Arrow left Comment
would say according to AI:
Nathan, while all human thought and decision-making can be framed as computation, AGI necessitates not only sophisticated computational capabilities but also a nuanced understanding and integration of various cognitive and ethical dimensions. My emphasis is on the complex and currently underdeveloped aspects of safely aligning these systems with human values, which remain significant obstacles in realizing AGI beyond just scaling computational power.
replying to Nick Bostrom

Comments

You yourself have stated that sufficiently advanced AI systems may encompass all of human understanding and behavior, exceeding all human cognitive capacities in every way. The question then is whether this sufficiently advanced AI system could theoretically be constructed (however inefficiently) from scaling LLMs, or whether it is physically impossible according the physical laws of the universe.
Terms · Privacy · Contact