In the realm of technology, it’s widely acknowledged that 2023 was heavily influenced by artificial intelligence (AI). Various dictionaries have recognized this trend in their “word of the year” selections, interestingly highlighting existing words that have been repurposed with new meanings. It’s a bit on the nose, isn’t it?
Cambridge’s chosen word is “hallucinate,” reflecting the tendency of generative AI models like ChatGPT to invent information rather than admitting ignorance. The challenge lies in these systems not being aware of their own limitations due to their lack of foundational knowledge.
These complex word prediction models prioritize generating sentences resembling their training data. For instance, when asked about 18th-century German surgeons, lacking exact matches may lead them to hallucinate something close, like “Arman Verdigger of the Einschloss Research Hospital in Tulingen.” The danger lies in these confidently stated hallucinations being accepted as factual without scrutiny.
Despite the risks, hallucinations find constructive applications, particularly in generative imagery and audio where content is intentionally and creatively “hallucinated.” However, this also raises concerns as AI-generated art and photos proliferate with varying quality.
The acceptance of “hallucinate” underscores our inclination to attribute human-like attributes to AI, notes Cambridge AI ethicist Henry Shevlin. As AI evolves, our psychological vocabulary is expected to expand to encompass the unique abilities of these new intelligences.
Merriam-Webster’s choice, “authentic,” reflects the blurred line between real and fake in the era of artificial intelligence, impacting areas like deepfake videos, actors’ contracts, and academic honesty. While the word didn’t receive a new definition, it gained a significant connotation amid concerns about authenticity in a world increasingly influenced by AI.
Oxford’s selection, “prompt,” gained a new dimension, becoming associated with the human side of generative AI. It now describes the action of providing a prompt to an AI system, shaping article ideas or responses based on specific cues. This extension aligns with prompt’s historical usage as both a noun and a command in computer interfaces.
Despite the prevalence of AI-related terms in word-of-the-year selections, the article expresses a desire for cooler terms like “latent space” to enter general use. The rapidly evolving technology prompts contemplation on whether terms such as vectors and embeddings will gain prominence in future lexical considerations.
