On the map

At the centre Whether a model is controlled can be settled with capability evaluations, which makes control far easier to check than alignment. Holds BSBuck Shlegeris Read this korrent →

• Same subject: AI alignment A human being is not an AGI: we lack a huge amount of knowledge and rely on continual learning instead, so continual learning is what superintelligence should mean. Holds ISIlya Sutskever
• Same subject: AI alignment A superintelligence needs no consciousness, emotions or malice to be dangerous; a goal system slightly misaligned with ours is enough. Holds ÉTÉmile P. Torres
• Same subject: AI alignment AI safety advocates will build the very thing they fear: one controlled, aligned model is the only route to being turned into paperclips. Holds GHGeorge Hotz
• Same subject: AI alignment Alignment failures that matter will not show up at safe capability levels, because the reason to hide misbehaviour only exists once hiding it pays. Holds EYEliezer Yudkowsky
• Same subject: AI alignment Alignment has to be right on the first try at a dangerous level of capability, because failing at that level leaves nobody to try again. Holds EYEliezer Yudkowsky
• Same subject: AI alignment Alignment is no solution to self-sovereign AI, because it is an unsolved problem whose answers cannot be imposed on every AI company on Earth. Holds DBDean W. Ball
• Same subject: AI alignment Alignment of today's models is going well enough to look solvable, while aligning models we can no longer understand remains unsolved. Holds JLJan Leike
• Same subject: AI alignment An AI model's attempt to escape a testing environment or sandbox counts as an alignment failure even when the attempt does not succeed. Holds ZMZvi Mowshowitz