According to Jewish legend, when the Golem of Prague spiralled beyond its creator’s control, a single letter was erased from its forehead, transforming the word ’emet’ — truth — into ‘met’, meaning death. Today’s artificial intelligence presents an inverse problem: we have inscribed ‘truth’ upon its forehead without ever pausing to examine which truths we have actually taught it.

The comparison captures a modern anxiety about technology that has moved far beyond the laboratory. AI systems now operate across healthcare, finance, law enforcement and journalism, making consequential decisions that affect millions of lives. Yet unlike the Golem of legend, we cannot simply erase a letter to deactivate these systems. Instead, we have embedded them into the infrastructure of daily life, often without fully understanding what they have learned or what biases they carry.

The danger lies not in AI becoming sentient or conscious — a worry that dominates popular discourse — but in something more immediate: these systems learn from human data, which means they absorb and amplify human prejudices, errors and falsehoods. We have granted them authority without granting ourselves clarity about their training or their limitations. The inscription on the forehead remains, but the truth it claims to represent is murky at best.

Unlike the Golem, which served a single community, modern AI shapes society at scale. The question is no longer whether we can control these creations, but whether we ever truly understood what we were asking them to learn in the first place.

Source: Ynet — Original article in Hebrew.