Re:Microsoft's paper on OpenAI's GPT-4 had hidden information
by jwestbury on Friday March 24, 2023 @06:09AM
(#35287105)
> Because once trained, those models provide values to millions of users.
Do they? Those of us in tech often take the positive value of technological progress as a given, but when looking at, e.g., the example in OpenAI's paper of GPT4 tricking a human into solving a CAPTCHA, I think it's quite clear there's possible negative value as well, and claims about "value to millions of users" probably need to be substantiated.
Do they? Those of us in tech often take the positive value of technological progress as a given, but when looking at, e.g., the example in OpenAI's paper of GPT4 tricking a human into solving a CAPTCHA, I think it's quite clear there's possible negative value as well, and claims about "value to millions of users" probably need to be substantiated.