Zoekresultaten

… and more specifically, I work a lot with natural language processing. So you can imagine that since last year that chat GPT came out. I've been quite busy with working out how these models work and what else do we need from Constance: That's also why we invited you more or less, because we've also been tinkering with ChatGPT sometimes for the research for some of our episodes or for other questions that we have. And we found out that [00:01:00] ChatGPT was quite woke in its answers! Darian: Or … giving us answers that seemed often conformed to a kind of liberal status quo bias and even to the extent that sometimes we would ask questions and we would get questions in return about “why are you asking that kind of question? Aren't you aware that this kind of thing that you're asking about is dangerous or is extremist?”, and that happened. Especially to me, or in one instance to me, I was doing a little bit of, I was trying to do some background research on a topic, eco fascism, so like far … we need to think about the limitations Constance: Okay, before we go into the content they produce, via the predictive models, I'm wondering, so if , it listens to us, we also have a transcript of this episode, so there's, the actual text will be online as well. Using the data from this kind of conversation, can it then mimic these kind of. More holistic ideas about life. And how would that look like? Jerry: exactly. Every piece of text, let's say goes into this language models, and then it …