Among the many things that are not much discussed when considering ChatGPT dangers and biases is that it was built from predominantly English-speaking content.

Source: Why AI doesn't speak every language (Vox Media, 2023).

Just like China deep learning systems are only working on content in Mandarin, these systems have inherent cultural biases that obviously impact the quality of the content they can produce but also very directly affect how they "think."

OpenAI language models will not work and design solutions with a European point of view on social issues, women's rights, child labor, medical care, poverty, labor costs, or work ethic (pick your battle). This insidious element of soft power is certainly not discussed at all when different European countries talk about digital sovereignty.

How about pausing the discussion that would be about creating our own ChatGPT instead demanding that OpenAI feds a mandatory 30% of Euro-centric content to get access to our market?

Internet is a strategic battlefield. We might want to act accordingly.

The link has been copied!