Gary Marcus

Rank 18 of 47

Score 94

@pretendsmarts The issue has never been that you can’t get the prompt right, but that the system isn’t reliable. I look forward to trying 4o images when they are available to fee customers as advertised.

See my original Horse Rides Astronaut essay.

3/26/2025, 10:49:31 AM

In reply to:

Andy Wojcicki

@pretendsmarts

43d

@GaryMarcus The new model doesn't fall for this.
This looks like Dalle3, but the strange part is that this usually goes through a prompt re-write, unless you specify to use unmodified input. The LLM part should correct it.
Even our own modest pipeline , gets 1/4 with no elephant (just the

Gary Marcus

@GaryMarcus

43d

“Can you draw a photorealistic beach with no elephants?”

The statement critiques the reliability of a system, specifically in generating images as expected. It expresses anticipation for improvements and references a previous work for context. The tone is constructive, focusing on system performance rather than personal attacks.

Principle 1:
I will strive to do no harm with my words and actions.
The statement does no harm and focuses on system reliability rather than personal criticism. [+1]
Principle 4:
I will engage in constructive criticism and dialogue with those in disagreement and will not engage in personal attacks or ad hominem arguments.
Engages in constructive criticism by addressing system issues and expressing hope for improvement. [+1]