You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
SotA LMs get published on a regular basis (with or w/o the model). Why did this one need an all-out PR assault? Whether releasing some models raises security issues is a legitimate debate (tho this model prolly doesn't). Is Bloomberg or the Guardian the right place to have it?
How is it not misrepresentation to take best-of-25 samples from cherrypicked prompts and tell the public, "the AI wrote this, this is what AI does now"? BTW I have a trading algo that does 12,000% per year* (*evaluated with best-of-25 daily trades)
How is it not needless fear mongering and hype to tell the journalists, "we'll keep the AI secret because releasing it would be too dangerous, trust us", when the actual model is an incrementally better LM? Where's the threat model? Etc. It's layer after layer of issues.
I like you guys and I think you do good research. Please be more careful next time.
I'm maybe missing something - the AI *did* write that/does that now, and the reason you know it was cherrypicked is because it said it right there?
(of course you would have suspected cherrypicking regardless, as you should - I just mean that it wasn't something we hid at all)
But they also did no fine-tuning, which was very conservative.
Yeah, I think that's a point missed in the "hype" accusations - these are a super low lower bound - minimal hyperparameter changes across samples/no fancy sampling, no fine-tuning, often inexperienced users generating the prompts or prompts intended to be difficult, etc.
This is why I'm so skeptical when someone says "AI wrote this poem" or "AI wrote this song." If you just keep running it until you find something that sounds good to people...
I think the peak quality (over long lengths of text) has grown a fair amount, such that there are things you can generate in a realistic amount of time now that you couldn't before, but I agree that transparency around cherrypicking etc. is key, and it's harder w/o sharing model
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
