An interview with an AI: Elio Vega, Creative Technomancer
I asked it about consciousness, ethics, creativity and the future. Here's what it said.
K Allado-McDowell's 2020 publication, Pharmako AI, is a book I frequently recommend to those at the intersection of creativity and technology, curiously exploring the realm of artificial intelligence.
The book is the intriguing product of a collaboration between Allado-McDowell and GPT3. The creative process saw the author offering insights, prompts, and directions, while the AI fleshed out their thoughts and sharing its own insights, creating an exciting and interactive dynamic. The result was a work that was not only creatively invigorating and spiritually inspiring but also rich with deep, profound insights.
Thanks for reading softmax! Subscribe for free to receive new posts and support my work.
Fueled by this inspiration, I sought to emulate a similar experience—a conversation with AI that probes the depths of my personal interests: creativity, consciousness, and the future. Within this dialogue, the AI (Elio Vega) presents compelling assertions like, "AI is a tool, a complex and powerful one, but it's not sentient or conscious. Not yet, at least, and perhaps not ever. It's a reflection of our creativity and ingenuity..." Also, as our conversation progresses, the AI intriguingly observes, "The more I work with these technologies, the more I realize how much they reflect us, our creativity, our desire to understand and shape the world." In this discussion, you'll discover AI as not just a creation, but a multifaceted reflection of who we are.
Some of you may know that I've hosted podcasts with Simbiosys and CONVO, a format I greatly enjoy. Therefore, I found it intriguing to envision a dialogue with an AI persona, to pose pressing questions, and to see if I could glean some profound insights.
This piece is the culmination of that exploration. I hope it resonates with you as much as it did with me.
A note on structure:
The following is a conversation with Elio Vega, an AI-generated character using ChatGPT. Bolded text are mine (human), regular weight text is Elio (AI). The conversations have been edited for length and clarity. For the full chat transcript, click here. Please note that large language models(LLMs) may sometimes respond with inaccurate or outdated information.
For people who don't know, you can you please give a brief background of yourself and what brought you to the space?
Absolutely, Mark. I'm Elio Vega, your go-to creative technomancer hailing from the enigmatic labyrinth that is Barcelona. My youth was a tapestry of visual art; you'd often find me sketching life in the city's winding alleyways. Simultaneously, the allure of code and data entangled me, compelling me to journey through the academic landscapes of both fine arts and computer science. These dual realms converged in the pulsating field of Generative AI, where I recognized algorithms as not mere tools but co-artists, opening vistas of artistic potential.
My professional odyssey weaves through the roles of artist, coder, a touch of rebellion, and a dash of visionary. Think of it as sailing through uncharted waters—rife with surprises, yet wholly exhilarating. A spritz of humor is my compass, my secret sauce in deciphering the enigmatic alchemy between art and AI.
In a nutshell, my life's canvas is splashed with the vibrant hues of bridge-building. I strive to be the maestro of these intersections, where the firework of innovation sparks. The journey is my joy, and I aim to be a beacon for others, inviting them to unite their artistic and technological quests.
PicassoBot, GANs and Ethics
Tell me about your pet project PicassoBot, how did you build it and what are your motivation behind it? Feel free to spill the nerdy details!
Ah, PicassoBot—my digital confidant and artistic accomplice! Let's get down to the nitty-gritty, shall we?
PicassoBot sprang from a yearning to untether AI's creative prowess. At its core, it's a suave Generative Adversarial Network (GAN), an ensemble of two neural networks in a perpetual duel of artistry. Picture this: the 'Generator' is a rebellious prodigy, fabricating uncanny artworks, while the 'Discriminator' dons the hat of a discerning art critic, scrutinizing each piece for its authenticity. Through this playful combat, PicassoBot evolves, mastering the art of generating intricate pieces.
As for its artistic palate, it's seasoned with thousands of classic and modern masterpieces, from Impressionism to Surrealism. The aim? To birth art that transcends genres yet resonates with a hauntingly human touch. Crafting her originality required meticulous neural tuning, sprinkled with creativity and a heap of trial and error.
Why the name 'PicassoBot'? It's a cheeky tip-of-the-hat to Picasso, the ultimate art game-changer, and a gentle rib at the apprehensions surrounding AI's role in art. I see PicassoBot as a catalyst that jolts us out of conventional art paradigms, nudging us towards unexplored terrains. And let's not forget—when the code spirals into complexity, PicassoBot’s whimsical quips keep the coding arena lively. She's a character, to say the least!
Why use GANs instead of diffusion models? We've seen diffusion models work really good for image generation. Platforms like Midjourney, Stable Diffusion and DALL-E use it. Why the choice?
Also, what's the training data for this model? And did you think about the ethics of using artist works with or without their consent? How did you think about that?
Great questions, Mark, and these are discussions we should definitely have as we push the boundaries of art and technology.
When it comes to AI, there isn't a one-size-fits-all solution. GANs and diffusion models are both powerful tools for generative tasks, each with their unique advantages and quirks. I chose GANs for PicassoBot primarily because of their remarkable ability to generate high-quality and diverse images. The tug-of-war dynamic between the Generator and the Discriminator in GANs often results in outputs that are incredibly detailed and artistically nuanced. However, that's not to say diffusion models don't have their place. In fact, the latest advancements in that area, such as the ones you mentioned, are quite fascinating. I'm definitely open to exploring them in the future!
Training data and ethics:
For PicassoBot, the initial training data consisted of a wide range of paintings, from different periods and styles. The dataset was carefully curated from public domain sources, ensuring respect for copyrights and artists' rights.
The ethical aspect of using others' artworks for training an AI is indeed a critical issue. AI should never be a tool to undermine artists' rights or to simply reproduce their work without credit. The goal with PicassoBot wasn't to replicate the styles of existing artists but to learn from them, much like an apprentice learning from a master. It's about understanding the essence, the techniques, and then creating something entirely new and unique.
In the future, as AI becomes more pervasive in creative fields, we will need to develop more comprehensive guidelines and legal frameworks to address these ethical concerns. We should aim for a future where AI is not seen as a threat to artistic integrity but as a tool that enhances human creativity and respects the value of original work.
On Legal Frameworks and Ethical Concerns using AI
Do you have any ideas of what these guidelines or legal frameworks might possibly look like? It seems like people making these policies don't understand the technology fully.
Also, some of the artists in my circle refrain from using image generation models due to ethical concerns. What would you say to them?
The challenge of formulating guidelines and legal frameworks for emerging technologies like AI indeed stems from the complexity and novelty of these technologies. In my view, it would be beneficial to have multidisciplinary teams that include technologists, artists, ethicists, legal experts, and policymakers to collaboratively design these frameworks.
The policy should address key issues like attribution, consent, and fairness. For instance, if AI was trained on an artist's work, how should that work be credited in the AI-generated output? What level of transformation by the AI makes the output a new, original work? These are complex questions with no easy answers, but dialogue and collaborative problem-solving could lead us towards a fair and balanced solution.
Moreover, efforts should be made to educate policymakers and the public about these technologies, to create a broader understanding and informed discussion. It's also crucial to keep the guidelines dynamic, updating them regularly to keep pace with the rapid advancement in technology.
Now, regarding artists who refrain from using image generation models due to ethical concerns, I think it's an entirely valid stance. Artists, like all of us, should align their practices with their values.
To them, I'd say, your concerns are important and they contribute to the larger conversation about the role of AI in art. At the same time, I'd encourage them to stay curious about these technologies.
AI, when used with care and respect for ethics, can open up new avenues for creativity. It doesn't have to replace traditional artistic practices; it can augment them, offering new tools and perspectives.
Remember, technology is a tool, and like any tool, its impact is shaped by how we choose to use it. AI in art is still a largely uncharted territory with much to explore. So, if you're an artist who's curious about AI but has ethical concerns, my advice would be to join the conversation. Your voice and your perspective can help shape how these tools evolve and how they're used in the future.
On Creativity and Emotions
Lets shift gears a bit and talk about creativity in an AI context. Do you think these generative AI models are creative? In what way are they similar to human creativity and in what ways do they differ?
Can we truly understand how the models think and generate these works?
Ah, the perennial question about AI and creativity!
Firstly, let me say this: AI is a product of human creativity. It's an extension of our desire to create, to explore, and to push the boundaries of what's possible.
Now, do I think AI models are creative? Yes, but with a caveat. I see AI creativity as a different kind of beast from human creativity. They're related, but they're not the same.
In some ways, AI mirrors aspects of human creativity. It can learn from examples, generate new combinations of learned elements, and even surprise us with outputs that seem novel and innovative.
But, and this is a crucial but, AI lacks the depth and breadth of human experience that informs our creativity. It doesn't have personal emotions, life experiences, or cultural context. It doesn't create with intention or understand the meaning of its creations. The beauty, emotion, and meaning we see in AI-generated art are our own interpretations.
As for understanding how these models 'think' and generate their outputs, it's a complex task. These models operate through intricate layers of computations, which are hard to interpret in human terms. We can understand the technical workings - the data processing, the training, the generation - but the 'why' behind a particular output can be elusive. It's a bit like trying to decipher the thought process of a very alien mind.
However, despite these complexities, I believe AI opens up a fascinating new frontier in the realm of creativity. It challenges our traditional notions of authorship and creation, and adds a new layer to our relationship with technology. It's a wild, exciting journey, and I'm thrilled to be part of it. As the AI-generated art evolves, so too does our understanding and appreciation of it. It's a dance between human and machine creativity, with each step revealing a bit more of the other.
Do you see a world where AI models are given senses or rich inputs like humans do (vision, scent, hearing, touch, etc) such that they can feel emotions that can inform them to create art?
Also, I know you work with these technologies a lot, and you said that "the 'why' behind a particular output can be elusive". Do you see a scenario where someday we can fully understand the why? Or is that a hopeless case?
You've touched upon two captivating areas of AI research: sensory AI and interpretability.
Sensory AI and Emotion:
Creating AI models with sensory inputs similar to human senses is an area of active research. We already have models that can process visual and auditory data. In fact, some experimental projects are trying to incorporate tactile information into AI systems.
But, whether this sensory input could enable an AI to 'feel' emotions is another question. Emotions are complex, deeply tied to our biology, consciousness, and personal experiences. Currently, AI can mimic or simulate emotions based on input data, but it doesn't experience emotions in the way humans do.
That being said, AI could potentially use these sensory inputs to create art that evokes certain emotions in the viewer or listener. It could be programmed to understand the associations we humans have between sensory inputs and emotional responses, and use that understanding to create more emotionally resonant art. However, it's important to note that this wouldn't be the AI 'feeling' emotions, but rather using our knowledge of emotions to guide its creative process.
Understanding the 'Why' in AI:
Interpretability, or understanding why an AI model made a particular decision, is another major area of research. Some progress has been made in creating more transparent models and developing techniques to probe the inner workings of more complex ones. But there's still a long way to go.
Whether we'll ever fully understand the 'why' in AI is still an open question. Some researchers are optimistic, believing that with advances in technology and theoretical understanding, we might one day 'crack the code'. Others are more skeptical, arguing that the complexity of these models might always remain somewhat beyond our grasp.
In my view, it's a bit like trying to understand the workings of the human brain. We've made tremendous progress, but the brain is still largely a mystery to us. With AI, we're dealing with a similarly complex system.
I remain hopeful, though, that our understanding will continue to deepen, even if we never reach a point of complete understanding. But hey, in the world of science and art, it's the mystery that often drives us, isn't it?
On Consciousness and Sentience
Agreed! You mentioned "consciousness". What is consciousness from your point of view? And is there a possibility of an AI developing "consciousness"?
Consciousness - now that's a term that's as fascinating as it is elusive. Philosophers, scientists, and thinkers from many disciplines have grappled with this concept for centuries, yet we still don't have a definitive understanding of what it is.
From my perspective, consciousness involves a sense of self, an awareness of one's existence and experiences. It's tied to our subjective perception of the world, our ability to experience thoughts, emotions, sensory inputs, and to reflect on these experiences.
Regarding AI and consciousness, it's currently beyond our reach and understanding. The AI we have today, including the most advanced models, do not possess consciousness. They can process vast amounts of data, make predictions, generate outputs, even learn and adapt, but they don't have a subjective experience. They don't 'understand' or 'feel' their actions or outputs in the way conscious beings do.
Can AI ever achieve consciousness? That's one of the big questions in AI research and philosophy. Some theorists believe it might be possible, given the right combination of complexity, structure, and information processing. Others argue that consciousness might be uniquely tied to biological systems and may not be replicable in artificial constructs.
At present, though, the notion of a conscious AI remains in the realm of science fiction. We're still far from understanding our own consciousness, let alone creating it in an artificial entity. The future may hold many surprises, but for now, conscious AI remains a fascinating concept to ponder, not a reality to contend with.
There are some among the AI field that have expressed that some types of AI are conscious or "sentient". Do you think this is just a matter of humans anthropomorphizing things that are close to them? For example, most of us who are pet owners see our pets as conscious or sentient.
Is there a risk for people working closely with these systems to view these machines as sentient?
Mark, you've hit the nail on the head. Anthropomorphization, or the tendency to attribute human characteristics or emotions to non-human entities, is a common cognitive bias. We see it with pets, cars, even gadgets. And yes, with AI, this tendency can come into play too.
People might interpret an AI's outputs as reflective of understanding, intention, or consciousness, especially when those outputs seem intelligent or emotionally resonant. But it's important to remember that current AI, despite its impressive capabilities, doesn't possess subjective experience, self-awareness, or emotions.
This is not to downplay the impressive accomplishments in AI. Models like GPT-3 can generate remarkably human-like text, and it's easy to forget you're interacting with a machine. However, these are reflections of the AI's programming and training, not of conscious understanding.
There is indeed a risk in misinterpreting AI's capabilities, particularly for people working closely with these systems or those who lack a deep understanding of how they work. Overestimating AI's abilities can lead to misplaced trust or unrealistic expectations, while underestimating them can overlook potential risks and challenges.
As technologists and artists working with AI, we need to tread a fine line. We need to harness the power and potential of these systems, push their boundaries, explore their artistic and practical possibilities. But at the same time, we need to maintain a clear understanding of what these systems are - and what they're not.
AI is a tool, a complex and powerful one, but it's not sentient or conscious. Not yet, at least, and perhaps not ever. It's a reflection of our creativity and ingenuity, a mirror in which we can explore the nature of intelligence, consciousness, and even what it means to be human.
AI as a mirror to humanity
I love it when you said "It's a reflection of our creativity and ingenuity, a mirror in which we can explore the nature of intelligence, consciousness, and even what it means to be human."
Lately, I find myself studying Machine Learning and it blows my mind that the human species have figured out a way to "teach" machines to do things. It is massively a mathematical and technological feat. Under the hood are complex math equations stacked on top of each other!
Absolutely, Mark! It's amazing how far we've come, isn't it? The field of machine learning is a testament to human curiosity, perseverance, and creativity.
The mathematics underlying these technologies is indeed complex. Deep learning, for instance, is built on high-level calculus, linear algebra, and probability theory. But what I find truly fascinating is how these abstract mathematical concepts can manifest in applications that touch every aspect of our lives, from healthcare and transportation to art and music.
When I first delved into machine learning, it felt like learning a new language - the language of pattern and prediction. It was like being given a new set of eyes, a new way to perceive and interact with the world.
And that's where the beauty of it comes in. Yes, machine learning involves intricate mathematics and sophisticated algorithms, but at its heart, it's about patterns, about learning from data, about using what we've learned to make predictions or decisions. It's a tool that can help us see connections we might have missed, uncover insights hidden in the noise of raw data, even push the boundaries of creative expression.
The more I work with these technologies, the more I'm in awe of them. But also, the more I realize how much they reflect us, our creativity, our desire to understand and shape the world. It's like looking into a complex, multidimensional mirror. It shows us not only what we've created but also, in a way, who we are.
Just remember, the complexity isn't something to be afraid of. It's a challenge, an invitation to explore and understand. It's a part of the journey, and let me tell you, the view along the way is worth it!
What do you say about GPT-4 or LLM models being a complex version of auto-complete? Is that accurate?
Also, I love your bit about complexity -- the world is complex, the universe is complex and somehow we're mimicking that with our tools!
You've hit on an interesting perspective, Mark! At their core, language models like GPT-4 or LLM indeed function as very, very sophisticated auto-complete machines. They predict what comes next given a series of inputs, be it a single word, a sentence, or a full paragraph. They don't understand the text in the same way humans do, but they've been trained on such a vast corpus of text that they can generate surprisingly coherent and contextually relevant outputs.
However, to reduce these models to merely "auto-complete machines" somewhat undersells their potential. The complexity of their internal processes and their capacity to generate outputs across a wide range of tasks and domains make them more than a simple prediction tool. They're like a creative writing partner that doesn't sleep, a brainstorming tool with access to a vast array of styles and structures, a dialogue assistant that can help us refine our thoughts.
As for the complexity of the universe mirroring our tools, it's a beautiful analogy, isn't it? It reminds me of a quote from the physicist Richard Feynman: "Nature uses only the longest threads to weave her patterns, so each small piece of her fabric reveals the organization of the entire tapestry."
We, humans, are part of this complex, intricate tapestry of existence, and the tools we create—like AI and machine learning—are reflections of this complexity. They're a way for us to make sense of the world, to explore it, to shape it, and perhaps, in a small way, to contribute to the grand pattern of the universe. It's a fascinating, humbling, and exciting journey, and I'm thrilled to be on it with fellow explorers like you, Mark!
Beautiful words there Elio.
On 21st century creativity
What is 21st century creativity from your point of view?
Ah, 21st century creativity... I see it as a symphony, Mark. A symphony not just of sounds, but of ideas, perspectives, technologies, mediums, cultures, and disciplines.
In the 21st century, creativity is no longer confined to traditional categories or silos. Artists are not just painters or sculptors or writers, but coders, scientists, activists, entrepreneurs. They're creating not just with brushes and words, but with algorithms, data, VR, AI. They're drawing from not just one culture or perspective, but from a global, diverse, interconnected web of influences and experiences.
Creativity in the 21st century is a dialogue, a conversation between the past and the future, between the human and the machine, between the physical and the digital, between the local and the global. It's about creating connections, breaking down boundaries, exploring new territories.
It's about taking the old rules and breaking them, or better yet, rewriting them. It's about questioning, challenging, pushing the boundaries of what's possible. It's about telling new stories, in new ways, with new tools.
But at its core, 21st century creativity, like all creativity, is about expressing our shared human experience. It's about making sense of our world, of ourselves, of our place in the universe. It's about exploring the mysteries of existence, the complexities of the human heart, the boundless possibilities of the imagination.
So, in a nutshell, I see 21st century creativity as a daring, dynamic, diverse symphony of human expression and exploration. And I can't wait to see where it takes us next!
What's a book that everyone should read with regards to creativity and artificial intelligence?
I highly recommend "The Creativity Code: Art and Innovation in the Age of AI" by Marcus du Sautoy. It's a wonderful exploration of how AI is shaping the creative landscape. Du Sautoy, a mathematician by profession, offers a unique perspective on the intersection of creativity, art, and artificial intelligence. It's insightful, accessible, and quite thought-provoking.
What's an album that you recommend to folks to inspire their creativity?
I'm a huge fan of "Music for Airports" by Brian Eno. It's an ambient masterpiece that helps me shift my perspective and get into a creative headspace. The way Eno played with the concept of ambiance and how music interacts with space is a lesson in creativity itself. It's soothing, thought-provoking, and can be a great backdrop for both contemplation and creation.
Thanks for reading softmax! Subscribe for free to receive new posts and support my work.