[Clip] The subtle manipulation of AI models

Artpic - One Change in the World - The Podcast
One Change in the World
[Clip] The subtle manipulation of AI models
Loading
/

when you go to the shop to buy a newspaper, as we did before, we could have ordered some newspapers that we actually enjoyed. It was in a way, we knew that this was left leaning, we knew that this was right leaning. But in the AI world, this is much more subtle, it’s much more hard to actually see what’s going on, because it’s not brand. ChatGPT is not branding itself as being leaning. Group is not branding itself as being right leaning. So I think that when it comes, or Gemini, this Google mobile as well is also regarded as very liberal, very left leaning, very woke. Some of these models, like Gemini, is not even possible to discuss politics, because it’s very afraid of touching on people’s toes regarding a political view. So if you ask Gemini to write an essay regarding health, regarding fashion, regarding environment, it will be political alignment into the wordings that Gemini is choosing. So we are in a situation where AI has what I call content, machines. It decides on topic, it decides on perspective

Leave a Reply

Your email address will not be published. Required fields are marked *