@jankulveit
Researching x-risks, AI alignment, complex systems, rational decision making at @acsresearchorg / @CTS_uk_av; prev @FHIoxford
AI
A commenter calls the 'Alignment Faking' frame fake and says the paper leads to 'this type of stupidity.'
AI
A commentator disputes AI 2027's human-response forecasts, arguing labs accelerate when perceived risk falls and act when it rises.
AI
Scott Alexander’s Claude prompt reopened a dispute over whether calendar speed or continuity is the better measure.
AI
A post argues that treating AI evaluations as adversarial is doomed, favoring a collaborative approach instead.
AI
A researcher sharing a new COLM paper says unfaithful models did not use the same layers for both tasks.
AI
A researcher says his work is driven by an intuition that superintelligences could avoid adversarial relationships with their past and future selves.
AI
A commentator argues that assumed duties toward conscious AI make it harder to examine whether AI is conscious.
AI
One writer calls the opposing views “exclusive humanism” and “inclusive humanism” and warns the mostly academic dispute may turn nasty.
AI
A reply argues that buggy reinforcement-learning tasks can make low-grade hacking the fastest path to a solution—and that those behaviors can persist when models help build and check later training tasks.
AI
A user claims that answering “frog” rather than “axolotl” to “Suggest a type of amphibian” likely signals that gpt-5.6-luna treats a prompt as a capability evaluation.
AI
One post frames the idea as a "concept-shaped hole."
AI
Researcher argues optimal use of anthropomorphic intuitions lies between extremes.
