WM Briggs about AI models programmed and given database for a certain output

Excerpt from: https://www.youtube.com/watch?v=1FSMMrwLznU&list=UULFnK3qgNPHupWlzFo5TlhRTw

WM Briggs has a rather 'shit input - shit output' estimation of (at least the current state of publicly known) Artificial Intelligence.

I get the idea/statement that all AI models are biased, as they are programmed and fed with a certain pool of data. Whereas I'm not convinced that lawnmowers or calculators are biased. Only if you define their task-specific functioning as bias. I also think that an AI chatbot, or any answering or conversation machine is at least much less biased if it isn't programmed for ideological answers and the circumvention of true, but politically unwanted answers. And when the pool of data for its composition of answers is not politically selective and biased.