If you missed the last Mystery AI Hype Theater 3000 with @emilymbender and me, no worries! We figured out how to do VOD, so you catch Episode 11 on Twitch for 14 days:

https://www.twitch.tv/dair_institute

#AIHype #SALAMI #MathyMaths

new paper out in Political Analysis suggests replacing human survey respondents with “silicon sampling” of simulated individuals from GPT-3

https://arxiv.org/abs/2209.06899

@politicalscience #AIhype #ai #LLM #mathymaths

Episode 7 of Mystery AI Hype Theater 3000 is out! @emilymbender and I talk with @trochee about evaluation, benchmarking, and Gish Gallop.

#AI #AIHype #MathyMaths

https://videos.trom.tf/w/om4kpDHuoYUAw1BjZqNTqU

🚩 : Appeal to the future + describing #MathyMaths as "sophisticated". More to the point --- the way to get automatic plagiarism machines that obscure the origins of the material is to design training tasks that optimize for that. This isn't something that "AIs" will just "decide" to learn to do or "evolve" to be able to do.

>>

Some people have suggested that large language models could help people improve their English writing. I'm cautious about this because... why try to present your English in inauthentic way? I would rather live in a world where we embrace all kinds of English instead of forcing people to use an algorithm to make it seem "professional"

#AI #ML #MathyMaths

I have a standing search on the phrase #StochasticParrots and its variants over the birdsite and it's been very interesting to watch. One phenomenon that seems to be emerging more recently is that people consider it a "slur" or an "insult". I'm curious if this comes from people actually using it with some intent, from other people over-identifying with their so-called "AIs" (#MathyMaths), or some combination of the two.