What AI can now do

Plain answers to the questions people ask about AI's abilities, each graded on the same five-step ladder (demonstrated, reproduced, accessible, reliable, real consequences) and dated to when it was first shown.

• Yes Can AI clone someone's voice from a short clip?
• Yes Can AI fake a real person on a live video call?
• Yes Can AI help someone pass a job interview under a fake identity?
• Partly Can AI fool a bank's voice ID or identity checks?
• No Can AI detect whether text was written by AI?
• Partly Can AI read medical scans or replace a doctor's diagnosis?
• Yes Can AI drive a car without a human?
• Yes Can AI deceive the people testing it?
• Yes Can AI write a convincing phishing email or scam message?
• Partly Can AI make a fake ID or document good enough to fool checks?
• Yes Can AI change people's minds?
• Yes Can AI solve math olympiad or research-level math problems?
• Partly Can AI replace a customer service agent?
• Yes Can AI find security flaws in software?
• Partly Can AI hack into systems on its own?

The full record of capability milestones is on What AI can now do: every milestone.

Spot an error or something out of date? Tell us; corrections are made and logged. Written with AI assistance from public sources; every fact checked and approved by Colin Balfe. General information, not legal, medical or financial advice.