How close is AI to human-level intelligence?

How close are we to developing an artificial general intelligence (AGI) — a machine capable of the whole range of cognitive tasks that human brains can handle? Some think that the large language models (LLMs) currently out there already have some of the ingredients in place. One point in these models’ favour is their underlying transformer architecture, which can find statistical patterns in a range of information beyond text, such as audio. Yet there are also signs that transformer-based LLMs have limits. For a start, the data used to train the models are running out.