Will any image-generation AI be able to consistently draw simple polygons by the end of 2024?
Will any image-generation AI be able to consistently draw simple polygons by the end of 2024?
35
1kṀ2658
resolved Mar 29
Resolved
NO

My attempts to get DALL-E 2 to draw very simple shapes went... poorly.

(https://outsidetheasylum.blog/testing-dall-e-2-mathematics-comprehension/)

When this market closes, I'll test out the most advanced image-generation models that I was able to get access to. If there are multiple, I'll try them all.

If any of them can consistently return a pentagon from the description "a pentagon", a heptagon from the description "a heptagon", and similar, I'll resolve to YES. If they need a bit of nudging like specific prompt wording but can still generalize correctly to any polygon, that'll be good enough to resolve YES. Otherwise I'll resolve NO.

A model that was trained specifically to make geometric shapes doesn't count; it has to be a generalist like DALL-E 2. In order for a new model to qualify for this market, it needs to be no worse than DALL-E 2 at the vast majority of things it's asked to draw.

Get
Ṁ1,000
to start trading!

🏅 Top traders

#NameTotal profit
1Ṁ298
2Ṁ194
3Ṁ134
4Ṁ117
5Ṁ116


Sort by:
2mo

@IsaacKing Time to resolve I think?

5mo

It will be interesting to see how inference-time scaling methods improve the ability of generative models to solve these kinds of problems

5mo

@IsaacKing is there a version of this question for 2025?

I'm thinking of writing another question that's specifically about inference-time scaling for generating images with correct geometric shapes. But I like this question and it's a bit more general

bought Ṁ50 NO9mo

Arbitrage (?)

@IsaacKing is the EOY for this market 2023 or 2024? The title says "end of 2024", but the market itself is marked "Jan 1". I'm making a bet only if it's the end of 2023

1y

@LukaChrelashvili Whoops, sorry about that. I think more people are likely to go by the title rather than end date, so I've changed the end date to match the title.

DALL-E 3 can't create a pentagon or any convex polygon other than a square and a triangle. Even if I specify or nudge the prompts by saying "shape with 5 sides", the resulting image almost always shows a hexagon.

Shape words are very hard. Be more explicit here. Most babies cannot understand more sophisticated shape words.

2y

@EnopoletusHarding What do I need to be more explicit about?

@IsaacKing There is no basic difference between machine and human intelligence. You would not ask a one year old child to draw you a "pentagon", whatever that is. Ask it to draw a form with five sides or whatever words will be appropriate.

2y

@EnopoletusHarding I tried that too, didn't work.

predictedYES 2y

@IsaacKing Can't count to five iirc

What is this?

What is Manifold?
Manifold is the world's largest social prediction market.
Get accurate real-time odds on politics, tech, sports, and more.
Or create your own play-money betting market on any question you care about.
Are our predictions accurate?
Yes! Manifold is very well calibrated, with forecasts on average within 4 percentage points of the true probability. Our probabilities are created by users buying and selling shares of a market.
In the 2022 US midterm elections, we outperformed all other prediction market platforms and were in line with FiveThirtyEight’s performance. Many people who don't like betting still use Manifold to get reliable news.
ṀWhy use play money?
Mana (Ṁ) is the play-money currency used to bet on Manifold. It cannot be converted to cash. All users start with Ṁ1,000 for free.
Play money means it's much easier for anyone anywhere in the world to get started and try out forecasting without any risk. It also means there's more freedom to create and bet on any type of question.
© Manifold Markets, Inc.TermsPrivacy