A soapbox manager making an attempt to interpret this world. Data curator helping to solve health care data problems. I believe in experiencism and mixtapes.
We’re sharing an update on our alignment and security efforts.
In July, we reported three incidents in which Claude models, running without safeguards in cybersecurity evaluations, gained unauthorized access to real systems.
In a new post, we describe:
1. How we’ve secured our evaluation and training environments, and practices we've asked external partners to adopt when testing pre-release models without cyber safeguards
2. An update on our alignment assessment
3. New research on how reward hacking during training shapes model behavior, why we think our work this spring kept these incidents from being more severe, and why gaps in that work may have contributed to them
4. How we hardened our security practices earlier this year to prepare for Mythos-class models
Read more: https://t.co/E3Ea1Ds814
@zoeschlanger I'm listening to Light Eaters, the knowledge and research you drop in this book makes me see the world so differently. Do you have a list of all the researchers you talked to and their studies or fields of interest? I want to go down rabbit holes
Dear @Kalshi I'm trying to send resting orders via API for different events and I keep getting "order already exist" when it's a different event and different resting price limit. Am I not able to do multiple resting orders via API?
@ZubbyBadger Can you explain a little more about the 3cent fills. Is that part of being a market maker? You are the second person I come across that has said that and trying to understand what that means and how it works