What a time to be alive!

I don’t like your optimism on the subject. AI has world ending consequences

Lol

He said it! He said the thing!

🤚

This really is just a cameo appearance isn't it

Artist: "Well, at least I have a job that ain't replaceable by any machine" AI: "Hold my bit"

@Lui Z The absolute stupidest take I’ve ever heard.

@Ougin Toga That's exactly what people would say when cameras first came around. What is there even to do, you just push a button and get a picture!

@viinisaari "artists" gonna be who are able to feed the most interesting concepts to ai and tweak it a bit.

Artists won't be replaced, art will just change. Similarly to photorealistic painters who were replaced by photographers.

Youtuber: My job is secure because I do things an AI cannot. AI: Hold my ensemble of models...

Ahah we're all watching Two Minute Papers. These new AI stuff are gonna help us make concepts faster. But wait for the day when Tiktok is flooded with viral content generated with AI only

ah, created by AI, distributed by Ai, made for humans, they way god intended....

@li'l T ahah right… but I’m afraid the tiktok model is like the future of all platforms now. Youtube is already showing my videos in a tiktok-like interface, and i can’t change it…

Who gives a shit to what happens to tiktok lol

Tiktok will be dead..it’s just a fad

Imagine a tool that can make storyboards, in wich you detail things like characters, angle, height of the camera , movement, action and the AI generates sketchs of the scene

@AKumar528 Matrix moment

😂 i see it creating an entire movie at this point

@Ulysse but can it give you variation of sketches in mere minutes even seconds?

Maybe we were living in such a simulation. Imagine what a planet sized computer could do

wow ur idea is revolutionary, u could pretty much write a whole manga, anime, story, movie, etc.

Humans: “A picture is worth a thousand words.” DALL•E: “I can make that in less than 20.”

@Muhannad Elmansuri tiger in atlantis*

@Josiah Ray whats wrong

@Muhannad Elmansuri ...😐

@MASOL Not code, words, as you would be explaining it to a human. Plus there is so much going on in the photo, that you could easily spend two pages to describe absolutely everything including colors and layout of every single thing featured in the pic.

@Muhannad Elmansuri you mean code

As impressive as it is scary. Though in fairness the ability to type something in and have an image appear in front of you probably would scare artists in 90s too. But Google Images is already here and leveraged _by_ artists to create new ideas. Maybe Dall-e would would be used similarly. Artists could start projects by generating ideas with Dall-e, then improve them manually. For a lot of people the dall-e result will be enough, but others will want further customizations.

@Novan Byworks The long-term benefit to the art world that will come of this is on the consumer-side of things, not the artist-side of things. Not sure if I like that but it's going to be interesting.

@doja Except that human beings are the perceiving consciousness. You will only be obsolete if you believe that your sole purpose on this planet is to do stuff, not to look and appreciate the beauty in the world. We are human beings, not human doings.

Ai image generation was the only reason I got back into commission work. Trying to take people's vague descriptions of what they want and turn it into a real image was the most frustrating part of commission painting. But just typing their requests into an ai and getting (often very loose) suggestions for composition cuts out a lot of time getting started.

@Novan Byworks The thing is this doesn’t turn everything into a jumbled mess. It can literally paint in any art style. It can imitate Salvador Dali, or Picasso, or Van Gogh. It can just generate anything. You could say “rubber duckies floating in a bathtub in the style of Salvador Dali” and it would come out looking perfect. It’s not just realistic photos….and this is only the beginning. The same is likely going to happen for music and some other art forms that don’t involve a human appearing in it (like dance or movies). But eventually this will evolve into completely computer generated movies…and they will be photorealistic and better than any human film ever made. Then where does that lead? Meta verse/virtual reality simulations that are indistinguishable from the real world. It’s exactly like The Matrix and other science fiction stories have predicted. The AI will eventually be able to create anything and do everything better and faster than us. Humans will just become obsolete.

In the future you’ll just have to write the movie script. When you’re done you have the movie. In fact you can watch it as you write it. And change it in real time to try different things. Imagine how Hollywood will be transformed.

It's scary to be honest, programmers won't be needed, you ask AI to create an application for you etc. This was discussed on the Official Podcast, if the perfect world where AI and Robots completely took over every single human job, and currency didn't exist would be 'perfect'

Or it never get released to the public

That would be amazing for amateur filmmakers to get ideas from. But it wouldn’t work as an actual A.I movie making machine such as the replies here are suggesting. If that was the case it would be boring and soulless. But then again, modern disney and marvel movies are boring and soulless yet people still enjoy them. So i guess to each their own.

omg .... mouthwatering

Well, you've spoken into exsitence. Theres this TV show called Westworld and on the first episode on the fourth season, it is exactly like that how you written it.

this is so insane, its going to change the world in a lot of aspects especially with art.

I have been playing around with AI art a bit and I think the technology will replace a lot of things. I don't think it will replace the culture of art though. I watch a lot of artists on Youtube and the reason I like it is because of the skill. When I watch someone draw a picture perfect image it makes me excited, and that is what gives it value. I don't look at my printer in the same way.

AI should only be used as a source of references to inspire your own art.

@something clever I was wondering where you were going with that lol

@SmudgeOfficialUK I did try to hint at it with the "make another ai to do a voice over."

@something clever ahh with ya. Sorry mate. Hard to tell on youtube these days lol.

@SmudgeOfficialUK I was making a joke.

Amazing stuff, sad that we can't all test it now lol. Definitely seems better for brainstorming rather than finished products, which would helps artists ideate or help clients explain what they want. The images it makes looks more like a first iteration of an idea, pros would usually make more iteration and improvements of those with specific feedback until a final product is achieved. In the case of art though, I think it could potentially replace the generic commercial art like the stuff printed on random cheap items, gift items, t-shirts, mugs, souvenirs etc. It's interesting to wonder about a world like that, reminds me of the anime series Carol & Tuesday that explores a future where AI generates all the mainstream music with people occasionally singing the generated lyrics, and a couple of girls who try to write & create genuine human music in such a world.

Dall E Mini is available to everyone by now

imagine explaining this to someone from the middle ages

Burn

Imagine typing "smoke on the water" and getting 035

They'd say it's work of the devil... and they'd kinda be right.

They've be like, nah, I have thigs to do. Going for a swim in a lake, rode a real horse, go for a walk in the forest, look at a perfectly clear night sky. Now, try to explain that to our kids

To be honest with you, as much as this is pretty impressive, in a few years when anyone can use this ( by another team ) then alot of people are going to be claiming a beautiful masterpiece as there art when in reality they can't draw for anything.

we’re fucked in a thousand ways.

@Nicolas Maldonado the difference is this tool has such a low skill barrier to use that artists are not needed anymore. When photoshop came out and improved everyone thought it would lose artists jobs because it was a tool. The difference is the skill barrier with software is still too high for most clients, this on the other hand is not, any client can type their brief into it a get their result instantly.

We need to completely rewrite copyright laws then. It's something we should have done already, given that current IP laws are so absurd and outdated, but in the future it will be absolutely essential. IMO, I'm in favor of abolishing IP laws completely because they do much more harm than good. They are responsible for most of the wealth inequality out there for obvious reasons (it's literally a monopoly enforced by the state). But a more lax law would be very welcome

It's a tool no different than a brush. You need to learn to use it. You also need to be creative with the input. Like MKBHD was so apple biased with his inputs, it wasn't very appealing to me.

@fried tofus Like artists and photographers never learn new tools and never improve a generated result

Lately I am in love with these kind of video by Marques more than tech product unboxing. Keep it coming.

As a graphic designer I was thinking the same thing before you brought it up... I feel like my job will be obsolete by automation, or something like this. At the very least I would like this tool so that I can get inspiration, or use an image from it and clean it up to be used

here is a prompt to break any AI: A person standing in front of a mirror, holding a mirror. The infinite reflections breaks any AI I've come across.

Marques saying "What a time to be alive" at the end of the video clearly reminds of Two minutes papers channel :D

For an artist it’s a horrific time to be alive.

Probably referential.

which also did a video on dal-e 2 !

look at the pined comment lol

this was definitely a reference!

This almost makes me think that it’s some kind of elaborated late April fools joke. Like I would love to have access to this it looks much more powerful than the normal Dall-e

9:25 It appears to me that the AI hasn't been trained to understand trademarks. I would guess it gets better results if you actually describe what you want instead of trying to say you want specific products. Also, I would have wanted to see the tiger+atlantis results if you had added "in photorealistic style" or "in steampunk style".

I'm guessing they purposely excluded trademarked products.

As far as I know, everything that artificial intelligence has created as such does not have copyright. So don't worry artists, it's not bad yet Personally, I think that DALL-E 2 is really unusual. He creates really fantastic things with just a few words. The beautiful thing is that you can create with this concept art and so on and use it in your work.

This is both amazing and terrifying. As a (mediocre) artist myself, a part of me is sick to my stomach imagining handmade art could, quite soon, largely become a specialized niche rather than a necessity. A few more years into this, I don't see why this AI's understanding of context and style won't become adept enough enough to create whole comics and keyframes for animators. Comprehending storyboard level of consistency seems like the logical next big step to this AI once they've tweaked its image quality to perfection (which it seems they nearly have). This isn't even Open AI's most potentially impressive invention. It also has a API for turning natural language into code. Imagine twenty years from now when making programs and video games is as simple as writing an outline. Between the recent breakthroughs of Open AI and DeepMind, general purpose AI has been showing a lot of power recently. This world is going to be very, very different in a few years.

As a teacher, this would be incredibly useful for presentations and class materials! I spend way too much time browsing and searching for clipart that I desire. DALL·E 2 would be such a great help and time saver!

There's a public one called Dall-e Mini and it's worse than Dall-e 2, but it is a great way to show ideas and you don't need consent for access

@Brian Clark not to billion dollar corporations. Art will be nothing more than a hobby

@Tolietduck84 Pumbleless because art made my humans is more valuable to us still

@Brian Clark no it doesn’t. Why would anyone pay artists when AI can do it better

Why would people bother learning art when it can be done better, instantaneously, by simply typing into a software

Part of me is excited with its potential and part of me is concerned. Does this mean that human creativity will be a non-desired skill in the future? I can see it being a creative "brainstorming" tool at its current state but throw in some more RND and in a few years it could become something entirely different, after all it's only at its infant stage.

"Does this mean that human creativity will be a non-desired skill in the future?" No. It can't read people's minds. It can only make things from your vague descriptions. Only people themselves can bring their ideas to life the way they want. AI isn't useless though. It can be used to produce references to inspire people's own art.

I mean, with humans, you can get some really deep and endless customization of whatever painting you want, the AI still has limitations. Though we are only beginning to rapidly drift to a world that seems more and more like science fiction every year, I'm pretty sure the technology won't be there yet, at least not until 2025.

@PixelPizza77 yeah, I've given it some thought It's not true anymore since AlphaZero. It used to be true in some what recent times...

@autohmae You sure about that? I'm pretty sure that's bull, and if it did happen it must've been a while ago, a computer now is certainly at least as good as a computer and a human

Having this kind of AI refined would be amazing for creating random ideas of things. Like I often get ideas of something like a spaceship with tthree massive engines and 3 compartments for fuel. Having an AI which can help put those random ideas to a visual sense could speed up the creative process of so many different industries.

No offense but the lameness of your example is hilarious :D ...intentional, I suppose.

Game. Changer. Imagine what games you could make with this like guessing the sentence input for an image

Next level. Saw this a while back and was amazed. When we get to point ai can make a vr world will be insane

while this is certainly amazing and mind boggling, a part of me has a sense of doom about AI like this. like, we will eventually get to a place where we no longer can tell what is real or AI anymore, and it will be so pervasive we will eventually stop wondering/caring. in essence, a reality fatigue. i’m learning to appreciate our time a little bit more knowing how much we make with our own hands, artistically and creatively, and as creators this may be our Golden Age. what will our descendants reality be like? kind of scary/sad to think about.

Ha...I'm reminded of that Matrix scene. If a steak looks, smells, and tastes like a steak, does it really matter if it's not real?

@ITSTONYSANTANA that’s not how automation works.

@ITSTONYSANTANA you completely underestimate AI. This is like saying “oh factory made industrial goods will show who the great metal workers are” no. There isn’t a human metal worker alive that can replicate machine made parts, same as this art. Artists will as a whole become obsolete in the next 20 years.

@Remus not really. It’s hard to photoshop realistically, and incredibly hard to photoshop completely original images in a hyper realistic manner. Soon when you see a cute video of a dog or cat online it will probably be AI generated

I am in love with doing AI art!!! I want to get my hands on this!

An easy example to understand what a general AI (often spelled AGI for artificial general intelligence) would mean is that it can start without what a painting is or how one can draw in the first place, but you could throw fully 1000 scanned art books (or PDF files) at it and after it has learned about art and drawings from those books, it could do DALL-E 2 stuff without any extra target specific programming. Now imagine such AI system watching every single tutorial video in Youtube and understanding it all...

@Mikko Rantalainen Yet.

@Jen Kem As far as I know, DeepMind Gate cannot generate images similar to DALL-E 2. It can do any of the 600 tasks it has been trained but I've yet to see any evidence that you could give it collection of art books and then ask it to do something similar to DALL-E 2 or even DALL-E mini.

2 minute papers just posted a video of exactly this. Google's Deepmind AI

I could watch these generations for hours. I want more

I don’t like the question, "Will this Ai take people's jobs." I think that question is incredibly relative. especially when speaking towards Graphic Design. Many companies have protected assets. To feed these through a system that is not in house, would mean you would have to sacrifice some privacy of your portfolio. Not to mention the shear complexity of aligning type and fitting brand standards. As a graphic designer, I see no threat at the moment. As an Illustrator/photographer, yes, the fear is there. That is if this tool is made for personal and commercial use. Even then, I'd assume whatever they would feed into the photo library, they would need explicit permission from the artists...and I don't think many would be happy to have their art appropriated cheaply. You can claim originality to these generated artworks, but it is fed by already existing images. Scary of course, but logistically, I don't see it flying smoothly commercially without the permission of artists.

There is a reason Google owns YouTube with its rights. All the platforms own rights to uploaded content

In other words: Yes, this will totally take away people's job. Though, it might take a couple of years more. As you said, it will get more advance and spit out more sharper text/images. High quality photo's and video's etc etc... For now, i don't think we immediately need to be "worried." As you said, it is a very good concept for things like when you are brainstorming and having images in about 10 sec. to see what could be done. But of course, you would want a specific high quality image in the end for the project. Which it can't do (yet.) But i guess they're very close. The thing is, it isn't available for most of the public. So if it will be mostly for very high level professionals like big corporates, well, let's say we still have a while to go. On the other hand, there are still people who just like to design their own stuff. People like myself who actually do like to design. Of course, when it comes to pro's for their jobs. Time is money. If you can make images in les then a couple of seconds. They won't hesitate to get this stuff and let them replace you for most of the part. They might still have some graphic designers hired to kind of control things i guess. But yeah, what a time to be alive. (In some cases.) I bet that, when they first found out the lamp and the telephone it was also a great time to be alive. I remember having one of the first cellphones and then the smartphone. Now that we can see kind of its peak design when it comes to smartphones. Well, the question becomes always: What's next? Like for tv's... We have the Micro LED from Samsung. "The Wall." Which looks like a nice (but very expensive) concept!

I feel like this is likely to become a tool for artists -- instead of taking 10+ hours to draw an image from scratch or photoshop it, they might spend a minute finding the right prompt and then just edit an image the tool gives them

@James Siewert imagine.

Yeah imagine actually WANTING to do ANYTHING for its own sake?

How fucking depressing the future of art sounds.

True, but this would in turn decrease the total workload of graphic designers, thus enabling them to get more work done in a shorter span of time, which would actually endanger the job.

I’m a software engineer and this is genuinely the first time in my life that an existing tech has felt that it’s impossible, my mind cannot comprehend that this tech not only understand things it understands what would be aesthetically pleasing to a human, I didn’t know a qualia like that is quantifiable

This would make a great wallpaper generator 😁

I've been playing with a bunch of these tools for a few weeks (Haven't gotten DALL-E access yet) and hoooo boy there's so much to unpack.

hi

@True Indian glide ai on colab notebook

Put me in the loop

?

please?

Amazing stuff--side note, I think we got scissors in the bears hands because, from the perspective of the prompt, if the bear did some kind of surgery on this grape plant, it would use shearing scissors, which are used to prune plants! It actually makes total sense. Also, the way Dalle messes up text is hilarious, because this mimics how we see text in dreams! Fascinating. Great vid ya'll!

I love this as a tool. Still scary, but man, imagine Tim had this starting from square one.

Sick! really need to be aware of all the universe of possibilities this AI represents, dangerous and awesome at the same time.

Okay, the one with the goat actually cracked me up. Awesome concept, can't wait to see where this is going.

When you asked about a tiger and Atlantis I was awed particularly by the first image you showed. It's crazy good. I would imagine a real artist creating such image in a span of weeks. It really blew my mind of the capabilities of such powerful AI. It makes me wonder if in the future should I employ a person to prepare me an artwork I wish for, or whether just a work of AI will suffice. If I were an artist I would definitely worry about my future prospects of work, especially if I happened to work out my distinct art style and AI could just learn it and "steal" it without me even getting noticed, let alone benefiting from that. Crazy world we are living in.

accidentally found your channel through another youtuber, david mbugua. wonderful presentation. the end times are nigh … this is where we’ll definitely lose the ability to tell real/genuine/factual from fabricated/fake/fictitious.

This might become a great tool for every designer. Great work, DALL-E 2 Team!

Great step for a none business oriented AI. This can lead to so much more artistic and social oriented applications ! Thanks for the video

This terrifies me more than I care to admit. Not the fact it can make thumbnails, just the amount of articifial intelligence it takes to create these images and how far we've come in the past few years. In ten years the media landscape is going to be so different.

@strayiggyTV this seriously worries me as someone who wants to be an artist in the future. I know humans can give you exactly what you want and people will pay you because of your art style and becase they like you as an artist, but what if it gets to the point where people just don't care??? They think that what Dall E generates is good enough and beats spending money and waiting days for a drawing that probably can't be half as detailed as Dall E? It makes me extremely envious of the artists right now. They may have to deal with this provlem in the future like me, but at least they already have things going for them. People who support them. How am I supposed to build my world when I'm competing with ai that will keep growing and growing?

@sphigel we can agree to disagree, but I don't see how eliminating jobs can exist without eliminating the need to pay for rent and bills and utilities. How would an individual earn a wage to sustain their life without a job, assuming all jobs are replaced by AI? Maybe it is the vagueness of the term post scarcity because I don't see how AI can replace the need for basic human necessities. Even if they made the means of production efficient the cost would never be zero. Would the people who own these corporations and companies with AI workers really be willing to offer their goods and services for free and run a deficit, defying basic microeconomics of profit maximization?

@Daniel Ninkovic Capitalism is literally the only economic system that has demonstrated any hope of ever leading to post-scarcity. Every first-world country on the planet is capitalist. Yes, some countries have more social programs than others, but those programs are funded by tax dollars earned through capitalism. I don't think we'll find much agreement here. UBI is completely unnecessary because if we ever truly achieve post-scarcity, an income will no longer be needed.

@sphigel in theory we wouldn't, but it really depends on the state of the world and the country you're living in. I mean look at the mentality in the US already, the pure vehement against welfare and UBI. A hyper capitalistic "boot strap" society that opposes any form of social reform as "socialism evil!" probably won't be super gunho about having people not be wage slaves and enjoy leisure time. That would imply people not earning money and therefore not spending their money on consumer goods. It would upend the economy unless the government taxes corporations, used those taxes to pay UBI, and then had UBI serve as the disposable income for people to spend on these corporations. Closed loop. That or money as a concept becomes useless, which I don't see as likely as there will still be a cost to maintaining the AI and various capital/machinery.

@Daniel Ninkovic sounds like post scarcity to me. Why would we need jobs when machines can do everything for us?

Exciting albeit scary too. That said, I do think we're looking at, a minimum of a decade before it will actually start coming for jobs. Reliability, total accuracy and the ability to offer even more precise images will all be needed before employers would actually turn to this over humans (as incredible as this tech is btw). However, I do think AI is coming for most jobs, including even mine as a writer. Again, I think we needn't actually worry until like 2035 or so. Just for those who are a little anxious about the immediate future. I see this technology being of assistance to the creative work space from around the mid-2020s but not actually threatening jobs till the mid to late 2030s, maybe even later. And even then, to actually prove triumphant over us humans, that's not going to be an overnight thing.

I honestly don't see why anyone should worry. Humanity has already lived that era in ancient Greece. Admittedly, while horrible, the slaves took care of all menial jobs for the citizens who would instead use their free time for other pursuits. Some of course spent it on meaningless indulgences but others would instead put their effort into thought, like philosophies, engineering, poetry, or skills like music, art, acting, sports or anything else. Those weren't their jobs. They were hobbies. Or rather, pursuits. Rather than work, it was something to do. In a modern world, if you want to do a hobby, you either need a job to support it or turn your hobby into a job. Neither is ideal for those pursuits. But in a world where AI and robots take away jobs? It frees a lot of people to do what they want rather than what they have to, without the slaves. Work on what they want rather than work on a job. Perhaps because of AI, money itself becomes meaningless. Perhaps, the true value comes out in handcrafted things that exist along with machine crafted things that are free for everyone. No one has a need that can't be covered by machines but everyone can get what they want how they want. You could get that tasty meal from a vending machine that makes it exactly as you want it from freshly printed authentic ingredients but there's this person who has a passion for cooking and you like their specific style over the vending machines "standard taste" that lives down the street right next to the interior designer who comes up with the wildest light installations and who would be happy to come over and do one for your room as long as you don't nitpick every detail. Rather than worry about losing jobs, perhaps you could consider how could you benefit from AI doing the job for you? The key is that humanity adapts. We always do. Like the greek society adapted to slaves taking care of things. We certainly can adapt to AI taking care of things.

13:35 - The jobs are kind basic the same but I see them both working together instead of competing. For example, instead of committing and putting effort for your graphic designer, you could go with dalle and show to your client multiple possibilities, then you come back with the chosen one and your graphic designer will create something on top of that. What I see is more of a shift of creativity going closer to the customer so they can share feedback and expectations early.

It's almost like how our minds create surreal things in dreams, but then interrept it in a way that only makes sense to us.

It was brought up months ago, so there's a good chance I'm remembering the details wrong. With that mentioned and in mind: I follow someone who does AI/MI as a researcher and they pointed out there was some (repeat) lawsuit where someone used MI similar to this to make "art" and then to get it insured, sold, something... as art and failed. The lawsuit was basically to try and get it categorized as art and they kept loosing the lawsuit because while the result was aesthetically pleasing or similar, it lacked uniqueness and with identical input, the same output could be produced by different people. The AI/MI researched basically said "yea, that's on purpose. If it could imagine something, that would be one thing, but if it is generating something. And imagining and generating something are different. Even if the imagined item is accidental or intentionally random, it was still purposefully done. While this would change only by changing parameters" I'd need to find the posts again, but an interesting thing to think about.

This is both incredible and kinda scary at the same time.

The idea that this takes jobs is silly.. if u know anyone who works in tins field would tell u this HELPS ppl like Tim &will one day elominate crunch

that is vitalik buterin words..u stoled

We still do really need to improve technology, and it'll be improved anyway. So the point is that we need to improve our society accordingly so technology improvement in it will only lead to everyone's good.

@Michael Eliot good on ya

Ai is going to advance futures technologies so much and i can't wait to see the results in time

I wish you'd tried something like "justice" or "truth" something hard to visualiser at all, wonder what it'd spit out

would be interesting to train dall-e2 on every accessable accuratly labeled content

You know damn well OpenAI has a dark version of Dall-E 2 that has the G+ ratings included. Imagine the possibilities.

"what a time to be alive" indeed! A video about Ai imaging wouldn't be complete without Two Minute Papers signature catchphrase ❤️

🙌📜

@Fernando Henríquez your time scale is off by magnitudes, I can't even imagine where AI will be in 2 years. Here are some of the most impressive AI models which came out just this year, if you haven't been keeping up: PaLM, Flamingo, Gato, Chinchilla, OPT and more You should check these out if you're interested!

Wish he would have put a link to that channel in this description.

Yeah, I wonder if that was a subtle shout-out to KárolyZsolnai Fehér.

Content and context 👇🏻 10:40 Limits 11:35 Quirks 12:15 Letter error has been mostly fixed now 12:35 image transformation 13:17 taking jobs fast - brainstorming 15:00 Two minute papers

I think this is fantastic. The future is going to be so interesting a creative with the help of AI. Great

This is the most amazing technology I have seen in the last decade .. mind blown

it's awesome tech... but it's clearly going to make artists work less valuable. I for one am a little scared of our future with AI.

I’ve had access to DALLE 2 for about a week now and I can’t stop using it. I’ve been trying to see if it can help me invent new product ideas.

use your brain

wtf are you doing here? well, big fan sir.

@maplejake not really it takes months to get around 2 months?

@D Tibor 💀

Can you ask it to create a pregnant man image??

"Will this take people's jobs?" Most definitely yes. Goodbye stock photo websites. At some point in the future this tech will be "perfect" and it will be possible with video as well. In the future we may have to make a conscious choice to value our own humanity and relying on technology may become a moral choice rather than one of convenience. What a time to be alive.

this would be wonderful for finding references for drawings!! i love it already

As an Artist I am so excited for the future of this. I can use this to create certain ideas or compositions I'd struggle with- and then go into more details myself. It's so amazing.

@Vili Evans exactly, 😊 it’s a really neat tool if used properly.

AI should only be used as a source of references to inspire your own art.

I really need it! Many times when I describe things in videos I search for the right image to add and I never find it. This could be the perfect tool!

Dear Marques, I absolutely love the new vibe of the channel. Before was just plain tech, now we see the truth of tech, and new things the internet might not have heard of. Keep up the great work!

Thanks Donald, very cool 👍

"Donald Trump being eaten by a giant Teletubby on live TV."

'DALL - E 23 Took Me Out For Dinner! Here's How It Went...'

Wonder what vids he’ll be making in the future 🤔

That's literally Mind Boggling! AI has so much progressed and still much to go ahead in robots and humanoid machines! 🤩😧 But thats scary that its so capable 🫣

as an tattoo artist i would love to try ´this out for design ideas, like start out with dall-e and then take it into procreate and smooth everything out

Hi Marques, I wonder if you could do a follow up on this video to show off some of the more incredible aspects. For example, glass and water will have perfect caustics. The light transport simulation is fantastic. There are many aspects to explore.

The Diffusion part seems like the tech they use in tv shows and movies where they zoom in on a blurry picture and enhance it to catch the killer

That'd be amazing for college/work presentations. As someone that can't draw, that just looks like a dream

@Daniel González I hate it. I hate the fact that people are voluntarily giving up their autonomy little by little for the sake of convenience. It's dehumanizing.

@Gabuda Ichamuda I understand where you're coming from (honestly I'm just too lazy to dedicate any time to learning how to draw) but this AI is anything but garbage. The fact that a computer can generate these images is mindblowing.

Dude, ANYONE can draw. This language of "I can't do X", barring some catastrophic physical impairment, is complete nonsense that is designed to sell you garbage like this.

👍👍👍

Learn how to draw. It is a worthy skill. Not to show off (do that if u want). But for the mere act of creation. Literally everyone can draw. Just the way everyone can talk or walk. The only difference is that we're taught those things. Drawing is the same. It's just another skill to be learned and practiced.

I can't wait for someone to simply create something that takes completely "standalone" AIs, complex as they already are, and chain them together, output of one to input of another. Or alternatively, create an AI that can take a general input, and simply decide which of the library of other AIs it should pass along the prompt to for the most useful output, and simply return that AIs output as its own output. Basically object oriented programming, only the method call is an entire AI by itself. You can give a writing prompt to a text generator, which is fed into Dall-E to create a frame for every sentence, have that frame cleaned up by a different AI which spell-checks text and corrects it, put that output in a spacial-reference AI to modify the ordering of that output to better match the intended script, get a final frame and when you have enough sequential frames, use interpolating AI to create in-between frames, clean up each of those frames, use image-to-sound to associate "movement" with noises, which themselves are generated by a different AI, then when you get an output, use a few other AIs to clean up the details and blurring, reiterate and eventually the original writing prompt spits out a fully rendered audio-visual short cartoon/movie based on the original writing prompt. Each of these AIs seem in the early stages of working, independently, but I haven't seen an AI which can act as a wrapper for other AIs yet. If the problem with AIs are that they are very good for only a narrow subset of problems, chain them side-by-side so you have an AI conglomerate that is very good at a reasonable (100 or so) related but Seperate narrow problems. Now you can approach not neccessarily general, but Wider Spectrum problems. Have 100 different AIs that can identify 100 different types of cancer? Make a Most-types-of-cancer identification AI. Stack as neccessary.

I'd be curious to see a similar model for prompt-generated music. Where I can ask it to generate "the Simpsons theme song, played on a banjo and didgeridoo, in the style of Mozart"

Finally I can describe my weird "Dreams" that I get while I'm asleep, to someone!!

*Artist :* Well, I'm just thankful that _my_ job can't be automated! *AI :* Here's a renaissance oil painting of you, holding my beer ; )

More videos about topics like this please! Your videography and interviewing skills are top notch and make these difficult concepts more understandable and entertainment.

bots

Lol

this OP's comment feels like it was writtien in a souless way without even watching the video lol

please do googles newest AI that explains why jokes are funny. it's mind blowing.

hi

im genuinely feeling anxious for Tim (and myself too)

"What a time to be alive" I was pumped down looking at my computer until I heard Marques saying that phase. Happiness came rushing to me.

Your content is always professional.

It will be very exciting to see similar technology in the future creating videos based on user's input. Scary to think what it will be able to accomplish in pornindustry, where people would be able to input whatever they think... If this gets to VR, humanity will go extinct

I love how the the thumbnail you asked for was “a robot hand drawing” at 14:30. That could either be a drawing of a robot’s hand or a robot hand drawing something. Because of the ambiguity, it just decided to do both in the same image

@Anup Sebastian I appreciate your lengthy comments even though the other guy is just being a prick to you.

@Mephobia Its difficult to explain to someone who does not understand. This is not a traditional program. We set certain goals for it to reach, and we can measure its performance. However, we do not explicitly tell it some sort of logic. We do not tell it what to look for, instead we allow it to figure that out on its own and it may be able to see patterns that we cannot see within the data, but we do not know the exact details inside the model once it is trained. Things used to be done that way and still is in traditional software, but not in ML/DL. You can use that sort of logic to tie multiple models together, but you cannot really program the exact rules within each model. You can change its structure, and the data it trains on, but not the way it makes its decisions. At first it is really 'stupid', and just guesses randomly, but it learns from the mistakes after each step. It corrects itself based on how far away from the right answer it was, and after many rounds, it gets most things right. Our algorithms only tell it how to learn and correct itself, and not how to make decisions. It learn how to make decisions on its own, and we don't know the exact details of it. At the end of the day the model is not exactly code, but a very complicated math equation that has a certain number of inputs and ouputs and billions of intermediate variables. So yes we don't really understand the details inside it and cannot say how these billions of variables interact with each other to make a decision like our brain does. Just like we don't know the exact neurons inside the brain causing you to make any decision. Dall E is certainly not a single model. It has multiple pieces, for example a language model and a 'drawing' model. We can program that the language model will output into the drawing model, but if we look at the output of the language model alone we won't really be able to understand the details of it, or what exactly it did to arrive at that output, or what it will look like when we see a final picture, nor do we know how the drawing model is going to 'interpret' the language model. We have some idea, but not like when we program the logic of it directly, in the old days.

​@Anup Sebastian If you work with this on a day to day basis, It's incredible no one has fired you yet. Of course you can read why and how the AI thinks. You can analyse every data from an AI, so the developers understands how it thinks and works. Like i said, it's not magic. Its a program written by humans and can be read by humans, even after an AI updated the code. It's not like "oh i wrote a program, but i have no idea what it is doing" You can always see what it is doing and why.

@Mephobia I don't think you understand. We don't know what is going on inside after training models. We do not know at what point in the 'code' a certain decision is being made, like traditional code. We cannot find bugs by looking at it like traditional code. We do not tell it any rules, it finds out. What we have is certain techniques to get machines to learn efficiently, and there is some research to probe what happens inside models once they have finished training or to see what happens during each stage, but at this point it is a black box. I know this as it is literally my job. I am by no means an expert on the forefront, like the people who worked on Dalle E, but yes I do this on a day to day basis. Yes you can say they are 1s and 0s, but we cannot really do anything if we can just see some 1s and 0s on the screen like in the Matrix. Code in real life isn't like that. There is nothing stopping us from saying say that the human brain is also just a much more complex system of 1s and 0s. And what you think of as original thought now, will someday just be able to be broken down and and understood as nothing more than complex math. No magic. Dall E does show some signs of independent thought, by adding details to images that you have not asked it for. You literally just send it a short request and don't need to be super descriptive. It is able to 'imagine' what the rest should look like, even if it is not something realistic or if it has seen in before in the training data. It decides the background, the lighting, the choices of color, the style of the art, and also makes it visually appealing, etc. These things are not coded into it and frankly we don't understand how it does this. This is what they mean by it being a little creative. It also surprises the people who made it. You can say it is creating stuff by understanding the context of what it has seen before, but this is also how humans artists create new work inspired or in the style of old work. And it is also capable of creating full on psychedelic artwork not based on anything at all just like humans can. This field only had its modern resurgence about 20 years ago, and we do not yet have the processing power to fully simulate 'general AI' like what you see in humans, nor do we understand it yet. Also that is not the goal at this point in time, as it is so far away, however, I wouldn't be surprised if we saw it in our lifetime. You saying it is not possible because it is not done yet is like someone in the 19th century saying airplanes were impossible because it hadn't been done at the time.

@Anup Sebastian ​ I guess you didn't get my point fam. My point is that it's not possible, because then it would already be created. It is not created which means that it's impossible to code a brain. And yes the developers do in fact know what is going on inside once it has learnt. Because it's all code, they can read the output, the algorithms, the logic behind that choice. I'm not saying AI don't exist, they do and some are very good. But it's all coded, its all 1's and 0's nothing more. No magic. No independent thoughts. No creativity. Just algorithms and data based on what the machine/program has beed fed. That little voice in your head that tells you "oh i have to do this" or "okay i make these chors and then i watch TV" that 'voice' is not possible to imitate. just an example. My points is that the brain is so complex and advanced that it would not be possible to make an exact copy. Yes AI's a good, but nowhere near brain level...

Hey, do you know if there is a tool that can take one image and make another out of it, with similar features ?

Damn, feel like I dodged a bullet when I left art school, now I feel like I also dodged a fucken cannonball. On the upside, I think this will be no different to the birth of photography, and will just become another tool. I can see art directors utilizing the shit out of this.

Damn, I thought I’d be dead before we got to this point. Having to question the legitimacy of things at this level is not good.

It takes me more than 10 seconds to open Photoshop and start a new document 😂 and this AI can spit out 10 images from nowhere that's pretty impressive

It takes DALL-E 2 about 25 seconds to generate 10 images.

As an animator and designer, i beginning to understand how truckers feel.

yes this will be inexpensive which will mean that book cover art and magazine illustrations will get more uniform, even more banal, even more predictable, and, worse, more illustrators (already being replaced by 'stock illustrations') will be put right out of work. "aesthetically pleasing" he says. To whom?

Dey took err jabs

For real though..

🤣🤣🤣

Ahaahahhahahaahahaha

As an artist, here’s my take. At least in it’s current state, I don’t think this will replace artists. Fully. Let me explain. In it’s current form it’s effective enough for the average person to generate a random image and be fairly pleased, but anyone who cares enough about image quality or any precise level of customization won’t be happy with what this can achieve. People who want logos, graphic design, character designs, fictional maps, anything that requires very specific detail, they’ll want a genuine person who can ensure everything is correctly represented and not just photomashed together to the point where if you zoom in you can see the seams. It also seems to struggle a bit more when it comes to inventing new and different art styles (since most people don’t give their style a definable name the AI can use to categorize it) and we know it struggles with text, so I think these areas will still be majorly human based. It really is more the realism that the AI does best, and frankly most people don’t hire artists to do realism unless they want a portrait done and the AI isn’t allowed to do faces. Professionals likely will only be using this for references and mock-ups, not for fully rendered and usable art. That’s not even considering the group of people who will likely avoid this simply out of fear and principle, not wanting a robot to be the one responsible for creating whatever it is they want. So overall, I think depending on what’s needed by the person it may take away some business, but I don’t think it’ll take away all art related jobs completely. It can’t replace portrait artists, in it’s current state it can’t even replace graphic designers or logo makers due to it’s text failures. If they somehow manage to refine it enough where it no longer has issues with text and somehow is able to produce images that are not fuzzy and visibly edited when zoomed in, then it’s a bit more of a concern, but even then I genuinely don’t think it’ll ever be able to fully replace artists due to that limitation surrounding faces, NSFW content, as well as the fact that it cannot possibly categorize every possible art style in existence because the majority of them outside of the classics do not have names to generate from. It would take someone an incredibly long time to generate art in the style of a specific artist they know online, where as it’s much faster and frankly more customizable to commission them. That’s how I see it, anyway.

@Kolwolf I agree with this as well. A lot of people, though, will be perfectly happy with the thoughtless piece of art the AI creates, which is why it’s a threat. People who really like art, though, will likely always look for human artists exactly for that thought and creativity they put into their work.

@GrimReaper I agree, but unless and until they find a way to eliminate the roadblock of allowing likenesses (and NSFW content), it can’t fully replace real people. Because ultimately there will always be a demand for portraiture (and NSFW art) and unless we find the right balance where people can request other people without it being a risk to our security that will always be something people have to source from humans. Sure, if it’s something cartoony you can certainly do things like you said and tweak things here and there until it looks like the person, but then if we allow that kind of manipulation with photorealistic art don’t we reach that same issue where it can be used in nefarious ways? Then there’s the fact that We’re already seeing a lot of indie game developers reverting back to traditional creation (using hand painted assets) in their games instead of using new technology that could achieve similar results, so I wouldn’t be surprised if we end up in a similar place with this AI if it ever becomes something easily accessible to the public. It’s one of those things that has so many complexities that it could absolutely happen, but it’ll take time and although I’m not necessarily optimistic I’m hopeful that in that time people will come to the realization that inventing AI to entirely replace our current jobs likely isn’t the best route for us. Not only economically, but just in general. It seems fun, but it’ll only lead to even more devastation in the real world.

What an "artist" mean to everyone is different for someone it's a cool creative gif for someone it is hyper realistic and detailed painting or basically anything at all but for someone it is the person behind bringing together all his life experiences and views on life to impact his way of going about art and expression and thought behind the outcome which in this case would be a digital art piece and if you ask me in many ways that is not something an ai can even begin to replace but the best it can do is to emulate it and for some people that is just enough and that is fine :)

It won't replace artists yet, but this technology is advancing at such a fast pace, it's insane. I believe that in 2 or 3 years you'll be able to do exactly what you described as the only roadblock between this replacing artist. Just generate a basic image with any prompt, and tell the A.I what I you to change or replace. "Add mountains to this map.", "Change this character's facial expression", "Make the hair flow in a different direction." etc, etc. We'll see it in our lifetimes, and I can't wait.

To be frank, I didn't understand the concept of most of @mkbhd's description. But the AI did. Now I am in a situation where I'm having to learn from a computer, earlier I used to give commands to a computer.

This would be amazing for art could get amazing references for what you are thinking of drawing.

I need this so I can stop searching for clip art for presentations. It would be so nice to be able to imagine a picture and have it appear effortlessly and not need to remove watermarks and ai upscale.

I will have access to Dalle-2 next week, I am planning on doing some weird camera designs that never existed

@Conrado Moreno I will !

Lucky! I'm still waiting

Hi Mathieu, huge fan. If possible, maybe ask it to emulate old film stock that we can’t find anymore?

Ask it to draw itself, would be curious to see what it generates

Could you ask it to draw an "impossible image"

See that's where you lost me "No adult content" The sheer amount of buckets you'd need to carry the money home after releasing a porn version of this is a huge waste

"DALL-E 2 is an AI that can draw anything you ask it for." "Now here are all of the things it can't do."

After words "What a time to be alive!" I've heard "squeeze that paper!" in my head 😅

What this tells us is that Marques has an underwhelming imagination.

While this might look like it's gonna replace the jobs of artists, I feel like it's an *extremely* helpful tool for artists. Being able to get a reference image of something that doesn't exist, in under 10 seconds. Amazing

It's not going to replace artists. Have you guys heard of artstyles? That's something that the robot cannot emulate yet.

Lmao not yet but three years ago it was just text eventually it will be so flawless and will even be better at thinking of ideas way better and people can and the desire for people to be creative will stop entirely which sounds scary but I bet it’ll actually be really cool

@Vili Evans Uh, so you're saying everyone should just learn to draw? That's never going to happen. I bet only like 10% of people know how to draw, and less than 10% of those are very good at it. Do also understand that you can't just 'learn to do things yourself' for EVERYTHING. If you 'learn to do something yourself' every time you need something, suddenly you're a jack of all trades, master of none. This is why specialization exists, and also why AI in the future will be very attractive for companies and people who commission artists already.

@Bose-Einstein Perhaps, but if you actually learn to do things yourself then there's no need to awkwardly try to get results by writing words that still won't produce exactly what you had in mind. Your concept of art is very limited.

@Vili Evans Think of it like commissioning someone. You have a back and forth convo about what you want, and if the conversation is done well, you get what you wanted and more someone. AIs in the future will be able to operate like this.

my brain is melting from how crazy and amazing this is

Welp there goes my dreams of being an animator.

For real though, technology is gonna replace so many jobs that it would be absurd for us to keep expecting everyone to have a job. It will simply not be possible

AI should only be used as a source of references to inspire your own art.

This is amazing. Terrifying. But amazing tech.

It's definitely Scary but not so much as Terrifying.. "Humans" can be or Are just as Worse. But this is pretty Amazing, agree. 👍

Calling AI terrifying is Antivax/5G/pancake Earth vibes

In future AI will read this comment and also will ask you why did you write that?

hi

type of thing that really makes me wonder what the real value is in art