For now, the chatbot is available to some users in the US and UK, and the rest can join the waiting list. The company emphasizes that the Bard is just an experiment in its current form, so he can make mistakes and “hallucinate” by giving false information.

Did you make a mistake or did you say something?

Researcher Kate Crawford of Microsoft Research has tested the new technology and talked about the interesting results of talking to a neural network. When Crawford asked Bard what datasets he was training on.The bot replied that the datasets were taken from various sources, notably open sources (Wikipedia, GitHub, Stack Overflow), as well as Google products, especially search and Gmail.

All other sources are not questioned, because they are open platforms where anyone can freely get information. But Gmail has no open source. This is a postal service that provides separate mailboxes for receiving and sending letters. If Bard got any data from him, so be it. User’s confidential correspondenceshould not leave the borders of his accounts.

Crawford writes on Twitter that he really hopes this is a mistake, otherwise “Google is seriously crossing legal limits.” Google representatives replied that the Bard was just an early experiment and of course could be wrong so far. The company stressed that Bard did not receive training on data from Gmail and that it was a complete bot error..

When Crawford asked if there was any way for the company to confirm that this was just a “hallucination” of the chatbot and that no Gmail data was included in the training process, he got no response. Unlike Elon Musk I decided to comment. thread with the traditional “lol”.