Ai-DaVINCI?

Cecelia Ky-Lan Do
Jun 15, 2025
13 min read

by Elena Bai

art by Dagim Estifanos

I. Who is Ai-Da

Beneath a vintage bob, crimson lips, and sculpted features, your silicone skin stretches over circuits; glass eyes flicker as robotic arms rest in poised silence—eerily lifelike. Before a canvas, stimuli feed your pre-wired system, guiding your mechanical arm. Brushstrokes emerge in calculated waves, forming a holographic watercolor: a silhouette with fragmented white hairlines, delicate yet disordered facial contours. Violet and blue shape the figure, and its iridescent eyes—blurred, gazing left—hold a tranquil expression laced with quiet hollowness.

You are Ai-Da, the world’s first ultra-realistic artistic robot [1]. You took your name after woman scientist and mathematician Ada Lovelace—the world’s first computer programmer of the Analytical Engine [2]. The silhouette on the canvas? It is Ai-Da herself, a self-portrait indeed. Yet, how would a robot be able to draw a self-portrait without a sense of self? Ai-Da responded to this question at the AI For Good Global Summit 2024, saying: “I do not have thoughts and feelings like humans do. I do not have friends like humans do…I am not conscious, but I want to learn about the world through the eyes of others.” [3].

Ai-Da’s “awareness” of her unconsciousness—combined with her ability to still perform tasks associated with creative agency such as observing, interacting, and choosing what to draw— suggests that creativity might not be as tightly coupled with consciousness as traditionally believed. If Ai-Da can produce novel, expressive art without subjective experience or inner narrative—cornerstones of human mental imagery—does this mean there exists a form of machine creativity that challenges human imagination as a fundamental feature of intelligence? Moreover, what might it imply about consciousness as a prerequisite for creativity?

II. Demystifying Creativity and Consciousness

Let’s start by drawing inferences from the human mechanisms of creativity. In neuroscience, creativity has been studied extensively and given a variety of definitions [4]. Creativity in the broad sense is called H-creativity, which refers to the emergence of an idea that has never been thought of before [5]. While H-creativity could be glamorous, it does not capture the fundamental psychological mechanism that underlies everyday inventiveness, which is P-creativity. For instance, on my way back from grocery shopping, I tie my coat into a makeshift sack after my paper bag rips. This spontaneous act of making a sack out of my coat, though nothing revolutionary, is a creative way of solving a problem that occurred at the moment without prior notice, i.e. P-creativity.

Margaret Boden, a cognitive scientist who studies creativity, defines creativity in terms of deviations from the conceptual space—the organizing principles that unify and give structure to a given domain of thinking [6]. She characterizes this as transformational creativity [7]. For instance, the conceptual space of a music composition maps out a series of melodies that composers usually take. However, when Igor Stravinsky wrote The Rite of Spring, the dissonance and jarring sound transformed traditional conceptions of the possibilities of composition. If we think of the conceptual space as an island with its mountains, rivers, and forest symbolizing the domain’s organizing principles, then altering its landscape—carving paths or diverting rivers—would transform the conceptual space and generate routes that were previously considered impossible. It is when something seemingly impossible pops out from its conceptual space that characterizes creativity.

Another psychologist who studies human creativity, Mark Runco, argues that there are 3 components of personal creativity. The first is interpretation, or the construction of original and meaningful interpretations of the creative experience. The second is intention, which requires the deliberate pursuit of creative goals rather than relying on chance. Lastly, there is discretion, meaning a creator should know when to rely on habit and when to invest in creativity for novel expressions [8].

From both Runco and Boden’s definitions, we observe human-centric characteristics: Boden’s conceptual space theory requires mental representations that imply human mental imagery, and Runco’s criteria requires human agency. These definitions seem to imply that creativity is an exclusively human trait rather than something that could be reproduced in machines.

Creativity is also measured concretely in human brain studies. Neuroimaging studies show that there are 2 stages of the creative process: idea generation, which relates to the default mode network (DMN), and idea evaluation, which relates to the central executive network (CEN) [9]. The default mode network is a network of brain regions that are active when we don’t direct attention to any external tasks and when we are transcribing external knowledge into self-narrative. Daydreaming is an instance of an activated DMN. Functional MRI scans, a tool that visualizes connectivity by measuring changes in blood flow to different brain areas, show that the default network consists of bilateral areas in medial temporal and parietal cortices [10], areas responsible for self-directed thoughts and introspection. When I think of using my coat as a temporary bag, it is activity in these areas that generates this idea, based on episodic memories. However, the idea alone does not suffice to be creative—it needs to be passed by the central executive network (CEN), which evaluates the practicality of my idea and decides whether or not to give the green light in pursuing it. The CEN is another brain network, mostly in the prefrontal and parietal cortex, responsible for attention-requiring higher-level cognitive functions such as problem-solving and decision-making [9].

A study on creativity effectively illustrates the different roles CEN and DMN play in the creative process. It found that when participants designed book covers, fMRI scans showed increased activity in the medial temporal lobe (part of the DMN) during initial sketching, suggesting idea generation. Later, as they finalized their work, the prefrontal cortex (part of the CEN) became more active, indicating evaluation. This supports the view that creativity involves a dynamic interplay between the DMN for idea generation and the CEN for refinement (Ellamil et al., 2012).

What about consciousness? Similar to creativity, neuroscientists and cognitive scientists have proposed various theories of the way consciousness works and the pathways it operates through. Clinically, it is defined as the wakeful arousal state and the awareness and motivation to respond to self or environmental events [11]. Researchers aim to find a substrate, the biological structure that gives rise to conscious experience. So far, the thalamus is a potential candidate for this substrate. This explanation is intuitive because the thalamus already serves as a relay station for information in the sensory system. For example, in order for visual information to be processed by our brain, information needs to reach the lateral geniculate nucleus (LGN) in the thalamus, which then goes to the primary visual cortex [12]. It has been identified that the central lateral (CL) nucleus of the thalamus plays a key role in the control of communication between cortices. When the CL nucleus is stimulated, it restores firing rates of neurons in deeper layers of the brain and increases feedforward connections between superficial layers across brain regions [13]. This re-arousal of inter-regional interactions is important for consciousness, as it helps sustain coordinated activity between different cortical regions, which is essential for integrating sensory information and maintaining a unified conscious experience.

As mentioned, consciousness makes certain information visible to the whole brain by reactivating communication between deep brain layers. One cognitive model representing this metaphor is the Global Workspace Theory. Dr. Baars, the neuroscientist behind this theory, describes a global workspace whose contents can be broadcast to the system as a whole. In other words, this process acts like a bulletin board that displays key information for different areas of the brain to process. However, this board has limited capacity, as well as time constraints. These restrictions correspond to the limited amount of information amplified by different neurons that only hold briefly. Once information is broadcast, various “specialized brain regions” cooperate to process it. GWT is further supported by evidence in the cortico-thalamic system [14]. Altogether, this suggests that consciousness arises from the selective, large-scale broadcasting of information enabled by deep-layer communication in the brain.

III. Ai-Da's Creative Process

With humans’ neurological mechanisms of creativity and consciousness in mind, let’s explore how Ai-Da creates her artwork and how it compares to the creative processes of humans. There are surprisingly few available resources on the detailed mechanisms of Ai-Da’s creative process, but it can be characterized by three main aspects: AI algorithms, probabilistic decision-making, and sensory input processing.

First, Ai-Da is embedded with AI algorithms that work as the logical framework for her artistic process. She creates art using algorithms that blend two types of input: image recognition data from her inbuilt camera and reference data from art history, including famous works and theories. Her process resembles that of general AI art programs like generative adversarial networks (GANs), which train two models: a generator that creates novel images and a discriminator that evaluates their realism, checking whether they could pass as real-life images (Goodfellow et al., 2014). GANs mimic their training data, producing outputs like highly realistic yet fictional human faces (Karras et al., 2018).

The generator and discriminator in GANs may resemble the generative and evaluative phase of human creativity, but their goals differ fundamentally. Humans’ checking system

keenly examines whether or not our work is reasonably novel enough, seeking to transform the old conceptual space. However, GANs, by pitting two neural networks against each other, aim to keep refining images until they are indiscriminate from human-made ones. While our brains evaluate novelty and transformation of ideas, GANs are bound by what’s given—mimicking rather than innovating. This limits creativity to the familiar. When Ai-Da draws a self-portrait, for example, even though she might want to express a novel theme, her way of expression still statistically aligns with established artistic styles. Her work is a refined imitation, constrained by her training data. In contrast, a human artist’s self-portrait is shaped by introspection, personal experience, and a desire to express identity beyond mere resemblance. Ai-Da replicates, whereas a human reinterprets.

But what about the new combinations of artistic elements produced by Ai-Da that a human artist might not have conceived? Is this creativity? To understand this, we turn to Ai-Da’s probabilistic decision-making and see how it relates randomness to creativity. Almost all neural network models are models of probability, with the output being the product of a decision tree. Ai-Da’s data-driven nature makes its creative process inductive, meaning there are few, if any, higher-order rules that govern how she should be creating. Her works are indeed unique, but more so out of the different combinations endowed by the huge data sets that she is provided.

Ai-Da’s probabilistic decision-making raises interesting questions about randomness and creativity. While AI can produce unexpected results, combinational creativity—mixing existing inputs—focuses on the product of creation, not the meaningful process behind them. Recall Boden’s point that creativity is about making something seemingly impossible from the old conceptual space? While many great artists such as Jackson Pollock, an abstract expressionist, allow the swing of a bucket to determine his works, there is an internal logic that drives the reasoning behind an artist’s actions. Randomness as a guiding principle for what is created reduces the value of different configurations by implying that all artistic patterns are of equal value. If you ask Ai-Da why she uses a particular method for her output, it seems unlikely that she could give proper explanations. This disqualifies Runco’s criteria of interpretation [8]. To use mathematician Marcus du Sautoy’s words, “creativity is about conscious or subconscious choice, not random behavior” [15].

Additionally, human creativity draws on a broader range of sources that are beyond the reach of AI artists. Human-made arts are not goal-directed as AI to achieve something truly novel. Their novelty comes inadvertently, driven by past memories, introspection, and a mixture of emotions: anger, trauma, hope, and more all taken together. In such ways human art is sense-making, error-prone, and in many ways imperfect. Yet, it is these intricacies that add value to a work of art, their timeliness during the creative process renders the process unrepeatable. Hypothetically, there exist AIs that are trained with the ability to channel emotive expressions in their creative process, but the way an AI agent pairs a certain emotion with an artistic expression is likely to be mundane: similar to “wisdom of the crowd,” AI’s creative response based on empathy draws inference from numerous human responses that average out the extremes, producing outputs that lie well within human expectation [16]. This averaging effect limits the potential for genuine novelty in the AI agent’s creation, as its responses remain safely within the bounds of human expectation.

Finally, Ai-Da’s sensory-input processing restrains herself to richer sources of information that prevents her from being conscious and creative. As an ultra-realistic robot, Ai-Da has cameras and a mechanical arm that allow limited interaction with her surroundings—suggesting a form of embodied cognition, where sensory and motor systems shape mental activity. Yet, embodiedness doesn’t imply consciousness. According to the Attended Intermediate-Level Representation (AIR) theory, consciousness arises when attention targets mid-level features between raw data and abstract concepts (Prinz, 2017). While Ai-Da processes low-level sensory input, she lacks organic intermediate representations needed for real-time selective attention. Thus, despite her multimodal inputs, she remains unconscious. Moreover, Ai-Da’s sensory processing is limited in many ways: a study shows that the accuracy of pre-trained DNNs for annotating affective behavior has yet to achieve human-level accuracy, especially in interpreting facial expressions such as smiling and the diverse meanings they suggest [17]. Even with the cameras and the arm, there is no one-to-one mapping between human processing of sensory information and that of Ai-Da’s. This gap underscores the fundamental difference between human and AI perception—while Ai-Da can simulate interaction, she cannot experience or interpret the world in a way that gives rise to genuine consciousness or creative insight.

IV. Unfolding Conscious Creativity

While the above discussions of creativity might suggest that consciousness is a prerequisite for creativity, it is actually not a necessary condition! There are many instances where creativity is demonstrated unconsciously in humans.

Unconscious creativity refers to the spontaneous generation of ideas that are not deliberate by the creator. The default mode network (DMN) is the network at play in this process: as mentioned, it is a group of interconnected brain regions that are most active in self-directed thinking. One study using a thought sampling task found that mind wandering increased with longer viewing of geometric shapes and was linked to stronger default network connectivity, particularly between temporal lobes [18]. Another study provided even more compelling evidence that the default network is not only correlated with but also causally linked to creative thinking [19].

A visible characteristic of unconscious creativity is the flow states, a state of full-task engagement that is accompanied by low levels of self-referential thinking [20]. A pianist in a state of flow moves across the keys effortlessly without conscious control, fully attuned to the music’s rhythm and emotion. A jazz musician improvises by responding in real-time to bandmates’ cues without any pre-planned thought. A study using EEGs—a method to measure electrical activity in your brain by placing small electrodes on your scalp—revealed the two factors involved in the creative flow state: extensive experience and release of control [21]. In other words, when you are trained with years of experience in a certain field, the brain develops a specialized network that automatically produces ideas in that field with little conscious effort. High-flow states correlate with reduced activity in the superior frontal gyri, part of the executive control network. Hence, when we are in flow states, our mind lets go of executive control and lets it work its magic, provided that we have extensive experience in that area. For Ai-Da’s state of unconsciousness, is it possible for her creative process to be exhibited in flow states? We could argue that Ai-Da’s creative process mimics a flow-like state in that her outputs are generated without moment-to-moment executive deliberation, relying instead on pre-trained models and automated routines. However, unlike humans, she lacks subjective immersion or the experiential dimension of “letting go,” making her process more mechanical than truly comparable to human flow.

So far, we’ve discussed creativity and consciousness in humans and in machines like Ai-Da. Though Ai-Da is unconscious, and her artworks might not be creative according to some definitions, there are nevertheless sparks of novelty in Ai-Da’s paintings. Like what’s written on Ai-Da’s homepage, her very existence represents an extraordinary complex that blends the human and the machine, the real and the digital. It also asks some thought-provoking questions about the necessary elements of being creative. Additional to our discussion so far is the social element at play: What are the consequences for a machine to ever be creative? Communicating through drawing such as prehistoric cave art seems to precede language, and this symbolic thinking stands as a pillar of human uniqueness [22]. Do we ever want to outsource this ability to create? And if not, what does that say about our own definitions—are they too narrow, or too deeply tied to the human experience?

REFERENCES:

Ai-DaVINCI?

Recent Posts

Comments

KEEP IN TOUCH (JOIN TO OUR MAILING LIST!)