Artificial intelligence is not trained with scripted resources like ChatGPT and Bard. Instead, it is trained by recording the reactions a person shows in their brain while listening to a podcast. These reactions are recorded by the "functional magnetic resonance imaging" method.
Later, after the person's permission is obtained, these reactions are deciphered by artificial intelligence.

Moreover, it is not limited to just podcasts:

Artificial intelligence has also managed to record the reactions that occur in the brain according to what the person sees and hears while watching a video. This was also demonstrated in the video above.

Of course, getting all these thoughts is not an easy process:

Artificial intelligence also needs to be trained individually before measuring the thoughts in the brain. In this process, a person has to stay under the MRI machine for 15 hours.
Also, when the person thinks about something different during the podcast, these thoughts are also transferred to the output. For example, if you suddenly start thinking about animals while listening to a podcast, these thoughts mix with your thoughts about the podcast.