Rethinking AI: Interaction Over Isolation

Artificial intelligence has long been viewed through the lens of isolated computation, where the prowess of an individual model or system is measured by its outputs and accuracy. Yet, this perspective often misses the mark in understanding the full spectrum of intelligence, especially in the context of co-creative and interactive systems.

Beyond the Bounded Agent

Traditional AI models, whether in classical AI or modern machine learning, have focused on the individual agent as the central unit of analysis. This approach has undoubtedly led to significant technological advances but has often neglected the interactive dimensions that can drive innovation and creativity.

Recent developments argue for a shift in focus. By examining frameworks such as distributed cognition, embodied cognition, and enactive participation, researchers propose that interaction should be the primary unit of analysis. This pivot aligns intelligence with evolving dynamics among agents, environments, and socio-technical systems, rather than isolating it within computational confines.

Interaction-Centered Intelligence

What exactly does this shift entail? The proposed Interaction-Centered Intelligence framework suggests evaluating AI not just on what it produces but on how it engages. This means looking at interaction trajectories, patterns of coordination, and the adaptive regulation of systems over time. In essence, intelligence is seen as an emergent property, constantly shaped and reshaped by the interactions it fosters.

Consider systems like the Drawing Apprentice or AI Drawing Partner, which exemplify this collaborative emergence. These platforms demonstrate that intelligence isn't static but rather an ongoing interplay between creativity and interaction.

Implications for Future AI

The question now is whether this new framework will redefine AI development. Can we afford to continue ignoring the role of interaction in AI systems? According to two people familiar with the negotiations, the industry is increasingly leaning towards interaction-based models, recognizing that isolated intelligence may not suffice for the complexities of real-world applications.

this approach could revolutionize areas such as explainable AI, hybrid intelligence, and enactive AI. By prioritizing adaptive participation and engagement, the next wave of AI technologies could offer more nuanced and responsive solutions.

Reading the legislative tea leaves, it seems likely that industry leaders and policymakers will need to update their frameworks to incorporate these insights. The bill still faces headwinds in committee, but the momentum towards interaction-centered AI is undeniable.