Notices by Prof. Emily M. Bender(she/her) (emilymbender@dair-community.social), page 11

“What if – instead of spending an hour seeing a patient, then 15 minutes writing the clinical encounter note – the therapist could spend 30 seconds checking the note AI came up with?” YIKES

This is a completely unrealistic expectation about what goes into verifying that kind of note and sounds like a recipe for overburdening the medical workforce/setting up errors.

/6

What if -- instead of seeing the process of creating clinical documentation as mere busywork -- the tech bros understood it as possibly part of the process of care?

What if -- instead of leading with the 'gee whiz AI' angle -- journalism in this space started with privacy harms, the fact that somehow tech companies get away with pretending healthcare regulation doesn't apply to them, and chatbots urging self-harm?

/fin

Either it's a version of ChatGPT OR it's a search system where people can find the actual sources of the information. Both of those things can't be true at the same time. /2

Also: the output of "generative AI", synthetic text, is NOT information. So, UK friends, if your government is actually using it to respond to freedom of information requests, they are presumably violating their own laws about freedom of information requests. /3

(This answer to my freedom of information request is raising a lot of questions not already answered in this answer to my freedom of information request...) /4

Here they're basically admitting they don't trust it to speak for them (and they shouldn't) but also that they think that some government communication is just so much BS. /5

And the juxtaposition here is just appalling. No money for actual public services (aka the function of government) but sure, let's keep increasing the budget for the terrorist^H^H AI cell. /6

The level of magical thinking here is astonishing: /7

It seems like there are just endless bad ideas about how to use "AI". Here are some new ones courtesy of the UK government.

... and a short thread because there is so much awfulness in this one article.
/1

https://www.ft.com/content/f2ae55bf-b9fa-49b5-ac0e-8b7411729539

Just got back from the Improbable Research show at #AAAS2024 and was reminded of this masterpiece:

https://www.youtube.com/watch?v=yL_-1d9OSdk

Working with @alex on the pod and some papers has been a dream. Now we get to do a whole book :) Huge thanks to Ian Bonaparte for expert representation and super excited to work with Hollis!

Folks, papier-mâché made out of good training data is still papier-mâché. So-called "generative AI" is never ever ever a good information source. This looks to me like lawsuits waiting to happen....

https://www.msn.com/en-us/news/other/hate-taxes-h-r-block-s-new-ai-chatbot-aims-to-reduce-your-tax-frustrations/ar-BB1hqspF

>>

I feel very vindicated for not making time to answer journalist's queries about papers that "prove" things based on hypothesized graphs and fabricated data.

Next up on Mystery AI Hype Theater 3000: @alex and I will take on the hype around AI and productivity: GPTs are *not* GPTs; "AI" is not here for your job ... but it might make it shittier.

Join us live at noon Pacific on Monday Jan 22:
https://www.twitch.tv/dair_institute

#MAIHT3k

In case anyone was still in doubt, Google is not at all interested in "organizing the world's information" (despite that language still being in their mission statement). Mixing non-information, LLM-extruded sludge in with authentic information is actually the opposite.

https://www.404media.co/google-news-is-boosting-garbage-ai-generated-articles/

I'd love to see the creation of platforms that require authenticity --- no synthetic media. The actual words + art of actual people only.

>>

Sometimes I get into arguments with people who say "But what about using grammar checkers/spell checkers"? We had both of those before LLMs were used to synthesize text, and they worked well enough, thank you.

This is funny, but also actually a really bad sign for general enshittification of the web. The most alarming detail here is that Amazon is actually promoting the use of LLMs to create fake ad copy.

https://arstechnica.com/ai/2024/01/lazy-use-of-ai-leads-to-amazon-products-called-i-cannot-fulfill-that-request/

Ready for some AI hell catharsis? Mystery AI Hype Theater 3000 episode 23 has dropped! "AI Hell Freezes Over" @alex and I somehow survive a harrowing journey through all the regions of AI Hell before the ice melts...

https://www.buzzsprout.com/2126417/14282708-episode-23-ai-hell-freezes-over-december-22-2023

Thanks, as always, to Christie Taylor for production!

Can't get enough Mystery AI Hype Theater 3000? This is your lucky week! Join us

Friday Dec 22, noon Pacific
AI Hell Has Frozen Over (If Only)
https://www.twitch.tv/dair_institute

... in which @alex and I will clear the backlog of Fresh AI Hell.

We're expecting this to be as cathartic as our previous pure hell episode (Ep 12, from May 2023).

#MAITH3K

A quick thread on #AIhype and other issues in yesterday's Gemini release:

#1 -- What an utter lack of transparency. Researchers form multiple groups, including @meg and @timnitGebru when they were at Google, have been calling for clear and thorough documentation of training data & trained models since 2017.

In Bender & Friedman 2018, we put it like this:

/1