Large language models like ChatGPT are often misused as search engines, but they excel at language manipulation and can be considered 'calculators for words'.
These models can be used for summarization, question answering, fact extraction, rewrites, suggesting titles, and creative tasks, but users should remain skeptical and fact-check their answers.
Key terms:
Hallucination: A phenomenon where language models create new facts that fit the sentence structure but have no basis in the underlying data.
Calculator for words: A metaphor for how language models excel at language manipulation tasks rather than being reliable search engines.
Browsing mode: An alpha feature of ChatGPT that allows it to work with search engines to answer questions using search results.