We've trained an unsupervised language model that can generate coherent paragraphs and perform rudimentary reading comprehension, machine translation, question answering, and summarization — all without task-specific training: https://blog.openai.com/better-language-models/ …pic.twitter.com/360bGgoea3
-
Hi there. We've released a paper, samples, and small model. Not our intention to hype our research relative to others (DM me specifics here and can probably help?). How do you think about intersection of increasingly general models and potential for malicious use?
-
-
I have been pushing for making reproducibility in
#AI conferences.@openAI is deciding to just go to media directly without engaging with research community. This is not science. It is hype. Media should boycott this@tsimonite -
In terms of engagement, one of our ideas here is to have a big public discussion about publication norms, so I'm really glad we're having this conversation. What do you think are the best next steps to have productive conversations here?
- Show replies
New conversation -
-
-
Hiya, here are some malicious use cases (outlined in the blog). What do you think?pic.twitter.com/HwWu3Vxma5
-
This is exactly hype and misleading the media. Where is any evidence that your system is actually capable of doing this? Which independent researchers have scrutinized your system? None. Media people don't count.
- Show replies
New conversation -
-
-
Interested to see your response
@AnimaAnandkumar and curious to know what release strategies/tech impact assessments you have at@NvidiaAI -
This seems to be a more nuanced issue than you frame it to be... how to retain the spirit of openness that has shaped AI, while advancing the goal of responsible research
- Show replies
New conversation -
-
-
Yes, eagerly awaiting the list of so-called malicious cases so that Dr. Anima can respond!
#GOANIMAThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Respectfully, they do list malicious uses in blog post. Potential for producing fake content seems clear. Disinfo. campaigns & spammers surely would happily put trained model to immediate use, before field has worked out countermeasures. IMHO,
@OpenAI has a tough line to walk. -
This is really tough, and I'm glad we're getting to have a big public discussion about it!
End of conversation
New conversation -
-
-
So what? Does open source and open sharing stop something from having mailcious use cases? Like the recent and horrifying use of GANs for generating realistic deep-fake videos of individuals in pornographic situations.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.