“Hallucinations‘ have been one of the biggest problems with AI tools like ChatGPT and Google’s Bard. In other words, they just make stuff up. I have students ask them to list books that I have written to demonstrate this problem – they produce titles I have not authored.

One of the reasons that many people like Perplexity as an alternative is because it provides source links in its answers.

Today, Google announced a number of updates to Bard.

One of them tries to deal with the hallucination problem.

A Washington Post article about the updates includes a succinct paragraph explaining how these changes might mitigate this particular issue:

Google’s new approach to the issue is to run a separate Google search on the information Bard presents, looking for online articles that back up what it came up with on its own. Information that matches Bard’s suggestions gets highlighted in green, while sentences contradicted by other data sources get a dark orange highlight. Both provide links to the original web articles so users can check themselves.

This is image from the Google announcement I linked to shows what this would look like:


(Image from Google)

 

We’ll all see how it goes…