[Garth Tater]

What a robot does is dependent on the premises of the programmer.

What a neural network (an AI bot) does is largely determined by the training data it works with. The programming determines how it learns but the data it works with determines what it learns. When you let internet users provide the data (by chatting with your bot) unexpected things can happen. Microsoft had a wonderful education in this field when their chatbot was hijacked by the cool kids on twitter and turned into a racist drug lovin' bot.

Microsoft’s racist chatbot returns with drug-smoking Twitter meltdown

[TigerLikesRooster]

A chatbot can be both gullible and honest at the same time. It basically plays back what it reads and hears. The data can be manipulated or honest. So it comes down to the filtering mechanism which is supposed to prevent manipulations. Then the question is, “Is this filtering fair and unbiased or not?”

[Bethaneidh]

Tx for the data. I have had interesting fun with chatbots. But the programmer decided that.