◀ IA · 17 NOV4/13
Segurança
OpenAI assina acordos com avaliadores independentes de segurança de IA
A OpenAI assinou acordos com avaliadores independentes: Apollo Research, METR, Transluce e Goodfire.
We have signed agreements with four independent assessors—Apollo Research, METR, Transluce, and Goodfire
openai.com
Os avaliadores terão acesso a recursos internos, como checkpoints de modelos, código-fonte e dados de treinamento.
assessors receive access to some of the same internal resources our researchers use, including model checkpoints, source code, training and evaluation data
openai.com
A Apollo Research vai avaliar se os modelos poderiam agir secretamente contra as intenções dos desenvolvedores.
will assess whether models could covertly work against their developers’ intentions
openai.com
O piloto com a Apollo começa nesta semana, segundo a OpenAI.
with Apollo’s beginning this week, according to OpenAI
techcrunch.com

