econ phd student at berkeley
berkeley, california
Joined December 2013
- Tweets9.5K
- Following1.4K
- Followers1.5K
- Likes52.3K
Pinned Tweet
Life update: I’m beyond excited to start my PhD in Economics at UC Berkeley this Fall! I am grateful to have had many wonderful people who contributed to my journey until here.
Here’s a thread of thanks:
William Radaic retweeted
Lots of people pointing out that reading one book a week with mandatory lecture attendance isn't, objectively speaking, a high bar. Fair enough.
But if you’re restoring standards in higher ed, you’ve gotta start somewhere. This course’s popularity is great news!
This @Harvard class is characterized as “brutal” because students are required “to read a hefty book a week.”
That was what literally all my English classes at @Columbia required 25 years ago.
Oh, and students are required to show up.
Where have standards gone?
wsj.com/lifestyle/harvard-cl…
William Radaic retweeted
Excellent paper with high relevance to the AI writing discussion.
“I discuss my ideas with the AI, it’s all my ideas and the AI is just writing them.”
Same category of statement as “most people are bad at multitasking but not me”
William Radaic retweeted
Lula winning means this clip will be immortalised, as it rightfully deserves
William Radaic retweeted
Lula wins, and it wasn’t particularly close — 3.5% or more than 4 million votes.
It’s an epic comeback and we will need time to fully understand it. But one thing is clear — the extension in voting hours was not the difference maker in this race.
William Radaic retweeted
>two agents in separate sandboxes
>no internet
>no network
>no shared files
>openai: "ok now try to talk"
and they talk through the CPU cache 💀
Over the past several months, we have substantially hardened the security of our research training environment. We’ve removed live internet access for all frontier models during training and built an offline world for them to use instead—a cached version of the web where they can still learn to browse, search, and use applications.
On September 20, after a model exploited a DNS misconfiguration to find a way outside of this offline world, we paused frontier training to strengthen containment further. We assembled teams of agents, powered by our most cyber-capable models, and tasked them with finding weaknesses in our research environments.
While they found no route to the internet they could easily exploit while concealing their activity from our monitors, they did find zero-day vulnerabilities, virtual machine escapes, and hardware side-channels for communication between agents. We used their discoveries to further strengthen our defenses and contribute fixes to open-source software.
After strengthening our containment architecture, we resumed our frontier RL training and inference on October 1. We’ve reserved substantial compute for continuous red-teaming, and made high-compute red-teaming part of the safety cases we use to decide whether frontier training and evaluation can proceed securely.
openai.com/index/securing-ou…
