AGI

In short: AGI (artificial general intelligence) would be an AI that can learn and solve almost any mental task like a human, not just one.

In more detail: Today’s AI is “weak” or “narrow”: it can write texts, recognise images or play chess very well, but only within its task. AGI would be flexible, transfer knowledge to new areas and keep learning on its own.

In Depth

• Levels: narrow AI (today) → AGI (human-like versatility) → ASI (superintelligence, clearly beyond humans).
• Status: AGI does not exist as a recognised system. There is no single definition, and experts disagree about how near or far it is.
• Questions: how do you test “intelligence”? Who is liable for mistakes? How does such an AI stay controllable (alignment)? What does it mean for work and society?
• Law: today’s AI systems fall under the EU AI Act with its risk levels.
• Distinction: current chatbots seem versatile but are not AGI. They make mistakes and “hallucinate”, so a human should check results.

See also: AI, AI phishing