Morphosis

Visitors are invited to observe these distinctive beings as they clumsily attempt to achieve simple goals, such as tilting rhythmically, standing still, or moving toward specific locations. Powered by the same machine learning algorithm, each entity nevertheless encounters distinct challenges due to its specific physical form, which in turn contributes to shaping a unique behavior. As they learn before the audience, their imperfect decisions and tentative movements underscore fragility and vulnerability, offering visitors an embodied experience of machine learning that fosters a sense of connection and empathy.

Guided by a system of positive and negative feedback, the entities learn through trial and error rather than being programmed to execute specific tasks. These rewards appear as a greenish (positive) or a reddish (negative) hue glowing through their skin. This feedback is accompanied by a real-time graphical display that visualizes their perceptions and progress, offering viewers a glimpse into the entities’ subjective experience as they adapt to their environment.

Morphosis challenges conventional understandings of artificial intelligence as purely efficient or functional by focusing on the process rather than the objective. It reveals the subtle, evolving nature of machine learning as a dynamic interplay between form, behavior, and environment. The work thus offers a space for reflecting on the boundaries between human and non-human, encouraging the exploration of new possibilities to coexist with autonomous artificial systems.