling her to my chest while she whimpers "why, why, why?" and knowing that she will be completely fine. The world where this cold could have killed her feels like it is terribly nearby. The children that immunodeficiencies do kill don't feel very far away, either.
There are a bunch of people in my mentions on the PEPFAR post that seem to imagine that empathy is a limited resource, that if we care about the distant dying children we cannot care about our own. This is not my experience. To love another person, particularly to love a child, is to hate disease, to see it clearly as humanity's oldest enemy, to wake up and fall asleep determined to destroy it. I want to end the common cold. I want to end HIV. I want to end every immunodeficiency out there. I want this selfishly, greedily, desperately. There is no altruism in it and no philosophy. Disease is the enemy and I want to watch it die. It is my enemy, and it is your enemy, and it is on the retreat. I am not asking for your empathy. I am asking you to fight.
gwillen retweeted
Watch the waves from yesterday’s M7.6 earthquake in the Caribbean Sea roll across seismic stations in North America.
gwillen retweeted
I don’t want to connect my coffee machine to the wifi network. I don’t want to share the file with OneDrive. I don’t want to download an app to check my car’s fluid levels. I don’t want to scan a QR code to view the restaurant menu. I don’t want to let Google know my location before showing me the search results. I don’t want to include a Teams link on the calendar invite. I don’t want to pay 50 different monthly subscription fees for all my software. I don’t want to upgrade to TurboTax platinum plus audit protection. I don’t want to install the Webex plugin to join the meeting. I don’t want to share my car’s braking data with the actuaries at State Farm. I don’t want to text with your AI chatbot. I don’t want to download the Instagram app to look at your picture. I don’t want to type in my email address to view the content on your company’s website. I don’t want text messages with promo codes. I don’t want to leave your company a five-star Google review in exchange for the chance to win a $20 Starbucks gift card. I don’t want to join your exclusive community in the metaverse. I don’t want AI to help me write my comments on LinkedIn. I don’t even want to be on LinkedIn in the first place.
I just want to pay for a product one time (and only one time), know that it’s going to work flawlessly, press 0 to speak to an operator if I need help, and otherwise be left alone and treated with some small measure of human dignity, if that’s not too much to ask anymore.
gwillen retweeted
“Sure, it would extend and improve millions of lives, succeeding where decades of other efforts have failed, but this would be bad because people would still be enjoying themselves. We must not forget that the goal is ending pleasure, not ending cancer.”
