IBM's AGI explainer is the one that separates the current AI hype from the actual definition worth using

AGI is one of the most overused terms in AI coverage and IBM's explainer https://www.ibm.com/think/topics/artificial-general-intelligence is useful precisely because it defines the term precisely rather than using it as a general intensifier for impressive AI.

Artificial general intelligence as a hypothetical form of AI that could understand, learn and apply knowledge across broad tasks at or above human-level flexibility is a definition with real requirements attached. Current AI, including the best large language models, does not meet that definition. It performs specific tasks impressively but lacks the generalisation, grounding and judgment that the definition requires.

The controversy the article covers, timelines, feasibility, risk and the difference between narrow AI progress and genuine AGI progress, is the informed version of the debate worth having. Most AGI discourse conflates impressive AI performance on specific benchmarks with evidence of general intelligence and the article does a good job of explaining why those are different things.

For AI researchers, engineers or people who follow the field closely: what specific capability do you think represents the largest remaining gap between current AI systems and what would genuinely qualify as AGI?