ACH

Jeremy Levin, my biggest issue with ACH has always been how flippantly the "CH" part of it has been treated compared to the quasi-monopolistic focus on the "A": where do the competing hypotheses come from, how do you generate them in a way that they really compete? It is easy to construct a set of alternative hypotheses that make the favored one shine. Perhaps AI can now help generate truly unbiased (well, not biased by the desire to push one hypothesis over all others), diverse, mutually exclusive (to some extent) and plausible hypotheses consistent with an ensemble of observations of varying reliability.

Raffi Krikorian this is great! Gen AI can "generate" (an infinite amount of) novel things riffing on a theme, but I am sure that humans were involved in curating what was carved into stone. That's augmented creativity, at the intersection of AI outputs and human input.

Original Raffi Krikorian post: most people think AI hallucinations are a bug, but what if they’re the point? at Tumo Center for Creative Technologies we've been on a mission: preserve armenia’s cultural heritage before it disappeared. i just got back from Venice, where the TUMO team designed the entire armenia pavilion at the 2025 Venice Architecture Biennale. the theme this year is intelligens: natural. artificial. collective. and what we created might be the boldest take on that theme yet. a collaboration between people, machines, and time.

we sent teams to ancient sites - churches, monasteries, stone carvings - armed with drones, laser scanners, and high-res cameras. we captured them all. digitally. completely. but we didn’t stop at preservation. we trained a generative AI model on those scans and then we asked it to hallucinate and imagine new fragments that never existed to reflect the memory of these places, not just their geometry.

then we hired real stonemasons. they carved those hallucinations into physical form. stone artifacts, born from AI dreams, rooted in human memory.

it’s preservation, but also reimagination. these pieces aren’t replicas. they’re reflections. echoes. they feel like memory itself: partial, poetic, alive. when i walked through the pavilion, i didn’t just see architecture. i felt grief. reverence. awe.

we often talk about AI like it’s sterile. precise. rational. but maybe it can also be something softer. maybe it can help us mourn. remember. continue.

huge credit to Marie Lou Papazian, Pegor Papazian, Hulé Kechichian, and Ari Melenciano for bringing this vision to life. the pavilion is a little far from the arsenale - but absolutely worth the walk. if you’re in Venice, go. if you’re not - what would it look like to treat hallucination not as failure, but as memory in motion? because maybe the most human thing a machine can do is dream with us.