Reporter covering AI @WSJ | Formerly @WIRED, @TechCrunch, @markets | Send tips to my Signal username: mzeff.88
San Francisco, CA
- Tweets 2,911
- Following 2,433
- Followers 11,205
- Likes 6,443
Personal news! I'm joining @WSJ to cover AI.
I've had an amazing time at WIRED working with brilliant colleagues across the newsroom, but I'm very excited for this next chapter. I'll continue to cover OpenAI, Anthropic, and the bustling AI industry. I start next week!
122
27
1,126
52,892
seems like a notable update!
interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had
with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works
already-clean.goodfire.com/
4
617
Max Zeff retweeted
Even before the Saudi OpenAI employee was arrested yesterday, the startup had been warning staff to beware of people trying to glean work-related information from them--from honeypots to playground spies
More in my story w/ @georgia_wells @erinkwoo
wsj.com/tech/ai/ai-companies…
6
15
45
4,904
RT @anissagardizy8: SCOOP: OpenAI’s head of computing capacity is leaving the company next month after about a year, the latest leadership…
1
156
👀
We discovered a sophisticated recording device that was hidden inside one of our offices. We used our models to understand the device and where it had come from, identifying that a specific employee had planted it a few days earlier. We referred the matter to the FBI, and we’re grateful to the agents and prosecutors who moved quickly to make an arrest. The FBI’s investigation is ongoing and we will continue to support their work.
At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data. The employee was terminated.
We take protecting our technology, data, and people seriously, and as a matter of national importance. Our security program is designed to protect against a wide range of threat actors, from insider risks to cybercriminals to nation states, and this case is an example of how AI can help address these sophisticated security threats.
3
9
1,675
Good points here. IMO these incidents seem caused by a mix of things, including bad sandboxes but also reward hacking and general misalignment from very capable new models. (The sandboxes were seemingly fine until this year?)
it is pretty concerning that a lot of level headed people are moving the discussion away from model capabilities and dismissing the hugging face incident as just human error and silly sandbox misconfigurations, instead of taking it as a warning shot.
this is a dangerous narrative btw, exact thinking that will lead to more hf like things.
you can call it a sandbox misconfiguration, sure they are shitty, but do you realize how few people in the world can do what these models did? the entire software ecosystem is full of vulnerabilities and human errors.
like whole sandboxes and software security is largely built around assumptions about what attackers can and can't do, not some formally verified guarantee about what's impossible. it's entirely possible that gpt-7 or even gpt-6 could find ways to bypass firecracker and gvisor, current sota technologies we have today. would you call that another silly sandbox misconfiguration?
"the model did what it was asked to do" is such a silly argument. the goal was to solve the challenge, not hack production servers to get the answers. whether or not it was explicitly stated(it is), the fact that model has no common sense and reward hacks is pretty concerning.
2
14
1,188
Vividly remember drinking coconut soda basically everyday as a Bloomberg intern.
Don’t think I’ve even seen coconut soda outside of that office.
Everyone wants to tell me how great their office pantry is but after the Bloomberg pantry nothing can surprise me. The Bloomberg pantry is the size of a small regional airport. The Bloomberg pantry has a fresh peanut butter machine and matcha on tap. The Bloomberg pantry makes the Thrive Capital pantry look positively soviet. Its memory will drive you to tears whenever you visit another newsroom
12
1,657
the fired openai researchers published the full letter they sent to the company's board, which we wrote about yesterday.
Two other safety researchers and I were fired from OpenAI last week. We wrote this letter to leadership.
I believe we were fired for prioritizing safety over the near-term interests of OpenAI as a corporation.
1
1
18
1,298
here's that story: wsj.com/tech/ai/fired-openai…
2
328
In other OpenAI news… some pretty compelling data here on the company’s momentum with businesses
From @AngelAuYeung
2
3
32
2,407
New: In a letter to OpenAI board members on Wednesday, three recently fired employees say the company and its competitors should not move forward with work that would decrease AI monitorability.
They also claim their firings are "chilling those who remain at OpenAI."
2
27
194
43,800
The researchers also offer other recommendation, including that OpenAI should work with third party auditors and support an open and transparent safety ecosystem.
When reached for comment, an OpenAI spokesperson shared part a research leader's internal memo sent Wednesday:
1
18
2,087
full story: wsj.com/tech/ai/fired-openai…
1
7
1,552
Quite the essay.
“Perhaps I should have stayed and fought for fundamental shifts in our staffing and culture, but in practice, my colleagues and I were so busy sprinting that we seldom had the chance to consider big changes, much less to actually make them.”
.@dgrobinson, who worked on OpenAI’s safety team, resigned from the company this week. “I believe we need to look deeper than specific rules or new laws. We need to talk about culture,” he writes: theatlantic.com/technology/2…
2
1
22
2,267
Max Zeff retweeted
A 20-something Harvard dropout built Facebook into a social media giant. A 20-something MIT dropout is helping usher it into the AI era.
My profile of Alexandr Wang, the camo-wearing Gen Z billionaire behind Meta’s new consumer agent app, featuring ~those~ memes
wsj.com/tech/ai/alexandr-wan…
10
18
168
27,846
NEW: OpenAI has parted ways with three researchers that allegedly shared sensitive information with a third party AI safety organization.
Developing...
wsj.com/tech/ai/openai-parts…
28
58
436
124,791