Loved Subbarao Kambhampati (కంభంపాటి సుబ్బారావు)'s 'reverse Turing' test: Just ask the LLM-or-human to convert a large amount of text into Shakespearean style. LLM can do this in *seconds* whereas even Shakespeare could not! You would immediately know you are not talking to a human. (Any task that is hard to do

I have heard concerns that this same vulnerability made many of the govt AI Safety committees get disproportionately populated by those supported with Moskowitz money..

[The rest of Prof. Edelson's thread is very much worth reading too..]

Over at WaPo, Prof. Gary Smith has excellent advice on how to fix the finances of US colleges!🤯

washingtonpost.com/opinions/2024/…

Happy Earth Day to y'all from #Tempe

No. Training LLMs on purely factual data STILL WON'T cure them of 'Hallucinations' #SundayHarangue

There is a persistent myth that LLM hallucinations are just a result of them being trained on un-curated and 'non-factual' data, and will go away with high quality/factual data.

Reread it after an eternity--and still won't change much.. ICYMI 👇

👉 thehill.com/opinion/techno…

For some definition of prediction, astrologers do some kind of prediction. #AIAphorisms

Hyper paper productivity is apparently alive and well in the era of ChatGPT..

Even in these days of  inflated resume expectations, it is perhaps more important to be known for a  contribution rather than a count; for a significant idea,  than for breaking a Guinness book record

Spoke at Megagon Labs yesterday. One unexpected bonus was having Tom Mitchell in the audience--in person and engaged--for the entire 2.5 hours!🤗

💡Idea for taming hype: Whenever a hype-fueling paper is shown to be off base by a new paper, the former's citations should be transferred to the latter debunker paper. After all, thou shalt not be allowed to enjoy ill gotten gains er. citations, no?

(No I may not necessarily