Having watched the video, I am deeply disappointed. There are examples of the use of AI, including a recreation of two existing persons. I will refer to avatars in this comment. While it may sound exciting to have a recreation of your child or yourself, there is a major disadvantage: the impersonation cannot learn normally, even if there are built-in limits in learning via artificial brain nodes. Key components of personality will be different because the avatar is not a human being.
In fact, an avatar needs to be limited. As the AI learns, either from impulses fed to it via behaviorist instructions to build its personality or by accessing information from the Internet, it has a different set of data to its disposition. It will be able to process more, and faster, but not really have a free, organic will.
What is more, the info which the AI-aided avatar feeds off of will have to be hierarchical and heavily segmented, to avoid influence by misinformation. But that ultimately means that the source of data will have to be artificial. The end of the internet as we know it – as individuals get served by AI lookalikes? Will there be consensus? Or will this be the root of conflicts?
Thorsten Koch, MA, PgDip
18 October 2020