What artificial intelligence can do - and what it cannot
AI can recognize patterns, generate text and support research, but confident answers are not always correct.
Artificial intelligence is a collection of methods that allow software to perform tasks associated with human intelligence. Modern systems learn statistical patterns from examples rather than following a complete list of hand-written rules. Some models can produce text, pictures or computer code; other systems help scientists study complex data.
These systems can save time, but they can also make mistakes. The quality of their answers depends on their training data, the task and the safeguards used. A convincing explanation is not a substitute for an independently verifiable source.
AI also raises public questions about privacy, copyright, education, employment, security and energy. People and governments can reasonably disagree about which rules are best. Good reporting separates what a technology demonstrably does from predictions about what it might do.
When reading an AI headline, ask who made the claim, what evidence was published, whether the result has been independently tested and what limitations remain. These questions are useful regardless of anyone’s political position.
All translated versions are original adaptations of the same explainers; each language has its own shareable address.
Our Sponsors
Promotional placements from businesses associated with the publisher. Sponsors have no control over independent news coverage.