Generative AI models predict text based on patterns learned from vast amounts of data. To “appear” more natural they don’t always “choose” the most likely word. This makes their output unpredictable and leads to errors. Since AI doesn’t “know” anything, it doesn’t fact-check its output.

Tom Cranstoun will explain why AI models hallucinate and the research going on to reduce this problem.

Tom Cranstoun’s career in IT has involved delivering digital transformations for global brands such as Nissan-Renault and Twitter. As a technical architect, his skillset includes translating complex technology into accessible IT solutions. Tom now brings his expertise to demystifying artificial intelligence.

Find out more about Tom on LinkedIn.