New paper from Anthropic show alignment in LLMs... It seems to me that AI safety needs a technical unlock because alignment is not going to ensure safety...
December 22, 2024 at 3:25 AM
New paper from Anthropic show alignment in LLMs... It seems to me that AI safety needs a technical unlock because alignment is not going to ensure safety...
Thinking one particular data quality metric: "Number of people who have changed last names in the past month" I heard that number is around 2 million. But regardless of the number, I find it difficult to use this metric because what can I do about this? It doesn't get me any closer to better data...
November 28, 2024 at 2:23 AM
Thinking one particular data quality metric: "Number of people who have changed last names in the past month" I heard that number is around 2 million. But regardless of the number, I find it difficult to use this metric because what can I do about this? It doesn't get me any closer to better data...