Redwood Research blog

Redwood Research blog

Home
Archive
About
Jankily controlling superintelligence
How much time can control buy us during the intelligence explosion?
 • 
Ryan Greenblatt
What does 10x-ing effective compute get you?
Once AIs match top humans, what are the returns to further scaling and algorithmic improvement?
 • 
Ryan Greenblatt
Comparing risk from internally-deployed AI to insider and outsider threats from humans
And why I think insider threat from AI combines the hard parts of both problems.
 • 
Buck Shlegeris
3
Making deals with early schemers
...could help us to prevent takeover attempts from more dangerous misaligned AIs created later.
 • 
Julian Stastny
, 
Olli Järviniemi
, and 
Buck Shlegeris
2
Prefix cache untrusted monitors: a method to apply after you catch your AI
Training the policy to not do egregious bad actions we detect has downsides and we might be able to do better
 • 
Ryan Greenblatt
1
AI safety techniques leveraging distillation
Distillation is cheap; how can we use it to improve safety?
 • 
Ryan Greenblatt
When does training a model change its goals?
Can a scheming AI's goals really stay unchanged through training?
 • 
Vivek Hebbar
 and 
Ryan Greenblatt
2

May 2025

The case for countermeasures to memetic spread of misaligned values
Defending against alignment problems that might come with long-term memory
 • 
Alex Mallen
1
AIs at the current capability level may be important for future safety work
Some reasons why relatively weak AIs might still be important when we have very powerful AIs
 • 
Ryan Greenblatt
Misalignment and Strategic Underperformance: An Analysis of Sandbagging and Exploration Hacking
A new analysis of the risk of AIs intentionally performing poorly.
 • 
Julian Stastny
 and 
Buck Shlegeris
Training-time schemers vs behavioral schemers
Clarifying ways in which faking alignment during training is neither necessary nor sufficient for the kind of scheming that AI control tries to defend…
 • 
Alex Mallen
What's going on with AI progress and trends? (As of 5/2025)
My views on what's driving AI progress and where it's headed.
 • 
Ryan Greenblatt
2
© 2025 Buck Shlegeris
Privacy ∙ Terms ∙ Collection notice
Start writingGet the app
Substack is the home for great culture