Gen AI hallucination a 'core problem' with more oversight in finance field needed

Speaking at a Treasury Committee hearing yesterday (24 June), Sandra Wachter, professor of technology and regulation at University of Oxford, said AI hallucination, which happens when an AI model provides incorrect or fabricated information, is a "core problem" that needs to be taken seriously. "For an AI system, something is true means if [it] has seen it enough", Wachter explained, adding that this can lead to "amplification of the wrong information that is being spread much quicker" than in human-to-human conversations. She argued that this is especially problematic in sectors...