The Cambridge Dictionary's word of the year is the AI-inspired “hallucinate".
The dictionary has also added a definition that relates specifically to its use on the subject of AI, typically when chatbots fabricate information and present it as fact.
“When an artificial intelligence (a computer system that has some of the qualities that the human brain has, such as the ability to produce language in a way that seems human) hallucinates, it produces false information," the Cambridge Dictionary said.
“LLMs are notorious for hallucinating — generating completely false answers, often supported by fictitious citations," is one of its examples of the word.
LLMs are large language models, which are at the heart of popular instances of AI such as ChatGPT.
One comparison you might think of is messaging autocomplete in a phone keyboard, and how its job is to complete the sentence in a manner that seems likely. It can guess what you might want to say based on the context, but it doesn’t actually know.
This gap between reality and the guessed result causes these hallucinations in popular generative forms of AI.
AI ethicist Dr Henry Shelvin from the University of Cambridge says the use of this term illustrates our tendency to humanise artificial intelligence.
“The widespread use of the term ‘hallucinate’ to refer to mistakes by systems like ChatGPT provides a fascinating snapshot of how we’re thinking about and anthropomorphising AI,” says Shelvin.
“Whereas these are normally thought of as human products, ‘hallucinate’ is an evocative verb implying an agent experiencing a disconnect from reality. This linguistic choice reflects a subtle yet profound shift in perception: the AI, not the user, is the one 'hallucinating.' While this doesn't suggest a widespread belief in AI sentience, it underscores our readiness to ascribe human-like attributes to AI.”
The power of that tendency to humanise is part of the danger of using current forms of AI, an assumption that the results will be factual when they are presented like the words of a confident, authoritative person. And it’s also why the current generative LLM forms of artificial intelligence have so rapidly become all people think of when AI is broached.
Software that tweaks your phone images after they are captured might be thought of as AI, just like the software that let the computer play against you in 1991’s Chessmaster 3000. But they are not as compelling as a chatbot that can answer like a human.
The Oxford University Press has not yet announced its word of the year, as it typically does so in December.
Last year, its pick was “goblin mode” while the Cambridge Dictionary’s 2022 word of the year was “homer”. This is a reference to the baseball shortening of home run, not the The Simpsons character or the late Greek poet.
