Gary Marcus

Rank 21 of 47
|
Score 29
2/17/2025, 2:44:34 PM
X
In reply to:
Alice
@AliceFromQueens
·
277d
@KtunaxaAmerika @GaryMarcus I'm sure many are attempting to build this
Noa Kt 🇨🇺 🇺🇸 🇨🇦 🇧🇪
@KtunaxaAmerika
·
277d
@AliceFromQueens A big advance for LLMs would be for them to come with a hallucination detector and a sarcasm/joke detector for their preliminary results, and say “hey, that isn’t right because x, what’s the real answer”. Basically automate what people who use them have to do. @GaryMarcus
Noa Kt 🇨🇺 🇺🇸 🇨🇦 🇧🇪
@KtunaxaAmerika
·
277d
@AliceFromQueens Yeah, when I searched … I didn’t get you.
lexaprofessional
@queasy_f_bby
·
1541d
my favorite bodega cat, blue/cheddar/trouble. HUGE attitude. love him
Alice
@AliceFromQueens
·
277d
@KtunaxaAmerika I'm not sure what that referred to and I suspect it was hallucinated. There were other pure fabrications in the research paper
Noa Kt 🇨🇺 🇺🇸 🇨🇦 🇧🇪
@KtunaxaAmerika
·
277d
@AliceFromQueens I missed your bodega cat content.

Was this pre Yom Kippur 2019?
Alice
@AliceFromQueens
·
277d
I asked Deep Research to solve the mystery of the person behind the account @alicefromqueens.

After debunking the Freddie DeBoer rumor (much to its credit!) it raised some weird possibilities, which it went on to reject as well.

I'd say it overweights 2020 election content

The statement discusses the practical challenges of implementing a feature in language models, specifically a hallucination and sarcasm detector. It is a technical observation rather than a substantive engagement with public issues, and thus does not constitute public discourse.