Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
[ad_1]
Xai has accused the “unauthorized modification” for an error in the AI Strong Grock Chatbot, which caused Grok repeatedly referring to “White genocide in South Africa” when playing in certain contexts in X X
On Wednesday, Grok began to respond to dozens of posts in the X on the White Genocide in South Africa, even in touch with undocumented subjects. A person’s label “@grock” strange answers caused by the X Account for the grock that responds to the users who want to
On Thursday, a change in a start from the official X, a change on Wednesday morning GROK BOT system – high-level instructions that direct the bot’s behavior – “Political Topic” high-level instructions that provide a “specific answer”. Xai says that pinch “broken” [its] Domestic policy and basic values ”and the company’s” thorough investigation. ”
For the second time, Xai has clearly accepted that the GROC is an unauthorized change in the GROK’s code, which causes the EU controversial.
Groch in February briefly censorship Donald Trump and Elon Musk, Khai and Xai’s founder of the founder of the United States, the founder of one of the Khai Engineering device, said he instructed the founder of one of the Xai Engineering naughty worker MUSK or Trump’n ignore the sources noting the spread of incorrect information and returned the change in the moment when this Xai began working.
Xai, on Thursday, said that similar events will make a few changes to prevent the future.
Starting from today, Xai Github will present a grock system of GitHub, as well as a change. The company says that Xai will change the system offering without a review of employees and ensure that it is “24/7 monitoring group to respond to incidents not charged by automated systems” and “24/7 monitoring group”.
Despite frequent warnings of the dangers of the dangers AI followed unchangedXai has a weak AI security path. Last Report When GroK was asked, he found that the women will undress their pictures. Chatbot may also be much more dwarf More than AI, like Google’s twins and Chatgpt, are very restrained to talk.
A study by Saferai, which aims to improve the accountability of AI laboratories, is weak in terms of security in terms of security between Xai and peers “Very weak” risk management experiences. Before this month, Xai Missed itself from the deadline to publish the final AI security framework.
[ad_2]
Source link
