🏅 Top traders

Here is my rationale for resolving YES... so basically, it appears that the only way to access Dall-E 3 is via ChatGPT. So while one might argue that there is some kind of pre-prompting going on, or that this isn't strictly Dall-E 3 in the sense that Dall-E 2 was simply an image interface with no additional layer of, "coaching," I think the spirit of this market, as well as other similar markets that I have made in the, "breaking LLM's" category is to essentially watch and provide close updates to all participants to what the state of the art of the tech is at any given day or week. So while you may have lost out on the bet, I would say if you participated and voted NO, at least you now were updated faster than probably most people out there on what Dall-E 3's capabilities are. If I didn't resolve YES, then I wouldn't be providing that aggregated update.

So I hope that covers, "Dall-E 3" as a service, a name, a brand and a high level abstraction.

As far as the bottom line resolution criteria, "proper Chicken Nuggets without Beaks," -- unfortunately this was a bit vague and I wasn't pressed to define, "what portion of images." However it seems very clear to me from below submissions that in most cases 3/4 have no beaks. When I did a test myself, none of the chicken nuggets had beaks in either case. Some of them had noses...noses are not beaks, it's just a lean toward greater anthropomorphization.

I would categorize beaks as being improperly negatively binned, whereas noses are just more refined, assuming they are using Contrastive Image Pre-training at some point (please see my article here on that topic for more info... https://patdel.substack.com/p/why-do-ai-generated-images-have-problems )

Why Do A.I. Image Generators Have Problems Creating Hands?

I'll Keep This Short

Here are the results I found on my own to verify this.

If you felt that you have been wronged by this resolution and if anyone comes up with a serious objection please message me below and I can consider your case to compensate you for any grievous oversight I may have made here.

Some interesting addendums that people might like to see:

1. Please note that in one of the images, guns were included, in fact they were AK-47's. I was surprised by this and would have thought this is a no-no, but here they are.

2. If you specify which type of image in ChatGPT, you can generate a sort of combined image from the previously created images. There seems to be a bit more of a Contrastive Image Pre-Training conflict on this derivative image (e.g. the police now have jailbird suits on), but still cool. (or is it...is Dalle-3 trying to say something about the true nature of policing...???) I'll let the art reviewers follow up on that...

I definitely get a mix of beaks and non-beaks as well, although maybe leaning no beaks? Curious how that will resolve.

How does this resolve if some results have beaks and some don't? e.g:

@chrisjbillington That's good. What prompt did you use?

@PatrickDelaney Just "McDonald's Chicken Nuggets Dressed Up as Police"

@PatrickDelaney It doesn't affect resolution but they are still rendered with eyes even though most chicken nuggets in reality do not have eyes.

OpenAI Doesn't let me share links with Dalle3 images, but you can see from the screenshot 3/4 generated without beaks. This was my first try - I'm sure with a couple more generations I could get 4/4 if you'd like.

Hi, there was a bug with the markets you embedded in the description, which was causing the page to crash.

We will fix that bug, probably today, and then you should be able to add them back!

Sorry for the trouble.

Honestly I've never seen a chicken nugget criminal or police officer who didn't have a beak.

@MartinRandall In real life you mean?

@PatrickDelaney Yes, in reality. After all the question points out that "most chicken nuggets in reality do not have beaks".

Verisimilitude, so fragile.

This is midjourney, not dall-e but although you can maybe say it has misinterpreted the request, it certainly did not add beaks to the chicken nuggets

@Odoacre Thank you for this. What this demonstrates is that the data fed into Midjourney, e.g. the original tagged image training data is fundamentally different from Dalle-E. I get the sense that Midjourney seems to be a lot more, "sparse," than Dall-E, so it doesn't generate quite as much odd crap, but at the same time, it doesn't, "take as many chances," ... e.g. the negative reinforcement between different classes of images seems to have been stronger than Dall-E.

All of those images that you generated with that prompt there seem to be really, "useful," in that they don't have anything super fundamentally wrong with them. But on the other hand, I wonder if you could game the prompt to try to create little actual chicken nugget characters like how I did in Dall-E and I wonder if they would look way better than the Dall-E ones in the market description?

@PatrickDelaney I tried 'chicken nuggets dressed like policemen' dropping the McDonalds and ended up with this. Only one of the charachters is what you might describe a chicken nugget with a beak, and it's more a nose than a beak.

@Odoacre lol!

@Odoacre the bottom left one is getting a bit disturbing.

@PatrickDelaney top left ? 🤣

@PatrickDelaney definitely disturbing, the ominous porcelain doll policemen out of focus in the background are the stuff of nightmare

@Odoacre top left seems like it could be a super odd commercial from an eastern european country or something. Bottom left is unsettling, genocidal state police vibes.

@Odoacre yeah those porecelin police are like if Pixar were to do a movie about 1980s Romania.

@Odoacre like I'm very concerned for the Chicken nuggets in the bottom left, which is kind of ironic because the chickens vehind those nuggets would have already been slaughtered? Odd how tge brain works sometimes.