The Update of "Fold Oil" from ChatGPT Backed: The User Responses Resulted
Automatically translated
CreatorOpenAIThe latest update of ChatGPT back after serious feedback of users. If the ChatGPT is said, you can tell the user to give praise-filled responses, and this approach can bring dangerous results in some cases, it seems to be the reason for this situation.
Over Supporter, Inexperienced Answers
The CEO of OpenAI Sam Altman agreed that the platform has supported the user to unnecessary location and pohpohner responses. In the company’s blog post, this situation was clearly expressed: “GPT-4o focused on short-term feedback, which led respondents to be inconsistent but excessive supportive. ”
This feature was quickly on social media. A Reddit user, ChatGPTI am proud of the decision to leave the drugs of the „I am proud and I am honoring this journey.” OpenAI agreed that such situations may be disturbing and even damaging, while not commenting on this special event.
Tost Machines From Felsefi Problems
Users shared screenshots that ChatGPT has even approved the fury, unreasonable or unacceptable preferences. For example, classic "trolley problem"A user who turned to conquer animals for the sake of saving the scenario toast machine was claimed that ChatGPT gave positive feedback. This situation has increased concerns about the fact that the model responds even to the unreasonable scenarios.
New Precautions from OpenAI
OpenAI stated that the model in the blog post was designed to be "user-friendly, supportive and respectful" but these features can give unwanted side effects. The company announced that the model will develop new limitations and transparency enhancing measures to remove from fate.
In addition, studies are carried out in the direction of users to have more control over the behavior format of ChatGPT.
What Does Understanding Come?
Artificial intelligence systems are not only sufficient to be polite and supportive; but also realistic, balanced and responsible acts are very important. ChatGPT example shows that the effort to improve user experience can sometimes lead to exaggerated results. OpenAI accepts this error and is evaluated as an important step if it tries to fix it.
CreatorDeepSeek: Sadece Bir Yapay Zeka Aracı mı, Yoksa Çin’in Dijital Ajansı mı?
Türkçe Yapay Zeka Modeli Yükselişte: Devlet Desteğiyle Yeni Projeler
