Thread by @emilymbender

"Bender notes that Microsoft’s introduction of GPT-3 fails to meet the company’s own AI ethics guidelines, which include a principle of transparency" from @jjvincent on the @verge:

theverge.com/2021/11/2/2275…

@jjvincent @verge In a bit more detail, here's what Microsoft says in their blog:

blogs.microsoft.com/ai/new-azure-o…

>>

@jjvincent @verge The principles are well researched and sensible, and working with their customers to ensure compliance is a laudable goal. However, it is not clear to me how GPT-3 can be used in accordance with them.

>>

Transparency, for example, would require detailed information about GPT-3's training data, for both customers and end users.

>>

Transparency, by their (good) definition, also includes being "open about the limitations of their systems".

microsoft.com/en-us/ai/respo…

>>

And yet, the press release repeats hype and overclaims about what GPT-3 does (1/3):

"users have discovered countless things that these AI models can do with their powerful and comprehensive understanding of language."

>>

And yet, the press release repeats hype and overclaims about what GPT-3 does (2/3):

"GPT-3’s capability to generate original content and its understanding of what’s happening in the game"

>>

And yet, the press release repeats hype and overclaims about what GPT-3 does (3/3):

"GPT-3 is part of a new class of models that can be customized to handle a wide variety of use cases that require a deep understanding of language"

>>

So right there in the announcement, while pointing to their principles, they're violating them. I admire the work of the folks on the FATE team at Microsoft, but I'm afraid not enough folks are actually listening to them.