69 points by dr_dshiv
2 months ago
> If GPT-10 is just like GPT-3, no extra bells and whistles, just bigger, and able to read whole novels and watch whole movies and answer subtle and open-ended questions about characters and their conflicts and motivation, and tell us when to laugh, I promise to post a YouTube video admitting I was wrong.
-- Gary Marcus
This isn't exactly what he described, but it'd make me pretty worried if I were him.
> This isn't exactly what he described, but it'd make me pretty worried if I were him.
Admitting when we are wrong should be the default, not a cause for worry. That someone is willing to do so in this age should to be celebrated (or at the very least encouraged).
Gary has directly said he "hopes" he's wrong, so I think he'd be surprised and happy. BTW, his primary criticism of current AI approaches like GPT is he thinks it unlikely simply scaling the size of training data for models like this will yield AGI.
I would be very happy if this was good enough to help me finish the bedtime story I've constructed for my 9 year old. I've clearly lost the plot and she keeps poking holes in it.
Reminds me of the credits from the Futurama soap opera All my circuits.