Unconventional Pathways Toward Strong AI 2026

Unconventional Pathways Toward Strong Artificial Intelligence is a residential research workshop held August 31 to September 4, 2026 at Castello di Baccaresca near Gubbio (PG), in the Umbrian countryside of Italy. Organised by the NiPS Laboratory of the University of Perugia, it gathers academics and researchers to debate forward-looking, non-mainstream routes toward Strong AI and Artificial General Intelligence (AGI). This is a scholarly forum for the AI research community, not an industry expo or product showcase.

The question the workshop asks

Recent AI progress has been driven overwhelmingly by deep learning and big data, with large language models and other foundation models showing impressive breadth across many tasks. The premise of this workshop is that, despite those capabilities, such systems remain fundamentally forms of narrow intelligence, often lacking robustness, reasoning depth, long-term autonomy, and genuine understanding. The goal is therefore to explore unconventional and forward-looking research directions that might contribute to genuine machine general intelligence, examining alternative physical, computational, cognitive, and theoretical frameworks rather than incremental tweaks to the dominant paradigm. In doing so it treats the limits of today's foundation models not as engineering details to be optimised away but as a starting point for asking what genuine machine understanding would actually require.

Contributions are invited that challenge prevailing assumptions and propose novel architectures, learning principles, and interdisciplinary perspectives, drawing on insights from physics, chemistry, neuroscience, cognitive science, and embodied intelligence. The call for papers welcomes both original research and visionary perspective pieces, reflecting the workshop's interest in bold and emerging ideas about the foundations, limits, and future trajectories of AI.

Topics on the table

The scope is deliberately broad and cross-disciplinary. Themes the workshop addresses include:

• Novel architectures for general artificial intelligence and new theoretical foundations of machine intelligence
• The limitations of current deep learning paradigms
• Agent-based distributed systems
• Neuro-inspired and brain-inspired AI
• Ambient-aware, embodied, and situated intelligence
• Cognitive architectures and models of reasoning
• Complex systems and emergent intelligence
• AI informed by neuroscience, psychology, and biology

This framing places the event firmly within the strong-AI and cognitive-architecture research conversation that runs alongside the more visible 2026 wave of agentic and tool-using systems. Where much of the field is focused on engineering ever-larger models, this gathering steps back to interrogate whether those approaches can lead to general intelligence at all, and what alternative paths might.

Who attends and what's notable

The confirmed participant list is international and academic, spanning institutions such as University College London, the University of Bologna, the ISI Foundation, the Italian Institute of Technology, Northeastern University, Sapienza University of Rome, the University of Perugia, and the University of Tokyo, including a contributor associated with the Whole Brain Architecture Initiative. The mix of physicists, cognitive scientists, complex-systems researchers, and AI theorists is precisely the interdisciplinary blend the organisers are aiming for, and it is the workshop's defining strength: ideas from statistical physics, neuroscience, and philosophy of science sit beside computer science in a way few mainstream AI conferences attempt. For attendees, the value lies precisely in this cross-pollination, since breakthroughs toward general intelligence may well come from the seams between disciplines rather than from any single field working alone.

The format reinforces that ambition. Rather than parallel tracks and exhibition halls, the program is a small, single-thread residential gathering. Following arrival on Monday, August 31, the days from September 1 to 3 are structured around morning and afternoon sessions of communications and discussions, with shared meals, an excursion, and a social dinner that give participants extended, unhurried time to argue through difficult ideas. Departure is on Friday, September 4. The intimate scale and the restored medieval setting are part of the point, encouraging deep conversation over throughput.

Practical details and participation

Participation is by application, submitted through the workshop's official form, with acceptance handling accommodation reservations at the venue. The published participation fee is 790 euro, with a 495 euro rate for an accompanying person; the participant fee covers four nights, six meals, three coffee breaks, three aperitifs, the excursion, and the social dinner, reflecting the residential nature of the event. The organising committee sits within the NiPS Laboratory at the University of Perugia. For researchers working on AGI, cognitive architectures, embodied intelligence, or the theoretical limits of current machine learning, Unconventional Pathways Toward Strong Artificial Intelligence 2026 offers a rare, focused setting to test unconventional ideas against a sharp interdisciplinary audience.