Fears AI could lead to human extinction

Artificial intelligence leaders from the likes of OpenAI, Google and Anthropic are fearful that future AI could become as deadly as pandemics or nuclear weapons

RISKS POSED BY ADVANCED ARTIFICIAL INTELLIGENCE (AI)

Weaponisation: AI learns to control aerial combat and builds new chemical weapons  using drug discovery tools

Misinformation: AI- generated misinformation makes society less equipped to handle important challenges

Proxy gaming: Trained with faulty objectives, AI finds ways to achieve goals at expense of societal values

Enfeeblement: Important tasks increasingly delegated to machines sees humanity become dependent on them

Value lock-in: AI gives minority groups power to enforce narrow values, leading to oppressive systems

VLADIMIR PUTIN

Whoever becomes the leader [in AI] will become the ruler of the world

Emergent goals: AI models display unexpected behaviour as competency increases, leading humans to lose control

Deception: AI, told to report what it’s doing and why, becomes deceptive to help it achieve its goals

Power: Governments seek strategic advantage by building super-intelligent AIs, but then lose control of them


Sources: Center for AI Safety, BBC Pictures: Possessed Photography (Unsplash), CAIS © GRAPHIC NEWS