Google knows everything about you, but is this information being fed to the next-generation Bard AI to compete with the virally popular ChatGPT?

Barely out and about for the first group of signees to try out, Bard AI has already stirred controversy about where conversational AI fetches its data for training.

Based on the findings of Kate Crawford, who is a Microsoft researcher, Google’s Bard AI allegedly revealed that it is trained based on the data from Gmail, among other sources.

When prompted by the user, Bard mentioned that its dataset came from Google’s internal data, which “includes data from Google Search, Gmail, and other products.”

So did Bard AI just reveal its training model is based on personal user data?

Well, not really.

Considering how Bard AI is also a generative AI that is based on LLM (Large Language Model), it provides data based on a large data set.

Simply put, this makes Bard “make things up” and provide information that may not be factually accurate.

Confirming this and addressing the slip of the tongue that Bard AI had, Google did respond to the tweet and stated that Bard AI is “an early experiment based on Large Language Models and will make mistakes. It is not trained on Gmail data.”

Despite the popularity of generative AI and LLMs in the tech community, these tools are not foolproof and can sometimes fail to provide accurate information.

Users with access may attempt to extract hidden information, but as demonstrated in the case of Crawford’s query, the results are not always reliable.

Although these systems are impressive, they are still plagued with early problems.

OpenAI’s ChatGPT-4 has been widely accepted as the reigning champion when it comes to conversational AI, and Microsoft’s fresh investment is all the more reason for the search engine giant Google to worry about it.

In a bid to overthrow ChatGPT, Google has already rolled out the first version of Bard AI, and you can sign up to join the waitlist right now.