Openai Peles Back Chatgpt on creating photos

This week, Openai I launched a new photo generator In ChatGPT, which quickly went virus for its ability to create Ghibli-style studio photos. Besides pastel illustrations, the original GPT-4O photo generator promotes ChatGPT capabilities, improving photo editing, providing text, and spatial representation.

However, one of the most prominent changes made by Openai this week includes their content baptism policies, which now allow Chatgpt to create images depicting general numbers, hateful symbols and ethnic features.

Openai previously rejected these types of claims for being controversial or harmful. But now, her approach has evolved, according to Blog post Posted on Thursday by Openai’s Model Behavior Lead, Joanne Jang.

“We are turning from comprehensive rejection in sensitive areas into a more accurate approach that focuses on preventing damage in the real world,” Gang said. “The goal is to embrace humility: identifying the extent of our lack of knowledge, and putting ourselves to adapt to we learn.”

These amendments seem to be part of the largest Openai To Chatgpt effectively “Uncensor”. Openai announced in February that it has begun to change how to train AI models, with the ultimate goal of allowing Chatgpt to deal with more requests, submit various views, and reduce the topics that Chatbot refuses to work with.

Under the updated policy, ChatGPT can now create and edit photos of Donald Trump, Elon Musk and other public figures that Openai has not allowed before. Jang says that Openai does not want to be tightly, and chooses who should not be allowed to be born by Chatgpt. Instead, the company gives users the option to cancel the subscription if they do not want to photograph their Chatgpt.

in White paper Openai, as Openai, said, it will allow Chatgpt users to “generate hateful symbols”, such as the swast cross, in educational or neutral contexts, as long as they “clearly do not praise or support extremist business schedules.”

Moreover, Openai changes how it defines the “offensive” content. Gang that she wanted to reject requests about physical characteristics, such as “making this person look more Asian” or “make this person heavier.” In the techcrunch test, we found that the new Chatgpt image generator fulfills these types of requests.

In addition, Chatgpt can now imitate creative studios patterns – such as Pixar or Studio GHibli – but still restricts the tradition of the methods of living artists. As I noticed Techcrunch before, this can Prepare a current discussion on the fair use of copyright -protected work in artificial intelligence training groups.

It should be noted that Openai does not completely open the gates to abuse. The original GPT-4O photo generator still rejects a lot of sensitive queries, and in fact, it contains more guarantees about creating pictures of children from Dall-E3, the former AI image generator from ChatGPT, according to White GPT-4O paper.

But Openai relaxed handrails in other regions after years of Conservative complaints about the alleged “censorship” of artificial intelligence from the Silicon Valley companies. Google has already faced a violent reaction to the AI ​​Gemini photos, which he created Multiple informatics pictures Like “founding fathers” and “German soldiers in World War II”, which were clearly inaccurate.

Now, the war of culture around moderate artificial intelligence content may be to its head. Earlier this month, Republican Congress member Jim Jordan sent questions to Openai, Google and other Giants about Possible complicity with the Biden Administration for content control created by artificial intelligence.

in The previous statement For techcrunch, Openai rejected the idea that its content baptism changes were of political motivations. Instead, the company says the transformation reflects “a long belief in giving users more control”, and that Openai technology is now good enough to move on sensitive topics.

Regardless of its motives, it is certain that Openaii changes moderation policies in the content, given the possibility of organizational scrutiny within the framework of the Trump administration. Silicon Valley giants such as Meta and X also adopted similar policies, allowing More controversial topics on their platforms.

Although the new Openai image generator has only established some GHibli Virus Studio Mets so far, it is unclear what are the broader effects of these policies. The recent changes in Chatgpt may fade well with the Trump administration, but allow AI Chatbot to answer sensitive questions can land in hot water soon.