OpenAI Debuts Next Version of Its Image Generation Tool

OpenAI is offering an early look at DALL-E 3, the next version of its image generation tool. From a report: The update allows DALL-E 3 to be summoned and controlled using ChatGPT and aims to produce higher-quality images that more faithfully reflect queries. OpenAI says that DALL-E 3 is significantly better at understanding the intent of prompts, particularly longer ones, compared to DALL-E 2, which debuted in April 2022. OpenAI says DALL-E 3 also does better -- but not perfectly -- in areas that have tripped up image generators, such as text and hands.

The ChatGPT integration will allow people to hone their request through conversations with the chatbot and receive the result directly within the chat app. OpenAI plans to make DALL-E 3 available to ChatGPT+ and enterprise customers in October DALL-E 3 will also be available sometime this fall in its public labs and for API customers.

Neat

[DarkRookie2]

I am continued to be unimpressed by all this AI and its supposed benefits to my life.
I see nothing more than a scam to run down batteries and wear out hardware.

Re:

[DarkRookie2]

Like what am I getting with this?
I can make weird pictures. I can do that in paint.
I don't work in a creative field and I have no way of printing a poster size something I like.
Maybe it will make the artists life easier. Won't mine. The art will still cost the same or even go up in price since you can add the AI buzzword.

Re: Neat

[beelsebob]

Youâ(TM)re getting better quality photos on your phone; systems that can spot bugs in your code; the ability to search your images; the ability to select text and copy it out of images; better internet search results; more accurate diagnoses from your medical images; rapid development of new drugs thanks to protein folding that mostly works; improved materials with properties we can choose; â¦

This is a bit like space, everyone always says âoewhat do I get from sending rockets to spaceâ while ignoring the amount of communication that is space based, and GPS, and satellite maps, and â¦

Re: Neat

[LazarusQLong]

thank you for this. Many people look at a technology and say, "Meh.", when in reality, it is these incremental increases that (eventually) lead to big changes. James Burke famously did three series wherein he traced simple technologies hundreds of years ago through to how they effected everyday life, Now.

ref: https://en.wikipedia.org/wiki/... [wikipedia.org]

Re:

[DarkRookie2]

I am sorry I don't see the appeal of something that I will never be able to use or make because I don't have the resources to do so.

Re: Neat

[drinkypoo]

You don't have a middle of the road gaming PC? That's all it takes to run stable diffusion for example.

Re:

[DarkRookie2]

Above average. But that is only part of the resources I lack.
Again what am I suppose to do with it?

Re: Neat

[drinkypoo]

Well if you don't also have imagination, nothing I guess.

I'm largely aphantasic so I have a hard time even picturing what I want to create. But the software has no such problem. If I imagine it, it can render it, or at least something like it that lets me zero in on what I want to see.

I experimented with getting chatgpt to write some poetry but it can't follow simple directions so that didn't help much. I wasn't really looking for artistic merit, only mockery, and it couldn't even manage that. So for me the

Re:

[DarkRookie2]

Exactly. Which why it annoys me greatly that it is everywhere being shoved into my face.
It is not a tool I am built to use.

Re:

[drinkypoo]

Just imagine how people in wheelchairs feel about stairs.

Re:

[DarkRookie2]

Again. I don't code. What is the point of asking it a coding question.
Again. I don't take pictures. What is the point of picture correcting software.

It does nothing for me.
I can see the point of going faster. I don't see the point of shoving a creative tool into my face and forcing me to use it.
There is a reason I am not an artist, writer, or photographer. I am not good at them and practice doesn't help.

Re:

[Talon0ne]

OK. Maybe AI isn't for you. You have yourself totally convinced.

Re:

[b0s0z0ku]

A 'puker pretending to be a real human adding detail that doesn't actually exist in the photo isn't "quality improvement" - it's pollution. Ability to do (say) facial recognition accurately isn't a benefit, it's actually a bad thing.

Re:

[LazarusQLong]

did you actually read what beelesbob wrote? Or are you replying here to the voices in your head?

Re:

[DarkRookie2]

I don't take pictures.
I don't code.
I don't save images.
Neat. Where it this. This I could use.
The only reason they would get better is that they will stop making it worse. Not that some new toy came along.
Doctor is too expensive to go to anyways.
Meds are to expensive to use anyways.
Cool. Bet my next computer case is made with steel and plastics.

So a bunch of worthless to me or too expensive for me to ever use. Just like space. Most of the tech is worthless to me. Kinda neat to see pictures of the Ea

Re:

[Anonymous Coward]

No, This is what AI gives me:

Warnings and bans on social media because AI thought a bill of materials was an explosive device, or that talk about an action request meant talk about an Armalite rifle, and thus a suspension or even a ban, with zero change of review.

Chatbots which run in circles and don't even have any type of useful solution, and forced me to go to a lawyer to cancel a subscription by serving the other party.

False claims that a picture has nudity or violence because it saw too many peach colo

Re:

[Otis B. Dilroy III]

Apparently you don't like transsexual porn as much as the rest of the internet does.

Re:

[gweihir]

Actually, I have seen two legitimate applications so far. One is creating lewd pictures with significantly less effort than a professional artist would need, but with some effort needed in separating out the good ones from the not so good and/or uncanny ones and the outright bad ones. Examples are Starfield lewd mods and the speed with which such pictures were created. Second example I have seen is automated translation (DeepL), where I personally estimate that you save about 66% effort or so, but the resul

Re:

[Tablizer]

> I see nothing more than a scam to run down batteries and wear out hardware.

If you can't beat 'em join 'em: get into the battery and hardware business.

When the next pandemic or civil war comes, I'm buying undertaker stocks.

I'm gonna stop being curmudgeon complaining about how stupid humans are, and get rich off of mass debauchery instead.

AI Killer App?

[RobinH]

I spent a bunch of time playing around with AI, and while I did enjoy a more conversational search style in Bing, and had a ton of fun with image generation, the AI generated content that's starting to show up on YouTube is just awful. I'm not a high school student writing essays, and my job doesn't entail writing a bunch of BS emails, so I'm still waiting for that killer app. I mean, if I could walk up to a vending machine and say "tea, earl grey, hot" and it made me a decent cup of tea, that'd be a star

Re:

[DarkRookie2]

It would be faster just to press the tea button.

Re:

[gweihir]

Indeed. And you would not get the occasional milkshake or tomato juice with that button either.

Re:

[HiThere]

WRT the vending machine "tea" that I've encountered, even calling it tea requires a bit of faith. Calling it some particular variety of tea is unreasonable. It's actually worse than most of the coffees.

Re:

[RobinH]

Well I presume it would be a machine with a huge number of possible recipes, like a replicator, but maybe we need to focus on building that technology first. :)

Re:

[gweihir]

My impression is that all the current wave of AI hype applications can do is simple things with a somewhat reasonable hit rate and sometimes a complete failure. It cannot do a bit more advanced things in many cases, and worse it often starts to hallucinate for a tiny bit more advanced stuff, often requiring an expert to spot the problems. This makes for very nice demos (as the hype nicely proofs), but rather limited usefulness as a professional tool.

Re:

[Wolfrider]

> the AI generated content that's starting to show up on YouTube is just awful

Bruh, you said it. Something approaching a majority of 'tube video thumbnails are getting to be obviously AI-generated and it's really annoying. More than half the time the thumbnail has nothing to do with the video, and is not even featured in it when you watch. Basically false advertising.

Better lewd pictures?

[gweihir]

Because that is frankly the only credible application I have seen so far: Lewd pictures with varying degree of nudity, where accuracy does not matter at all. Not that this is a bad application in any way, it is just means the tech is not nearly as spectacular and fundamental as some people seem to think.

Re:

[CAIMLAS]

That's the axis on which all new technology is judged, though: can it help me get off faster/better/harder?

Almost every useful internet technology to date has seen its earliest application -adoption- being the use of exchanging smut. That's what drives things to prominence.

I have no idea where we go from there, but I'm sure things weren't quite so clear in the early days of BBS what you'd do with a computer's EGA graphics card, either...

Re:

[gweihir]

Well, this means the tech will stay around and that actually is a new thing for AI approaches. It does not mean it will go far beyond this. I did forget the second use I found and that is LLMs can reduce the effort for making translations to some degree, because an expert probably need only a third of the time to fix an AI translation than doing it completely would have taken (my own estimation based on one case where I did that).

In both cases those are niche and not high volume applications and they in no

Re:

[Tablizer]

6 finger'd hands are considered a mistake, but 2 wankers or 3 tits is a selling point.

Re:

[SoftwareArtist]

I know you're just trolling, but seriously? That's the only credible application? At least try to make your trolling sound plausible.

People creating websites, who now can generate professional looking artwork for it without having to hire a professional artist.

Small business owners creating logos for their companies that look far more professional than anything they could have created before.

Amateur game developers who can use it to create good looking art for their games.

People with ideas for graphic nov

Re:

[gweihir]

I am not trolling. I just have a much better appreciation of reality than you do. Obviously, other pictures of the same, not very impressive, quality level also qualify, but they _still_ need to be checked over and selected by an actual expert.

Re:

[SoftwareArtist]

I am not trolling. I just have a much better appreciation of reality than you do.

LOL. Here's a hint for you. People who aren't trolling don't say things like that. They respond with substantive arguments. Trolls, on the other hand, ignore logical arguments and just say, "I'm smarter than you!"

Re:

[gweihir]

Nope. Because _you_ are trolling. Apparently you are unaware of that.

Re:

[iMadeGhostzilla]

I do not want to waste time looking at art that didn't take blood, sweat and tears to make by the artists who wants me to look at it.

Besides the principle, all AI art I have seen so far has a certain lifeless -- dead -- quality to it that makes it sometimes fascinating and almost always unpleasant to look at. I expect people will soon tire of it, and many have already. The best outcome may be that the human generated stuff that looks AI-cheap will, I expect, also be rejected. Content producers will have to

A fist full of dolls

[Tablizer]

> also does better -- but not perfectly -- in areas that have tripped up image generators, such as text and hands.

So instead of 6 fingers, a hand will have 5.5.

Conversations?

[nospam007]

"The ChatGPT integration will allow people to hone their request through conversations with the chatbot and receive the result directly within the chat app. "

The picture is OK but it really needs more cowbell.
And the tits should be wayyyyy bigger.

I’m looking forward to using it.

[WDot]

It looks like DALL-E 3 has some high quality images, but integrating it into ChatGPT is the thing that will make it compelling. There are people (including me) who already subscribe to ChatGPT Plus. I’ve thought about subscribing to Midjourney, but the friction between joining the Discord (which is their interface for generating images) and signing up for an account usually keeps me from doing it.

People who are wondering why they would want AI generated images don’t have much imagination. If