Why LLMs Make These Mistakes

This week I discuss the latest BBC study on AI answer quality from public data sources. As I discuss, the BBC and EBU found that 45% of news queries produce erroneous answers, so the reality has set in: public domain AI engines are neither “superintelligent” nor are they perfect.

Yet they are very self-confident. So we, as users, need to be careful.

As you’ll hear, there are three things to consider here, and you can read more about this in my latest article on the topic. For those of us in corporate roles, the message is clear: data quality must be our #1 priority, and this is a whole new domain for HR and other service functions.

Like this podcast? Rate us on Spotify or Apple or YouTube.

Additional Information

BBC Research Findings

Interpretation of the Findings and the “polluted corpus” problem

Claude’s admission of the “polluted corpus” problem

AI Thinking Skills You Need To Stay Safe (podcast)

Galileo: The World’s Trusted Agent for Everything HR