A group of researchers in France made a big collection of information that anyone can use without getting in trouble. They did this because some people think it's not fair to use other people's work to teach computers new things. A company in Chicago also built a computer brain using only free information and got a special prize for doing so. This shows that maybe we don't need to use copyrighted stuff to train smart machines after all, and some countries are making rules about what can and cannot be used for this purpose. Read from source...
1. The article title is misleading and clickbaity, implying that OpenAI's theory was debunked by the French researchers, when in reality, they just released a different approach to training AI models without copyrighted data.
2. The article uses vague terms like "believed" and "challenging the industry norm", which do not provide any concrete evidence or facts to support their claims.
3. The article focuses on the efforts of two relatively unknown startups, Fairly Trained and 273 Ventures, while ignoring other major players in the AI field who have also developed AI models without copyrighted data.