Artificial intelligence, or AI, is machines and computer programs that can imitate smart human behavior, such as making predictions and decisions. Yet the people who study it have trouble pinning it down. John McCarthy wrote an early FAQ on the subject. He wrote that there is no single yes-or-no answer to whether a machine is intelligent. Intelligence involves many mechanisms, and AI has learned to make computers carry out some but not others. He also wrote that most work in AI studies the problems the world presents to intelligence, rather than studying people. AI researchers may use methods people do not use. In his FAQ, a solid definition of intelligence that does not depend on human intelligence did not yet exist.

The Nobel Prize committee noted that when people talk about artificial intelligence, they often mean machine learning that uses artificial neural networks. The 2024 Nobel Prize in Physics was awarded for foundational discoveries and inventions that enable machine learning with artificial neural networks. Machine learning works best when there is a lot of data, such as thousands or millions of examples.

Large language models learn to predict the next word through pre-training. The model produces probabilities for possible next words from patterns in its training data, and one of the highest-probability words is picked to continue the text. Predicting the next word is not the whole story of how a chatbot is built, though, because pre-training is only one of several key steps in making today's large language models. Backpropagation is the most common way to train a neural network.

Descriptions of hallucination say it follows from how generative models work. A language model rebuilds text from patterns in its training data, rather than looking facts up in a fixed store. NIST, the U.S. National Institute of Standards and Technology, says users may believe false AI content, often because of the confident way the response is written, and then act on it or pass it on. A 2024 Stanford study of three popular language models found hallucination rates of 69 percent to 88 percent when they answered specific legal questions. The models often did not recognize their own errors. They did worse on harder tasks that needed a careful reading of legal ideas or texts.

NIST warns that performance gaps between groups or languages can come from training data that does not represent everyone. This can lead to wrong guesses about how well a system works for some people. In a 2018 MIT study of three commercial face-analysis programs that guessed gender, the error rate for light-skinned men was never worse than 0.8 percent. For darker-skinned women it was more than 20 percent in one program and more than 34 percent in the other two. The error rates were consistently higher for women than for men, and for darker-skinned people than for lighter-skinned people.

NIST's 2022 report said that to understand AI bias, people must look beyond the computing and data to the wider society in which an AI system is built and used. NIST also says it can be hard to measure the harms of AI, because some negative impacts are difficult to assess. The OECD AI Principles, adopted in 2019 and updated in May 2024, promote AI that is innovative and trustworthy and that respects human rights and democratic values. NIST's AI Risk Management Framework is meant to be voluntary. Organizations of any size or sector can choose to use it to manage the risks of AI.

As of November 2026, the EU's AI Act, adopted in 2024, sorts AI systems into four levels of risk. AI systems seen as a clear threat to the safety, livelihoods and rights of people are banned. The Act sets no rules for AI judged to be minimal or no risk, and it requires that people be told when needed to keep trust, such as when they are dealing with AI.

A 2024 U.S. Department of Energy report said data center load growth had tripled over the past decade and was projected to double or triple by 2028. As of early 2025, MIT reported that data centers worldwide used about 460 terawatt-hours of electricity in 2022, and that use was expected to approach 1,050 terawatt-hours by 2026. In early 2025, researchers estimated that a query to a chatbot such as ChatGPT uses about five times more electricity than a simple web search. Water is also needed to cool the hardware that trains and runs generative AI, which can strain city water supplies and local ecosystems.

In 1950 the mathematician Alan Turing published a paper that asked, "Can machines think?" In the game he described, a person asking questions tries to tell which of two hidden players is the machine and which is the person. The answers are typed. In 2014 a program was said to have passed the Turing test after it fooled 33 percent of judges. But the trial was very small, and the result was not reliable.

A March 2025 report from the Association for the Advancement of Artificial Intelligence included a survey of 475 people in its community. Of them, 76 percent said that scaling up current AI approaches is unlikely or very unlikely to lead to general intelligence. Most people surveyed disagreed that factuality, meaning not making up answers, or trustworthiness would soon be solved. Most agreed that trustworthiness is ill-defined as it is currently formulated.

People should check what an AI tells them before they trust it. A good way to check is lateral reading, which means opening a new tab and looking the facts up on other reliable pages. If a chatbot names a book or article, a person should search to confirm that it really exists. An AI should not be asked to check its own work, and a person should look for outside sources.

Words to know

• backpropagation: training method
• hallucination: confident mistake
• voluntary: by choice
• terawatt: huge power unit
• ecosystems: living communities