Cambridge Dictionary has announced hallucinate as the Word of the Year for 2023.
The news follows a year-long surge in interest in generative artificial intelligence (AI) tools like ChatGPT, Bard and Grok, with public attention shifting towards the limitations of AI and whether they can be overcome.
AI tools, especially those using large language models (LLMs), have proven capable of generating plausible prose, but they often do so using false, misleading or made-up ‘facts’. They ‘hallucinate’ in a confident and sometimes believable manner.
The Cambridge Dictionary – the world’s most popular online dictionary for learners of English – has updated its definition of hallucinate to account for the new meaning.
The traditional definition of hallucinate is “to seem to see, hear, feel, or smell something that does not exist, usually because of a health condition or because you have taken a drug”. The new, additional definition is:
“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.”
AI hallucinations, also known as confabulations, sometimes appear nonsensical. But they can also seem entirely plausible – even while being factually inaccurate or ultimately illogical.
AI hallucinations have already had real-world impacts. A US law firm used ChatGPT for legal research, which led to fictitious cases being cited in court. In Google’s own promotional video for Bard, the AI tool made a factual error about the James Webb Space Telescope.
Wendalyn Nichols, Cambridge Dictionary’s Publishing Manager, said: “The fact that AIs can ‘hallucinate’ reminds us that humans still need to bring their critical thinking skills to the use of these tools. AIs are fantastic at churning through huge amounts of data to extract specific information and consolidate it. But the more original you ask them to be, the likelier they are to go astray.
“At their best, large language models can only be as reliable as their training data. Human expertise is arguably more important – and sought after – than ever, to create the authoritative and up-to-date information that LLMs can be trained on.”
The new definition illustrates a growing tendency to anthropomorphise AI technology, using human-like metaphors as we speak, write and think about machines.
Dr Henry Shevlin, an AI ethicist at the University of Cambridge, said: “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. Inaccurate or misleading information has long been with us, whether in the form of rumours, propaganda, or ‘fake news’.
“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.
“As this decade progresses, I expect our psychological vocabulary will be further extended to encompass the strange abilities of the new intelligences we’re creating.”
Addressing hallucinations – if they can ever be fully fixed – may define the future success and uptake of generative AI.
