‘Extinction risk’: Could code-writing AI wipe out humans via software backdoors?

Industry luminaries are warning of near-imminent doom unless AI is tamed. Given that today’s generative AI models are writing semi-decent code, shouldn’t we worry we’re preparing the ground for Skynet?

ai models artificial intelligence (ai) backdoors code dev & devsecops doom generative generative ai humans industry machine learning (ml) near risk secure software blogwatch skynet software warning writing