Cleo Abram
Why DALLE, Midjourney, and other AI image generation tools are so controversial…
Check out Storyblocks and sign up for the Unlimited All Access Plan: https://www.storyblocks.com/cleoabram
You’ve probably heard about the AI that can make any image you want. With these, anyone can make art in seconds. Artists, understandably, have concerns about that. This technology has quickly become VERY controversial.
Since we got access to DALLE-2 a few weeks ago, my friend Justin and I have been sending our creations back and forth. It’s a lot of fun. Justin is a designer and animator who works with me on Huge If True. He’s an artist. I’m… not so much. What I noticed as Justin and I were making these is: His were way better than mine! It didn’t feel like the AI was leveling the playing field between us. It felt like it was giving me new skills, but it was giving HIM superpowers.
So, we set up a little competition to test that… and you voted on the results.
Along the way, we’ll explain how these artificial intelligence image generation tools (Open AI’s DALLE-2, Midjourney, stable diffusion, and more) really work, why they’re so controversial, and why they matter for our future.
00:00 Why am I hearing about AI art?
02:00 How to use DALLE
04:15 Thank you Storyblocks!
04:46 Round 1: Cleo with no AI
05:50 Round 1: Justin with no AI
06:36 How do AI art generators work?
08:09 Round 2: Cleo with AI
08:54 Round 2: Justin with AI
09:24 How does DALLE use artists’ work?
10:25 Should artists worry about AI?
11:12 Why is AI art so controversial?
12:08 What’s next for DALLE?
13:07 RESULTS
Subscribe to see the next episode: https://www.youtube.com/cleoabram?sub_confirmation=1
For more art and animation from Justin: https://www.the11percent.com/
Sources and further reading:
– OpenAI’s tutorial for DALL-E 2: https://openai.com/dall-e-2/
– Free Hugging Face DALLE mini: https://huggingface.co/spaces/dalle-mini/dalle-mini
– Aditya Ramesh’s detailed explanation of how DALLE-2 works: http://adityaramesh.com/posts/dalle2/dalle2.html
– Midjourney’s community showcase: https://www.midjourney.com/showcase/
– Marques Brownlee’s DALLE review: https://youtu.be/yCBEumeXY4A
– Joss Fong’s explainer on how AI image generation works: https://youtu.be/SVcsDDABEkM
– “Generating Images from Captions with Attention” https://arxiv.org/abs/1511.02793
Be featured in an episode – upload questions for me to answer! https://www.dropbox.com/request/Edocsb2kErpueQ7F9T1Q
tw
I tell different stories in different places:
You can find me on TikTok here for short, fun tech explainers: https://www.tiktok.com/@cleoabram
You can find me on Instagram here for more personal stories: https://www.instagram.com/cleoabram
You can find me on Twitter here for thoughts, threads and curated news: https://twitter.com/cleoabram
Bio:
Cleo Abram is an Emmy-nominated video producer and journalist. Cleo produces detailed explainer stories about technology and economics. She wrote the Coding and Diamonds episodes of Vox’s Netflix show, Explained, was the host and a senior producer of Vox’s first ever daily show, Answered, as well as a host and producer of Vox’s YouTube Originals show, Glad You Asked. She now makes her own independent show, Huge If True. Each episode takes on one big technology innovation or idea, explains what it is, and helps people imagine the ways it could improve the world we live in by answering one simple question: If this works, what could go right?
Vox: https://www.vox.com/authors/cleo-abram
IMDb: https://www.imdb.com/name/nm10108242/
Gear I use:
Camera: Sony A7SIII
Lens: Sony 16–35 mm F2.8 GM
Audio: Sennheiser SK AVX and Zoom H4N Pro
Music: Musicbed
Follow along for more episodes of Huge If True: https://www.youtube.com/cleoabram?sub_confirmation=1
—
Welcome to the joke down low:
What did Vincent say when he couldn’t find his car?
Where did my Van Gogh?
Find a way to use the word “car” in a comment to let me know you’re a real one 😉
Hey DALLE-5, make my next YouTube video
I love the rest of your content, but the way this blithely skips over the blatant copyright infringement issue is offensive. When these algorithms are trained from either licensed material or public domain material then fine, until then it's 100% just profiting at the expense of actual artists past and present. This is nothing like the transition to NLEs – just because you purchased an early copy of Adobe Premiere didn't mean it would spit out Casablanca with talking bananas, you still had to shoot (or animate) that yourself, cut it together, etc. Making things more accessible and stealing the work of others is not the same thing.
I guess I have too outlandish requests of DALLE because I have so far been very disappointed in it with my prompts.
😍
Radios and cassetes will take away jobs from musicians and make concerts more rare! / The Television is going to make radios a thing of the past! No more radio announcers jobs!
And you know the list goes on, people fear change but progress demands it.
We cant stop the évolution of some of these phenomena unless we live in a dictatorship
I've used these AI image generation tools, and everything I produced in some way seemed to be copying some pictures that a human created. My concern is that if everyone uses these tools instead of creating actual art, then where does the new image input come from? Would we eventually just have the same inputs recycled in a bunch of different combinations? Eventually it would be like inbred images with nothing really new.
Humans are SOO SSSTTTTUUUUUPPPPPPIIIIIIIDDDDDDDD… We've been getting warned about the dangers of AI tech… The "illusions" being experienced by these LLM are precursors to the singularity we OBVIOUSLY don't fear enough. Are the rules of engagement robust enough to protect humanity? And what's to stop bad-actors from bypassing safety protocols and simply take the brakes off? And, here's the most interesting question, if necessary, can we stop developing AI even though we clearly don't want to? Thank you, "Science", for creating unfathomable danger and possibly dooming us all.
terrible for tactile artists, great for everyone else. i bet low-paid niche duct-taper work (as far as virtual jobs go) will be the safest for longest, since they're working at companies that already don't care to optimize performance.
People would not hate AI that creates art so much if it didn‘t steal from real artists‘ works to generate new pieces.
And making a computer generate art for you doesn‘t make you an artist.
You're kinda missing the point… it's taking other peoples work without their consent. Thats the big issue.
When people take insparation it just isn't the same thing as when an algorithm pickes up a data set.
Really cool. But, It's really expensive to use.
Glad I switched to healthcare; the one area plp dont want to talk to AI or machines, they want a human.
It seems like you delivered quite a basic point about "AI is just a tool", but didn't address ANY of the concerns other than the most literal "AI will take our jobs".
Sure, it can be compared to many things, but AI even as we know it now is quite unique and dangerous in many respects. Potential for fraud and forgery or stealing the intellectual property to name just a few.
It feels wrong to ignore such obvious and almost inescapable problems, while focusing solely on positive message and potential of the technology.
This woman gives me the creeps. Is she real or vinyl? I have never suffered through such pretentiousness in all my YouTube days.
Who pays for this channel? This is nothing but corporate and right wing disinformation and propaganda.
That's because people in general think inside the constraints of what they know . Smart people who saw the applications to this technology went outside the norm and did what the guy did with the spread sheet and turned a gimmick into a job set. Now you have people who hire themselves as AI prompters making money from people looking to use AI but don't have the skill set to use it to it's full potential. Much like with the advent of Photoshop, Fruit Loops, App development. It's a case of those who watch the guy in front pull further away.
Great vid!!
Now we make 4k Images with no flaws, 3d models, animations, voice-overs and so on and every month with no joke i see new AI evolution skills that come from the open source community not the company's.
And 9 months later, the point we're at is not that much different. Images can be almost photo realistic now, but artists are still safe because they can create better looking stuff and utilize those other tools better than anyone with just AI tools can.
BUT.
Free version of ChatGPT helped me write the first chapter to a book idea I've been thinking about for years, in a single afternoon. Just the rough sketch, but still, 10k words in a single afternoon of playing around with the tool. I probably generated 10x that in text, I had a second word document open to cut, paste and rewrite things to get closer to what my idea was. But what I finished in that one afternoon is still the pace that NaNoWriMo writers challenge themselves to write in a WEEK. To write – a very rough first draft of a – book is a matter of weeks. Free online for everyone to access!
I've also had a tabletop game idea bouncing around in my head for a year or two. And again, it took like two afternoons of sketching and talking with ChatGPT to iron out rough first version of the rules and a deck of 30ish cards, good enough for initial testing if the game is actually fun.
I'm a programmer by trade, and my last creative writing happened 20 years ago. The power of these tools enable almost anyone to try out their creativity. I almost certainly will not have a literary masterpiece in my hands, but getting that first draft out is already feeling a lot better than I thought it would ever feel. I'm motivated to try. I want to playtest my game, I want to write the next chapter to the story. I want to come up with more solutions!
I've also used ChatGPT for work once. I know barely enough SQL to get around a database. I needed a script to change the collation of every character based column in a database (From a latin alphabet to a greek alphabet supported format). This was my first time using ChatGPT. For those curious: Cursor to go through each table in the database, a second cursor to list each column with any char-type data in it. Cycle through each of these columns, and alter the column, conserving the name, data type and size, but changing the collation. 67 lines of code, with comments.
Researching this thing it would've taken me several days, if not weeks (at which point it would've been faster to write each column name and attributes individually). It took me two hours – again, using ChatGPT for the first time!
These tools will be extremely powerful for a lot of people in their work, but also the free tools will help a lot of people in their personal projects too. Once video and audio generation improve, I can already see personalized animation projects becoming accessible for a whole lot more people.
People behind Stable Diffusion released Stable Doodle just yesterday to the public. You can doodle something in a window, give the similar generative prompts as you do for the generation in general, and it will do it's damnest to keep the doodle in mind as it generates what you drew as a prompt. Give that tool a year and give it that same "erase the part you don't like and it'll generate a new version" option, and you are looking at a tool that could potentially bring anything, ANYONE draws, to life.
AI generated art compared to human created art is like fast food chain compared to fine dining/craftsmen's restaurant/small family restaurants. AI can generate accessible art for everyone who wants them widely and cheaply, but it can't overthrow the human desire for humanity, experiences and actual human interaction. I would hope that AI art can lift up the beauty standard of the masses, as it makes art so accessible, but I would also hope that human artists can be protected and respected for their exclusive contribution of the beauty humanity. I think the existence of AI art really gives human an opportunity to ponder, what is it that makes human art human, just like now when people decide to take a break from fast food and go to a great restaurant, it is for the experience of humanity, not just to solve hunger.
At 14:24 when it shows a few AI art examples, are those text promps the true prompts to get those images? From what I have seen, it usually takes a lot of different descriptive words and also lots of experimental renders and adjustments to get the ai art to look really good like that.
A great example of the mindset that people should have for that technology can be found on the Corridor Crew channel, the behind the scenes video of Anime: Rock, Paper, Scissors. Their channel as a whole is focused on making cool sketches with the newest technologies and disecting movies scenes from different jobs perspectives.
It’d be hard for me to find it but I remember watching a podcast with video game developers, one of whom is now a director. He said his 40hr/wk first job in the industry is now the click of a button in a game engine. This didn’t put him out of a job, it freed him up to do other things, create more complex systems, etc. these systems might ALSO become a click of a button later, but that’s how it works.
I have no problem with the creation of an AI tool that allows for more creativity and broader inclusivity. What i do have a problem with is the companies developing these tools making billions using the work of UNCREDITED & UNCOMPENSATED artists to train the AI tools…
At least it will make a real artists hand made work more valuable especially if it's a painting.
I wonder if it's possible for DALL- E 2 to produce a set of parameters (prompts) that would consistently and accurately reproduce art from across an artists ouevre. So, an artists entire collection could be viewed, and it could be perhaps categorized into 5 different lenses (filled out prompts) for the different combinations the artist subconsciously created. Then, perhaps have the lenses themselves recombined for new art, or to perhaps screen for other artists / fakes. I know Remedios Varo said she had made Giorgio de Chirico fakes for money. DALL – E 2 sounds like a great way to objectify the study of art, and authenticate.
Oh the times we are living in is really annoying me!!!! 😮Life is not about gun it is about finding who you are ❤- which is a creation of the Almighty GOD – REPENTING AND COMING TO JESUS CHRIST is our task as long as our heart beats ❤
and this your fun will get stuck in you one day…… In ALL of you 😢
I do like semantic aswell but there is a difference between given out "tools", and creating a 100% design idk..
Wait till the terminators start rolling off the production line run by robots controlled by ai programmed by a psycho
As an artist, when I first heard of this I was terrified. What does this mean for me and my future? Than I realized, AI is not an overpowered artist. AI is a medium.
Literally none of you gave a shit when it was truckers, cashiers, and others you deemed as "lesser than" getting replaced by AI. Now that it's upper middle class college educated people getting replaced it's a big crisis and moral panic.
I guarantee you could find a ton of you making fun of truckers telling them to "Learn to code lol" on Twitter a few years ago when self-driving trucks were the buzz.
Brilliant subject to discuss. I draw portraits of people, pets, and animals as well as buildings snd landscapes in colored pencil. I dont mind people doing their own thing with PC art. Ive never used computer programs to improve any of my own work, and Ive never even used Photoshop. To me its a waste of time. Im just old-fashioned and could care less what computers could do.
you entirely missed the point of ai art being literal theft. all it does is steals from real human artists. ai could not make a single art word without the already existing human art pieces
This is really cool, but as a creative, i’m still worried, not just in a “HUMAN JUSTICE!!!! RIGHT AND WRONG!!” way, but in a “I am not capable of doing any not creative job, so if I am put out of work by something like this, used by consumers of art rather than artists themselves, then my life is going to be not so good, and I’m terrified.” way.
Sorry to sound mean, but Aditya Ramesh has a wild voice. I thought it was computer generated to start with.
There has to be ways for an artist to be compensated for their original works. The owner of a large language model A.I. should not be able to just take other people’s hard work without permission. If ever there were a need for an online micropayments system to be created (perhaps based on blockchain technology) so that artists get paid, this would be it.
11:14, No offence but what you are talking about is a different thing. I am a young artist and AI art is scary because I feel like this might replace artists. Some companies are using ai art instead of artist and there are a lot of artist out there who want to be recognized and hired for their artworks and talent, me included. Making artworks takes time and patience to make. AI can be used for inspiration, it is not a bad idea, but it can not be used for doing the work for you. It feels like cheating. Art is about practice and the talent which is why what makes it special, imo.