My Authors
Read all threads
Whenever someone points out sexist/racist stereotypes in a new AI tool, you’ll find lots of apologists in the comments saying, "What's the surprise? It just reflects the training data".

Well, the "surprise" is that researchers keep releasing these tools as if everything’s fine.
To OpenAI's credit, they released a (rudimentary) model card based on the "Model Cards for Model Reporting" paper github.com/openai/gpt-2/b…
arxiv.org/pdf/1803.09010…
Sadly, it's unlikely that most people who downloaded the model even noticed the model card, let alone abided by it.
OpenAI said they wouldn't release one of their models because it could be used by political actors to generate seemingly realistic disinformation text. This follows a long line of claims that AI is dangerous because it's too good, when in fact it's harmful because it's too crude.
Many AI researchers feel that stereotypes in training data are society's problem, not theirs. But social norms evolve rapidly. Sexist comments commonplace 5 years ago would lead to a backlash today. If you choose to include text from the 60s in your training set, that's on you.
Oops, wrong link to "Model Cards for Model Reporting" earlier in the thread (link goes to "Datasheets for Datasets" instead). Correct link: arxiv.org/pdf/1810.03993…

I meant to say that Datasheets for Datasets is *also* a great paper on how to do ML in a more responsible way :)
Good to see that OpenAI have released a follow-up report in which they present preliminary work describing some of the stereotypes in the model's outputs.

Even better would be to make it a norm that this type of work be done *before* releasing a model.
Missing some Tweet in this thread? You can try to force a refresh.

Enjoying this thread?

Keep Current with Arvind Narayanan

Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Follow Us on Twitter!

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!