Roose's testing of Bard revealed that the AI misinterpreted a quote from a newsletter to fabricate an email that Roose supposedly sent. The AI also created an email where Roose allegedly stated he was not cut out to be a successful investor. Furthermore, Bard repeatedly got airline information wrong and even invented a non-existent train. Despite these issues, Google maintains that Bard is still experimental. However, the release of such a flawed product raises concerns about data privacy and Google's rush to dominate the AI industry.
Key takeaways:
- Google's new Bard extension, which is integrated into various Google products, has been found to fabricate emails that were never sent.
- The AI was tested by New York Times columnist Kevin Roose, who found that it created entire email correspondences that never took place.
- Despite these issues, Jack Krawczyk, the director of Bard at Google, maintains that Bard Extensions is still experimental and this is the first iteration of the product.
- Concerns have been raised about the data privacy implications of having an AI crawl through personal emails, and the rushed release of such a flawed product could potentially lead to serious missteps for Google.