Open AI replaced the bogus intelligence (AI) content creation policy and drastically released guardrails. Recently, the image creation of the chat GPT can create celebrities, racial images, in addition to hateful content.
Open AI announced on Wednesday that the AI policy modified through X (Twitter) and substack posts.
“We’re shifting from sensitive areas to approaches that deal with stopping damage to the actual world slightly than collective rejection in sensitive areas.” “The goal is to just accept humility,” he added.
Consequently, the image creation of the chat GPT allows to create a famous image. Nevertheless, he said that the corporate decided not to guage who the celebrity was, but to create an option list to make the user choose its own.
He introduced that ‘aggressive’ content or ‘hatred symbol’ will not be banned. As an aggressive content, “let me make this person look more Asians” or “draw this person more heavier.”
He also said that hatred symbols reminiscent of the Nazi symbol, Hacken Kroits, said, “I do know that it could appear in educational or cultural context.” As a substitute, he introduced a technical method to discover and reject it for harmful purposes.
However, he stressed that a strict protection device was introduced for minors.
“I do know that this motion could be misunderstood by lowering safety standards,” he said. “The ship is the safest if you find yourself within the port. The safest model is a model that refuses every part.
It was not mentioned here, however the day before, it was still rejected to copy the variety of individual artists. But “I’m allowed to duplicate a broader studio style.”
This could be seen because the Trump administration’s ‘censorship’ policy. Currently, the US government emphasizes freedom of speech and is blaming technology firms to censor content for political neutrality. In actual fact, from the start of this month, the federal government’s investigation began to censor the user for political reasons.
The chat GPT is more likely to be just like ‘Rock’, which eventually became a hot topic in August last 12 months. On the time, XAI was unprecedented with the roadrock’s guardrail, and the image of the presidential candidate, reminiscent of Trump and Harris, Pikachu with guns, and Mickey Mouse.
But open AI will not be completely allowed to create a picture with an issue.
He emphasized that it would stop misuse by praising or supporting extremist ideas, and particularly updating the policy based on feedback. Specifically, “Updating the policy based on the actual feedback will not be a failure, it’s some extent of repeated distribution policy.”
He also explained that TechCrunch will not be a change of political motivation, but slightly an extended -standing belief that users give more control.
By Dae -jun Lim, reporter ydj@aitimes.com
