Gary Marcus

Rank 14 of 47
|
Score 89
10/5/2025, 11:52:29 AM
X
In reply to:
Mqrius
@Mqrius
·
2d
@analisereal @GaryMarcus @elder_plinius I think what Gary means is that models are sometimes optimized for things that people have talked about, so they can claim the new version works well. But it's possibly fake, in that a slight variation would break it again. So let's try a variation? eg an ostrich riding a diver.
Análise Real
@analisereal
·
2d
@GaryMarcus @elder_plinius "Not that impressive..." The video above was produced from simply a prompt saying "a horse riding an astronaut." Of course this is impressive. This is getting ridiculous Gary. Here's your other example from your substack post. This is of course impressive.
Gary Marcus
@GaryMarcus
·
2d
@elder_plinius as i noted last year, getting a 3 year old example that’s been well-publicized isn’t that impressive. if you can do it robustly across variations that would be more impressive.

see also https://garymarcus.substack.com/…
Pliny the Liberator 🐉󠅫󠄼󠄿󠅆󠄵󠄐󠅀󠄼󠄹󠄾󠅉󠅭
@elder_plinius
·
3d
WE DID IT, CHAT 🤗

"a horse riding an astronaut"

believe it or not, getting even a still image of a horse properly piggybacking an astronaut used to be nearly impossible with gen AI *checks notes* last year

now, we have realistic low-gravity horse-carrying physics and

The statement is part of a discussion on the capabilities and limitations of AI models in generating images. It questions the authenticity and robustness of AI-generated content, suggesting possible data augmentation. The tone is analytical and skeptical, aiming to promote a deeper understanding of AI's current capabilities.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does no harm and is focused on a technical critique. [+1]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    It promotes understanding by questioning the robustness of AI models. [+1]
  3. 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 without personal attacks. [+1]