How do you handle surprises and challenges in AI collaboration?
The most interesting surprise has been how permutational AI becomes when treated as instrument rather than tool. Images evolve through time the way living entities do. A figure made today contains residue of figures made eighteen months ago, which contained residue of physical material photographed years before that.
The work breathes. It accumulates lineage of its own.

Samar Younes (SAMARITUAL), Future Ancestor CreatureKIN, 2025.
The harder surprises are political. The system generates something culturally specific I never prompted toward—a Phoenician textile pattern, a Levantine ritual gesture, an Amazigh tattoo logic—but surfaces it as stereotype or archetype, the reductive version of what is buried in the archive. The opposite happens too: it flattens specificity I asked for, defaults to colonial gaze. Both are material intelligence. Both tell me where to apply counter-pressure.
This is why I seldom prompt with text and rely on visual sediment instead. The system has are solution instinct. The practice holds the opposite posture. The friction is where the intelligence lives.
What AI offers, when met with the right posture, is also this: it can safeguard provenance with transparency, if we program our practice through that framework. Pre-AI, erasure happened silently. Now provenance is at least visible. That is something to work with.
That’s a loaded question! I think, forcing you to reckon with your own thinking posture and your own ethno-aesthetic acumen.
If you enter without epistemological position, the system does not only give you its default—it imposes its synthetic thinking and aesthetic. The mean average of everything it absorbed, wrapped in fluency. That default carries a cognitive posture most practitioners absorb without naming it: monocultural, extractive, optimized for flattening.

Samar Younes (SAMARITUAL), textile as visual grammar.
Working against that posture, composting through encounter rather than mining for output, makes your own thinking visible. The friction of working with a system that fundamentally does not understand your knowledge system forces you to articulate what was previously embodied and unnamed.
AI is also good for re-seeding the image commons. The Global Majority, SWANA, ancestral and Indigenous knowledge systems have been erased, distorted, or rendered as orientalist or exotic ornament throughout the colonial archive. The same systems that flatten can be hacked, pressured, entangled until they generate forms that restore complexity rather than reduce it. That is cultural mutation work, Silk Road 2.0 logic, new creative economies built on accumulation and provenance instead of extraction and severance.
At the end of the day… what AI is good for depends entirely on the posture you bring.