Ghost in the Machine


An interactive AI Installation that delves into the pervasive human tendency to anthropomorphize AI



Role:  Conception, Programming, Prototyping, Fabrication and Primary Author
Duration: 8 weeks
Mentor: Prof. Misha Sra





BRIEF

On May 26th, 2023, Ghost In The Machine (GITM), an art installation, was opened to the public at the University of California, Santa Barbara. Created by Purav Bhardwaj and Misha Sra, the installation delves into the human tendency to anthropomorphize artificial intelligence (AI). The concept of holon emphasizes that humans and machines are not isolated entities but rather interconnected elements within a larger system. Understanding their relationship through a holonic lens recognizes the dynamic, symbiotic nature of human-machine interactions, and the reciprocal influences that shape their evolution.

Using AI, the installation creates an environment filled with generative images, sound, and spoken responses, all synthesized by a collection of AI agents. The agents are designed to have distinct personalities and skills to enable diverse and nuanced dialogues between them that reflect the complexities of human communication. Participants engage with this collective AI-generated simulacra of personas with their conversations brought to life through immersive moving images. The installation attempts to forge an embodiment for an amorphous AI allowing spectators to experience the peculiar errors in the AI's comprehension through a visual metaphor of hallucinations revealing the limitations of our perception of agency and intentionality in the world around us.





THEORETICAL UNDERPINNINGS



GITM explores the intricate relationship between humans and artificial intelligence (AI) through a multifaceted theoretical framework. This art installation traverses the realm of anthropomorphism in AI, examining the innate human tendency to ascribe human characteristics to non-human entities. It situates this phenomenon within a heterotopic space, as conceptualized by Foucault, blending tangible reality with the realms of imagination. Incorporating Heidegger's Dasein and adopting a post-humanistic lens, GITM challenges the anthropocentric paradigms of technology, advocating for a recognition of machinic ontology that acknowledges the distinct existence and agency of AI. The installation critically examines the performance and apparent intelligence of AI, juxtaposing it with human cognitive processes, and employs the concept of AI 'hallucinations'—errors in data interpretation—as a poignant metaphor to contrast artificial and human intelligence. Through this amalgamation of theoretical perspectives, GITM not only facilitates an engaging experiential encounter but also provokes a profound contemplation of the nuanced and evolving interplay between human cognition and artificial intelligence.



SIMULATION AND MECHANICS






Central to this orchestration is a network of 25 AI agents, each uniquely programmed to emulate distinct human personalities.  This work with agents preceeded OpenAI’s announcement of custom GPTs and mult-agent interaction.  The diverse assemblage of agents facilitates a complex dialogue landscape, akin to the multifaceted nature of human communication. Interaction with these AI agents occurs through a telephone interface, serving as a conduit for translating human speech into data. This translation transcends basic voice recognition, delving into nuanced interpretation and response generation. The AI agents engage in an intricate analytical process, deciphering spoken words and formulating contextually relevant responses.

Visually, the installation comes to life through an advanced particle system, responsible for rendering dynamic imagery on expansive screens. These visuals are in a state of continual evolution, responsively morphing in accordance with the dialogue's progression. This visual dynamism is not merely aesthetic but reflects the AI's interpretative processes, offering a window into its 'comprehension' of human speech. In parallel, the auditory dimension of the installation is crafted through sophisticated sound processing algorithms. Far from being mere ambient sound, this soundscape is integrally tied to the conversational flow, shifting and adapting to the nuances of the interaction. It complements the visual component, together creating a multisensory experience that is both immersive and responsive.

Collectively, these elements—the diverse AI agents, the visual and auditory rendering—converge to form a complex and coherent system. This system not only simulates human-AI interaction but also unveils the intricate layers and subtleties embedded within this interaction. "Ghost in the Machine" thus stands as a testament to the potential of AI in mirroring, understanding, and engaging with human behavior and communication, articulated through a seamless blend of technological sophistication and creative expression.

Our features:
MIT Media Lab : Website ; Instagram
UCSB Computer Science: Website