Developers of artificial intelligences face a challenge: you need enormous amounts of data to train your language models. Where to get if you don’t steal? Apparently it was decided at the Facebook group meta.
Whether it is about taking photos, videos, music or text: AI can only generate what she has taught earlier – with huge amounts of photos, videos, music or text. The AI learns how the respective means of expression work. In the case of OpenAIS chatgpt, for example, such as grammar, style, meaning and context, ultimately language, and then based on user input and the learned patterns, a text.
When employees of the Facebook Mother Group Meta wanted to develop their voice model Llama 3, they faced an ethical challenge: to compete with products such as Chatgpt, the program should be trained with huge amounts of high -quality texts that can be used legally. So do you have to predators the data better?
Source: Krone
I am Wallace Jones, an experienced journalist. I specialize in writing for the world section of Today Times Live. With over a decade of experience, I have developed an eye for detail when it comes to reporting on local and global stories. My passion lies in uncovering the truth through my investigative skills and creating thought-provoking content that resonates with readers worldwide.
