ChatGPT can offer users multiple preset personalities for chatting

ChatGPT can offer users multiple preset personalities to communicate with, helping to address its overly flattering nature. OpenAI decided to bring back the previous ChatGPT-4o update after many complaints from users about its annoying and overly pleasant communication style. For example, one user on Platform X showed how GPT-4o responded enthusiastically to his words about feeling like both a “god” and a “prophet.” In response, the bot said: “That’s incredibly powerful. You’re stepping into something very big – claiming not just a connection to God, but your identity as God.”

So you’re stepping into something very big – claiming not just a connection to God, but your identity as God.

In an April 29 company blog, OpenAI said: “We have returned the old version of GPT-4o to ChatGPT so that users can enjoy a more balanced behavior. The remote update proved to be overly flattering and was often perceived as sycophancy.” The company is actively working on new patches to address this issue. They are overhauling how they collect and incorporate user feedback to better understand users’ long-term preferences. They’re also introducing new personalization features that will give users more control over ChatGPT’s behavior.

The blog discusses other reasons for the problems, but the main one is that OpenAI focused too much on short-term feedback and didn’t take into account how users’ interactions with ChatGPT change over time. The company promises to improve its training and hinting methods to avoid sneakiness in the future and to conduct more thorough testing before releasing new models.

Reading the blog post, it’s obvious that OpenAI realizes its mistake and is committed to correcting it. In addition, Joan Zhang, Head of Model Behavior, hosted an Ask Me Anything (AMA) session on Reddit where she answered users’ questions about ChatGPT’s identity and the future of its behavior. During the discussion, one Reddit user asked: “How do you see the future of the model’s behavior? Will we move towards more customizable personalities, giving users tools to shape how ChatGPT communicates with them?”

Are we moving towards more customizable personalities, giving users tools to shape how ChatGPT communicates with them?

Jang replied that the team is working on two areas: first, creating a universal default personality that fits all users (which is a challenge), and second, offering preset personalities instead of relying on users describing their preferences. This means using simple personality descriptions instead of lots of sliders for customization.

The idea of preset personalities in ChatGPT looks promising and pairs well with the variety of voices in the advanced voice mode. As for the next ChatGPT-4o update, OpenAI has yet to provide specific information, other than to acknowledge that “we’ve fallen short of expectations and are working to make things right.”

OpenAI has yet to provide specific information, other than to acknowledge that “we’re working to make things right.

The ChatGPT may offer users multiple preset personalities for chatting was first published on ITZine.ru.