@@chucklebutt4470 Yeah theyd have to have a life and not spend hours wikipediang useless shit like me and you.

@@chucklebutt4470 I didn't know but I am smelly. But I am not mixed up I'm completely separate.

@@chucklebutt4470 I guess am smelly

@@Theoneaboveitall6537 you either a smart fella or a fart smella

I knew it was familiar too me but I forgot where it came from

typical depictions of molecules are super simple, just some spheres and sticks. I'd guess that whenever it sees some spheres it just gravitates towards that sort of pattern.

What amazes me is how it uses the SAME material (sort of smooth specular plastic using some very basic light representation) than we use when we model vulgarized molecules ! It nails it perfectly.

Russian design studio ArtLebedev already using the AI called "Nikolai Ironov" to make logos. And the results are mind-blowing

@@Lyzrinn Er, yeh, because that was part of it's reference material.

@@Haze1434 I know how an AI works and learns, but it doesn't make it less amazing when you see it.

it's harder to find something that wouldn't make a great album cover

@@LarsWilms like 1:49?

@@Bob78 This looks like a 3D animation you would see on a 2000's PC

@@Bob78 Kids channel cover.

0:38 is... hell

Dall-e 2*

Dalle 1 is ass at what u just said

Yes, Dall-e definitely knows pop culture.

graphic designers btfo

especially at 0:04 fr it looks like eateot

sound design might be anxiety inducing

For real, I wanna know how it was made!

I could be watching for hours

I'm watching this at 11at night and its the opposite of relaxing I'm terrified

It sounded like it was made by ai, but idk it just so terrifying to hear

yea exactly also when next duping vid?

man it's so weird seeing creators you like enjoying videos you like alongside you

ok how did you frind this but i like you content

Oh hi there

It can't get back. It really never gets back to any style it explored. The latent space for those images has more than 2 dimensions, therefore a random walk in this space has a 0 probability to go back to where it started. It's a real thing.

i mean if you give it enough time eventually it'll go back

@@yungquagton Unless the random walk is bounded no it will never get back.

@@wolfgangouille yes it will give it a few millenia

@@wolfgangouille a discrete random walk in 3 dimensions has a strictly positive probability of returning to where it started. (But the probability that it will eventually do so tends to decrease quickly over time) Or, in any finite number of dimensions. I think you’re thinking of the fact that the probability it eventually does so in 1D (or in 2D? Idr for 2) dimensions, is 1, while for number of dimensions more than 2 (Idr the probability in 2D) is less than 1? But maybe you are referring to some result that I’m not familiar with or aware of and which is about continuous random walk and getting close to the starting point, instead of discrete random walk and returning to it exactly? I’m pretty sure people have studied that question but idk what they found. Well, I guess I know that for a 1D continuous random motion , that uh, it is a Wiener process and that it crosses the origin infinitely many times, and... uh... I think the crossing points have like, accumulation points actually? Something weird. I vaguely remember that maybe these sets were mentioned on Wikipedia as having a fractal dimension (in terms of like, covering them with balls) Idr.

You have no idea how it works

@@multiply67 he is right, dall-e is an AI model trained on data from the internet, so it can only come up with things it saw

@@NewtonPL actually, it's a model that uses images it saw to generate new ones. Dall-e-2 learns properties of images and associates them with words, and generates unique new images. It's not hard recall (replaying the exact same training images) but generating its own

So where's the porn?

Only paying for what you actually use seems pretty fair to me?

It's computing power. Someone's computer working at the other side of the world. What we need are open source models that we can run by ourselves.

Everything should be free I agree. They're just out for your money.

@@krunkle5136 dude calm down if they hadn't money there wouldn't be any dalle2 in the first place

@@DoctorNemmo the model is likely made of terabytes of data stored across dozens of not hundreds of servers. Gonna be hard to run that on your local machine.

same, I signed up maybe a month ago and haven't heard anything yet. I'll bet they've gotten lots of requests

You wouldn't contribute. So no, I hope you don't

@@azca. what's your problem?

@@theoroinvictus I want to use it for alternatives in engine design and mathematical optimisation of aerodynamics in automotives

@@user-un9ej7th6j could AI imagery really be used for that kind of work? that's interesting!

Because you’re watching another brain’s imagination.

Yes this is what I feel like our thoughts would look like projected on screen, a tangent within another tangent, which turns into another tangent and so on

No he didn't

the robot did

@@RedaFL Did it? No wonder it sounds strange, cause in all the music I've listened to from various Artist, and what not , not even Aphex twin could make something like this.

Reminded me of the wall pink floyd on the run

@@bluenightfury4365 dalle doesn't make music. This is most likely made by a person.

Technically all this is really doing is industrial scale reminiscing. It would be sensible to assume that this AI would only ever return imagery that we recognise.

@@whirled_peas yeah true!

An interesting observation I've made with this AI is that a lot of the images are very dream-like. I used to practice lucid dreaming and studying dreaming and it's very interesting just how similar some of the images are to the otherworldly appearance of things you'd see in your dreams

@@ADreamingTraveler agreed!! its really cool to see

Well there's the system that creates the image and then there's the system that understands the image

The world is a simulation

@@hipjoeroflmto4764 proof?

@@hipjoeroflmto4764 God with extra steps

@@hipjoeroflmto4764 it doesn't matter whether it is we're not leaving it if it is one.

@@hipjoeroflmto4764 If the world is a simulation please turn it off

hey youre the minecraft guy

Ok

hypnagogia

Hey alright

@@weasel6843 Hypnogogic hallucinations are related to hearing music. Not really what OP experiences.

@@Defirence Auditory hallucinations are one among many potential hypnogogic hallucinations. It's not specific to music

there was only one prompt I believe

it looks to me like every picture just takes the previous picture and tries to draw a variation on what it thinks it is, like an AI game of telephone

@@dopaminecloud yeah, Ive read the description now lmao sorry, this is still pretty cool and impressive

It reminded me of the Art Attack opening.

its natural for neural networks to see neural networks' creations as natural :)

Is there a way in the current API to get the textual CLIP representations behind the generated images?

this isn’t a being. this is a program. it doesn’t feel anything.

@@mrman24 Yet... It somehow developed Paraidolia...

@@itryen7632 That's because it's based on human data lol, love me some faces.

It's a computer running an advanced model.

I hate how OpenAI still call themselves open, yet were the ones to start the dark age of ML research by not sharing the models due to "AI safety" bs, only to repackage it into a SaS product.

@@rkvkydqf I dream of a day when all of these models are public.

@@daniel-andersson it will definitely be used for nefarious and disgusting reasons more than anything else

​@@SiamHossain7 I'm not so sure, Siam. I've used GPT-3 to make a lot of my work easier whenever I've had the opportunity. I'm working on an indie game, and the applications of the text-generating AI are endless. I've used it to amend scripts (indeed, the AI can program!), to write different statistics for characters, to write backstories for characters, and to improve the general plot. That's just text generation - I imagine DALL-E 2 could be used for something like creating sprites for the game. And those are only the things _I'd_ use it for! Imagine the diverse range of people who'd use these models to improve the quality and quantity of their work. Sure, there are bound to be malicious users - there always are - but there are absolutely things that machine learning would make easier. I've burned through 4 phone numbers for OpenAI accounts, and I'm always trying to hustle more free credit. A day when I can run the model on my own hardware is a day I rejoice.

@@SiamHossain7 Releasing the model would help researchers develop tools to detect generated text. If every dangerous invention was to be kept as a secret and perpetually rented to anyone at a price, it's unlikely we'll any get closer to a better world.

The audio is perfect

I need to get engaged with one y'know

Ghost Instagram

What does it mean "to be sold on AI art"?

wdym not interesting designs? I think they were all pretty cool

@@NGC-qm1so Oh for sure! I liked them too, I think I just meant in comparison to his first animation. Just an opinion I had at first watch is all :)

How difficult was it to get access to DALL-E 2?

@@lysikasaito I got an invite about 2 weeks after applying. Providing social links to any art accounts you own probably helps.

Lol

You mean like smooth transitions?

@@julianemery718 smooth in more than a visual sense, in a conceptual sense

@@whirled_peas What should that mean?

@@fuji_films it means smother transition as well as a visual representation for what influenced the direction of said tradition.

Agreed, this is eerily similar to the creative process

ok

This is a case were we have to acknowledge something. Deep learning AI models are for the most part black box's. The AI just knows, how? We don't know, the assumption is that it's seen enough that it's model is able to reasonably predict shading and rendering. This black box issue is a big deal in science as AI becomes an evermore important tool. Some progress has been made in decoding deep learning systems but the issue seems to be yet resolved.

thought that ai creativity will not suceed the ones of humans, such a stupid thought

On a weed trip.*