which imo means interp people have a duty to be responsible/measured in how they present their work
I think it would have been nice if Nick/Goodfire had shared their findings beforehand not only with their friend (roon), but also with other interp experts who could provide a more independent assessment.
I think a great example is when Anthropic asked Neel Nanda and his team to replicate their work on J-space and write an independent critique in advance.
If Nick’s work had been accompanied by a review from Neel or other experts, we might have been able to avoi... (read more)
I think it would have been nice if Nick/Goodfire had shared their findings beforehand not only with their friend (roon), but also with other interp experts who could provide a more independent assessment.
I think a great example is when Anthropic asked Neel Nanda and his team to replicate their work on J-space and write an independent critique in advance.
If Nick’s work had been accompanied by a review from Neel or other experts, we might have been able to avoi... (read more)