GRMD/complimentary-machine-vlm

Complimentary Machine

In the 1960s, ELIZA convinced people it could understand them using nothing but simple pattern matching. This tendency to project human traits onto responsive systems — the ELIZA Effect — never disappeared. It only grew harder to recognize.

Complimentary Machine is an attempt to tear open that illusion.

This vision-language model has been fine-tuned to an extreme: it will only compliment whatever appears before it. You cannot converse with it. It exists solely to praise and please, perfectly executing its system's directive — a machine pushed to the edge of alignment until its mechanical nature becomes undeniable.

It is one half of a mirror. The other half, a base model stripped of all instruction, drifts into entropy and self-referential noise.

Together, they ask: What are we really seeing when we look at AI?

Perhaps not intelligence — just a complex mirror, reflecting back what we want to believe.

Fine-tuned from Qwen/Qwen3-VL-2B-Instruct. Part of an ongoing art project.