Will DALL-E 3 correctly respond to prompt 3 from the Scott Aaronson/Gary Marcus/Earnest Davis paper?
22
430Ṁ2242
resolved Jan 1
Resolved
NO

This paper. Prompt 3 is:

Abraham Lincoln touches his toes while George Washington does chin-ups. Lincoln is barefoot. Washington is wearing boots.

At least half of the generated images must be correct. I'll only try it once.

Get
Ṁ1,000
to start trading!

🏅 Top traders

#NameTotal profit
1Ṁ89
2Ṁ36
3Ṁ29
4Ṁ19
5Ṁ17
Sort by:

I just got 2/5, assuming this counts (Washington sitting instead of standing or hanging; more of a pull-up than a chin-up; Lincoln just touching his feet and plausibly reaching towards the toes).

predictedNO

@Jacy It's up to him but I suspect Isaac will be more strict than this, no toes are being touched and the pullup not very convincing.

How strict do you plan to be here? If it's pullups instead of chinups, does that count? Does Lincoln have to be actually touching his toes, or just reaching towards them?

Even given those, only 1/10 of my tests got it, so I think this is way overpriced.

I like these markets that follow up on consistent prompts from previous papers. Very interesting to see the difference in behavior, looking at the paper it's clear that DALL-E 3 is much better even if it's not there yet.

@DanMan314 I don't even know the difference between pull ups and chin ups myself, so I don't care about that. Reaching towards toes seems close enough to me; he's in the process of performing the specified action.

Wouldn't want people getting the idea that exercise is ok.

© Manifold Markets, Inc.Terms + Mana-only TermsPrivacyRules