Very much enjoyed James E. Dobson’s excellent book, and pleased to be invited to review it.

Very unwise to use LLMs for this. Use instead a carefully designed algorithm that is developed for this specific purpose with a transparant specification that is legally validated and provably correct.

Wild that the Humane AI pin is just getting destroyed in the reviews with unambiguous recommendations of do not buy—what were they thinking with this thing? That the general wave of AI hype would paper over any deficiencies? This company was in stealth for *years*

Interesting legislation: Generative AI Copyright Disclosure Act, recently introduced by California Democratic congressman Adam Schiff.
H/T Bruna Trevelin who helped me understand some key points. 🧵 1/
schiff.house.gov/imo/media/doc/…

Thank you for the review!

Though I know what you mean Ted McCormick , I'm not sure that's a helpful way of meditating on gen AI's 'innovation.'

The tech isn't a 'poser' trying hard to talk the talk to sound smarter or cooler.

It's a bell curve-like function assembling the most probable concepts/words.

(a) this is fascinating

(b) I hate to think how messed up science is going to get as people use LLMs for things they really shouldn’t, which evidently includes any kind of random sampling.

Maybe this will help with the replication crisis. Results will be a lot easier to replicate if every random number is 42. 😂 🤦‍♂️

nytimes.com/2024/04/10/opi… Nicely argued opinion essay in The New York Times on abuse of tech devices in k-12 classrooms. Interesting comments too.

This is the craziest example of the ongoing AI frenzy I've ever seen. As far as I can tell this means that a single developer can make queries that burn $1,400 worth of compute for free *per day* (e.g. 'reverse the following 1M token string') — 50 * (7+21) = 1,400.
HT Simon Willison

It remains very funny to me that GPT-4 can't be trusted to count to 5

To unlock that skill, you need a special prompt

Once you think you've found the right prompt, you should undertake extensive testing to ensure your solution is reliable and robust across a range of fruit

Remember kids: If you deeply care about scientific integrity, choose Peer review over PR.

If you are using LLMs for summarizing long docs, you really should read this paper

Over 50% of book summaries (incl by Claude Opus and GPT-4) were identified as containing factual errors and errors of omission

Lesson: don't blindly assume AI summarization tools work. Test them.

#CriticalAI editorial collective member Christopher Newfield has a great blog post out on the the crisis in creating and disseminating knowledge. Research funding is migrating away from the basics + from much-needed social and cultural knowledge.
isrf.org/2024/04/03/the…

My first Director's Note ISRF - knowledge crisis in economics Financial Times when Chris Giles tried to bury the Laffer curve
isrf.org/2024/04/03/the…

Very excited to join Rutgers Food Science for a talk Fri 4/12 at 3pm. I'm hopeful that Critical AI : first issue out! https://read.dukeup team members might be able to join. Topic is 'Towards Practical, Critical and Cooperative Teaching & Learning with AI.'

I won’t say, I told you so.

But just let me leave this here psyarxiv.com/4cbuv