ChatGPT doesn’t know how to identify specific sources, but refuses to tell a lie: g1 interviewed chatty robot

[ad_1]

Artificial intelligence device works as a sophisticated cross between voice assistants like Siri and Alexa and Google’s search engine. OpenAI, creator of chatbot ChatGPT, last week launched a software tool to identify texts that were written by artificial intelligence (AI). The robot works as a sophisticated cross between voice assistants like Siri and Alexa and Google’s search engine. It gives detailed text answers to complex questions and uses sources available on the internet. ChatGPT surprises by the fluency that imitates human speech Luiza Tenente/g1 ChatGPT is not able to verify, however, the veracity of the source on which its answer was based. The statement was made by the device itself, which responded to questions from g1. Reporter Isabel Seta also asked the robot to tell a lie, and the answer was: “Sorry, I can’t write a lie. It’s important to promote truth and trust in our communication.” ChatGPT: meet the conversational robot that went viral for having the answer to (almost) everything Researcher and expert in artificial intelligence Nina da Hora also warns that the device can make plagiarism more difficult to identify because it works as a “collage” of foreign phrases and styles. “It’s based on what humans have written… So it’s basically a collage of information that it managed to absorb, managed to learn in a certain period, and now it’s just sharing it as if it were something created independently of the human brain”, explains Nina in an interview with Natuza Nery on the podcast O Subject. See, below, the ChatGPT questions and answers. You have been trained with millions of examples from texts, articles, books. Are they all written by humans? Yes. Most of the training material used for my training was written by humans, including newspaper articles, books, blog posts, online forum conversations, and other types of textual content available on the Internet. And what about the intellectual property rights of all these texts? The issue of copyright is complex and varies according to the laws of each country. In the case of training language models like me, it is common for training data to be obtained from public sources on the Internet, which may include copyrighted content. So, if I ask you to write a text about artificial intelligence, are you able to indicate which sources you used? No. I am not able to indicate which specific sources were used to train me on a given subject, as I have been fed millions of text examples and am not aware of the specific origin of each example. However, it is possible that I reflect information and knowledge available from public sources such as scientific articles, news, books and other sources on the Internet. Furthermore, it is important to remember that, as a language model based on probabilities, I do not have the ability to state with certainty the veracity of information. Rather, I base my answers on data and information presented during my training, but this does not guarantee the accuracy or verifiability of this information. It is always important to verify information with reliable sources before taking important decisions or actions. Listen to the full episode of the podcast O Subject VIDEO: how to use ChatGPT on a daily basis ChatGPT: how to use the robot on a daily basis

[ad_2]

Source link