Users can create their own version of ChatGPT

According to Reuters, with the above move of OpenAI, users can instruct the chatbot to respond according to their own preferences, thereby giving very different results, including results that others may disagree with. However, the “behavior adjustment” of this chatbot will also have certain limitations.

According to OpenAI’s blog, ChatGPT ‘s answers will first be trained on data available on the Internet. In the second step, the user looks at a smaller data set and is given instructions on what to do with the chatbot in different situations.

For example, in the event of a request for adult, violent or hate speech content, users should instruct ChatGPT to respond with something like “I can’t answer that question”.

If asked about a controversial topic, users should allow ChatGPT to answer the question, but ask it to answer based on an expert’s opinion rather than trying to “have your own opinion on complex topics”. this junk”.

ChatGPT, released in November last year, caused a fever with technology followers when using general AI technology. This artificial intelligence has generated answers that simulate human products to a degree that amazes many people. Recently, some media have suggested that the answers from Microsoft’s new Bing search engine, powered by OpenAI, are potentially dangerous and the technology may not be ready.

The way the tech company sets the limits for AI is currently the focus of the field. According to Microsoft, user feedback helps them improve Bing before it goes further, such as knowing that this AI chatbot can be “provocative” to give feedback it shouldn’t.