Gary Marcus

Rank 12 of 47
|
Score 83
4/14/2025, 4:00:37 PM
X
In reply to:
Gary Marcus
@GaryMarcus
·
296d
@AlexanderNaume2 “horse riding astronaut” is not solved in general form; i just tried this slight variation and Grok was 0 for 4.
Gary Marcus
@GaryMarcus
·
296d
Reasoning, solved! @grok @MarioNawfal
Alexander Naumenko
@AlexanderNaume2
·
296d
@GaryMarcus I found this one

and this
Reve
@reveimage
·
315d
✅ Horse riding an astronaut
✅ Clock reading 5:30
✅ Full to the brim glass of wine
Ethan Mollick
@emollick
·
315d
Even with multimodal image generation, ChatGPT only gets 1 out of 3 of the hard challenges of image creation:
✅Horse riding an astronaut
❌Clock reading 5:30
❌Full to the brim glass of wine

Presumably, the training image data is mostly not-full glasses and clocks set to 10:10
Gary Marcus
@GaryMarcus
·
296d
@AlexanderNaume2 did they show some new drawings? i am not seeing any actual data of any sort whatsoever. what is the actual task that they showed alleged improvements on?
Alexander Naumenko
@AlexanderNaume2
·
296d
@GaryMarcus I haven't seen anything written, it's only my speculation. But how does that hypothesis compare to the possibility that transformers suddenly produced dozens of those drawings?
Gary Marcus
@GaryMarcus
·
296d
@AlexanderNaume2 did they really do that? is there something written about what they did?
Alexander Naumenko
@AlexanderNaume2
·
296d
@GaryMarcus I was impressed that they asked people to draw astronauts riding horses specifically to train their transformers! Forgetting that it is easy to come up with no elephants on a beach and so on.

The statement and images are part of a humorous and lighthearted conversation about AI-generated images. The tone is playful, focusing on the creative and sometimes absurd outputs of image generation models. It does not engage in substantive public discourse or address societal issues.