Rodger Werkhoven, an independent creative director at OpenAI, is coming to the first AI Weekend, which will be held in Rovinj from 19. to 21. of September, to share his insights on the latest achievements of artificial intelligence and its application in the real world. Rodger Werkhowen in addition to being involved in the creation of ChatGPT and DALL· E programming tools, also used artificial intelligence in art. For us, he reveals everything about the artistic collective CollaborAitors, the first AI-generated film “Let’s be firends”, and how they won the prestigious award at the 69. International Short Film Festival in Oberhausen.
In the spring of 2022, you joined a select group at OpenAI and contributed to laying the groundwork for the successful release of ChatGPT. How challenging was it?
Well, ChatGPT wasn’t a thing yet. Not even a concept. It’s actually quite an funny story. It’s incredible how much of a role ‘chance’ played in it. When I received my invitation from OpenAI to join their group of external creative professionals for their DALL-E 2 research, they were extremely excited about this much more complex and capable text-to-image. DALL-E 2 almost overshadowed all the attention for their GPT models; it sort of emerged naturally. Since February 2019, I had been experimenting with the GPT-2 model through an external organization. At some point, more active GPT researchers asked OpenAI for an easier workflow to investigate the GPTs with less hassle. To facilitate this, OpenAI’s developers years later added a simple question-and-answer feature that helped get to the core of GPT research more quickly. This Q&A method turned out to be so useful by accident that it took on a life of its own and even became a product: ChatGPT. But that was much later. We were discovering everything about DALL-E 2 long before that. The world was excited about an AI that can create images from a single line of text. At the beginning, we were told that the experiment would be discontinued if the results were satisfactory, but the plans changed because the research was expensive. OpenAI decided to see if it was possible to make money with DALL-E 2. We discussed the different billing models and whether students should be given free access. We were able to invite people for an art program, and DALL-E 2 went into beta. OpenAI has become more and more relaxed about sharing the results publicly. At the beginning of the project, we were not allowed to publish realistic faces due to fear of identity theft and other ethical issues. Also, the waiting list for DALL-E 2 quickly grew to one million people. Meanwhile, we mapped out the weaknesses of the models such as: issues of bias, gender role, racism, anything that is potentially offensive and that you wouldn’t want children to generate. We’ve worked with OpenAI developers on solutions like ‘Latent Prompting’ to ensure inclusiveness. By the time of commercialization, the DALL-E 2 model had become sufficiently secure, inclusive, user-friendly and profitable. In September 2022, the waiting list was lifted, and one million people were given access, in addition to the ChatGPT tool, which became available to the public in November. Everything OpenAI learned during the development of DALL-E 2 served as the foundation for the launch of ChatGPT.
Did you expect so much success and global acceptance of ChatGPT?
Not everyone foresaw that ChatGPT would have 100 million users just two months later, considering it was initially intended to support GPT testers in their research. But it makes sense. When do you need images from DALL-E 2? And when could you use some textual support? Well, for job applications, homework, correspondence with your local government, lawyers, your employer, complete social media campaigns, you name it. ChatGPT immediately had much more relevance.
AI in Art and Creativity
You also contributed to the creation of the world’s first video experiments generated by AI, and you are also a co-founder of CollaborAitors – an art initiative that combines art, technology, and science, and which produced the award-winning film “Let’s Be Friends”.
Since I was not allowed to generate and reproduce realistic human faces, I first generated a series of very theatrical Shakespearean actors, of all ages and ethnicities. After all, you can’t use them in a fake passport. Then I thought, “How cool would it be if we could ‘act’?” I mentioned it to OpenAI’s developers, but the video wasn’t part of their plans. DALL-E 2 wasn’t meant to create cool stuff for the creative industry. That’s when I looked outside of OpenAI and came across the “My Heritage” algorithm. They used a licensed AI program trained for human facial movements, and were able to make photo-portraits of real people naturally move, look around, and smile… I wondered if this algorithm would also accept AI creations. And yes, the first-ever GenAI DALL-E 2 animated creation has become a reality! I published my findings about the multimodal setup between the two GenAI models, and the creators of the other, D-ID AI, responded. They were impressed and asked me if there was a market for it. I said, “The number of people who could animate photos of loved ones is finite. The number of AI-generated characters that creative animators and storytellers can create and bring to life is infinite. In OpenAI alone, there are already a million people on the waiting list. Do I need to say more?”. The D-ID also had lip-sync and speech AI, so they asked me if I wanted to try it. So I created the first-ever AI animated, speaking, AI-generated entity. Then I asked a Dutch hip-hop legend, DJ Alien Rotterdam, to write a rap about AI – hip-hop is a culture. Thus was born the first-ever AI-generated, AI-animated, AI-dubbed rapper, without human motion tracking. All separate AI creations. In the end, the famous Dutch multimedia artist, Arno Coenen, suggested that we show the project at the Dutch GOGBOT festival of art and technology. Then, in order to create the film “Lets be friends”, we founded an art collective – CollaborAitors. We premiered the film in 2022, and a year later, the film won an award at the prestigious 69. International Short Film Festival in Oberhausen.
Do you think artificial intelligence will soon find its place in all branches of art?
Yes. Because true artists are inquisitive minds. Of course, that research can be ‘limited’ to delving deeply into whatever an artist’s medium is, say ceramics. You don’t necessarily have to work with AI. But Arno Coenen, among many other disciplines also a ceramist, finds it interesting to see how he can push the boundaries in ceramics with AI. We recently won a major international pitch in Taipei, Taiwan. With AI, we designed a 6-meter-high dragon that will be entirely executed in ceramics. Through meshy.ai, we offer local schoolchildren the opportunity to participate in the design. Local ceramists will realize the whole thing, bringing employment to those local craftsmen. The limits of our own creativity are our only restriction now. I’ve even seen dancers dancing with AI as a partner.
Being a creator with artificial intelligence – have you ever felt that AI surpasses you in creativity? How do you deal with the idea that your own creation could be better than you?
Today’s AI tech is not creative at all and only very limited in problem-solving. AI often comes up with accidental combinations that we humans find creative. That’s not the same. We give meaning to the output of AI, which is the result of a process based on our input. And very often, people find that output creatively underwhelming. That’s largely due to the limited input of those people themselves. I see well-educated creatives with an art academy background generating very interesting output. Sometimes I see AI output that makes me envious almost ;-), realizing that it’s because of what that amazing, talented person instructed the AI to make. AI output also regularly surprises me positively. But such an AI has no idea what it’s producing. It needs me, the human, to assess the output’s value. Creativity and intelligence go hand in hand. In humans and in AI. I look forward to the moment when AI does become genuinely creative. Then it will have developed from a tool to a full-fledged creative partner. But then we will also have reached the point we call ‘The Singularity’. If AI becomes more creative than humans, it will have achieved the status of ‘Superintelligence’. I’m curious to see what that means for creativity and art. I wonder if we humans will still be able to comprehend that art with our limited intellect. I don’t think so. Even the ability to judge art depends on the intelligence level of the beholder. There are animals that create ‘art’ to impress their potential mating partners, such as the bowerbird and the white-spotted pufferfish. But those animals are not capable of understanding Homo Sapiens’ art. Only the cockatoo seems to appreciate our music…
Have you ever wondered, ‘Am I a creator or just a coder’? Have you ever had creative blocks while working with AI and how did you overcome them?
Python, C++, Java, are just languages. Just as arranging kanji sequences is the way to communicate with Chinese people, code is the way to communicate with computers. Fortunately, humanity has now created AI transformers for those who are not proficient in any language, to help them convey their ideas. You are not your language. You are your story. I’m lucky never to have a creative block… That’s one of the many great advantages of an ADHD brain. Sometimes I have no interest in a creative task… If it’s not challenging enough, for example… Then I simply raise the bar higher than where my client set it… I find the task fun again and hopefully surprise the client with output beyond their expectations… That sometimes backfires, of course. But an intelligent client also finds it exciting and cool to step out of their comfort zone… The reward can also exceed expectations…
AI and Business Is there a plan to introduce financial or other compensation for artists whose works are used as bases for generating AI artworks?
Certainly. There is consideration of compensation plans similar to those in the music industry, among other things. But that’s quite old-fashioned thinking, in my opinion. I once met an artist who ‘complained’ (but turned out to be lying) about being prompted. Truth: He wanted to be prompted because he saw it as PR… After all, where social media is saturated, one can quickly gain a lot of attention from many potential fans in the few most well-known GenAI models? Quite a few unknown artists would be willing to pay a cent for that, similar to ads on social media. I recently spoke with a creative director at Disney – whose name I’m not allowed to mention – who doesn’t mind if AI enthusiasts generate Disney and Marvel franchises. As long as they don’t try to profit from it afterward. Do you know how much an ad on Facebook costs Disney to keep a franchise in the spotlight? Finally, take a cue from Brian Eno. Early on, as an artist, when tablets were still emerging, he had apps made that allowed people to create his visual and sound art themselves. Eno earned from app sales. Other artists also earned from ads in free versions.
It is said that AI is a crucial tool in the business sector. Is this the case or is it all just a marketing trick?
We have been able to conduct business very well for thousands of years without AI. The Dutch East India Company – founded in 1602 – was a pioneering example of a joint-stock company, where investors could buy and sell shares of the company’s stock. It only becomes necessary for entrepreneurs to embrace new technology if failing to do so puts them at a competitive disadvantage… In that sense, entrepreneurs will now indeed have to adopt AI.
How would you convince a skeptical CEO to invest in AI? What are the specific benefits that AI brings to business that traditional methods cannot?
ChatGPT-4o says:
- Enhanced data processing and analysis
- Improved efficiency and productivity
- Personalization at scale
- Real-time monitoring and adaptability
- Predictive capabilities
- Cost reduction
- Enhanced security
- Improved customer service
- Innovation in product development
- Handling complex, data-intensive tasks
I would add: - AI removes business insight-clouding emotions like overconfidence and fear from
business processes.’
What advice would you give to CEOs who want to start implementing AI in their companies? Where to start, what are the first steps they should take, how can they educate their teams on the benefits and challenges of AI?
Start by asking these questions to your respected and loyal current advisors and suppliers. They won’t risk the relationship by pretending to understand AI implementation. If they are not there yet, but inquisitive, go research together. If they have no ambitions in that direction, they will do their best to help you as well as possible for the sake of the good relationship. If they can’t help? Then contact me: www.linkedin.com/in/rodgerwerkhoven.
How would you describe your relationship with AI – is it mentorship, partnership, or something else? Do you consider AI as a tool, collaborator, or even a competitor in your work?
My relationship with AI is playful. I see AI as very cool, very handy, very awesome toys, with which I can achieve surprisingly serious and seriously surprising relevant, usable results at my pace: much technology is too slow and too dumb, too limited for me. Not AI. My ingenuity, my own creativity is decisive. Not what AI can or cannot do. Take this example: I’m a big Kraftwerk fan. Their 1981 album ‘Computerwelt’ (Computer World) was prophetic. And their rhythms and sounds were so innovative… they inspired completely new, very diverse music genres like Hip-Hop, House, and Techno. What set Kraftwerk apart from other bands at the time was their playfulness. While other bands saw their new synthesizers and drum machines as tools, Kraftwerk had a very playful approach. They even included beep-making children’s toys to generate their sound, which would become the sound of the future – now. Unconventional in what you can or cannot do with a synth or toy. They even pretended to be ‘Mensch-Machinen’ (human machines), robots. When they played the song ‘Taschenrechner/Pocket Calculator’ live, the audience was allowed to press the toy-synth buttons while Kraftwerk made nerdy dance moves themselves. Dare to play! Also with AI! That’s where innovation comes from. Just like I made GenAI’s first-ever animations with a tool intended for ‘consumer service and educational purposes’.
How do you see the future of business in a world where AI is becoming increasingly present? How do you think business models will adapt and what are the potential risks of over-reliance on AI in business?
To start with the risk: over-reliance on AI in business can lead to disinterest in how processes work, how results come about. Then people quickly accept that the results are just the results. “Computer says ‘NO’.” is then seen as insurmountable. If AI can’t solve it, then it’s unsolvable. That would only stifle innovation. We will also see very exorbitant examples of Murphy’s Law. Because even with AI, things that can go wrong will go wrong. We recently had a similar preview with the CrowdStrike BSOD debacle worldwide. Such a disaster can happen if people think ‘it will be just fine’ and therefore handle tech nonchalantly.
How do you think business models will adapt?
With growing intelligence in business models, they will yield more and more while costing less and less… And since money is that human invention to represent real costs, those business models will increasingly cost less energy, less environmental impact, less human and animal suffering. If humanity fully embraces this AI revolution, with 1000% dedication, curiosity, positivity, playfulness, creativity, and with respect for the tech, each other, and with respect and love for all other business models and ecosystems brought here by billions of years of evolution on this small, fragile planet, then our still somewhat primitive man-made business models will be able to integrate and assimilate with naturally evolved business models as in biological ecosystems. Result: Utopia becomes reality.
