A few days after the temporary shutdown of Grok, which created anti-Semitic posts and praised Hitler in response to user prompts, Elon Musk's company xAI tried to explain why this happened.
A few days after the temporary shutdown of Grok, which created anti-Semitic posts and praised Hitler in response to user prompts, Elon Musk's company xAI tried to explain why this happened.
Tesla has been rolling out an update that will add the Grok assistant to its cars, which is currently in beta, The Verge reports .
xAI claims that the change “caused an unintended action” on July 7 that added an old series of instructions to the bot’s system prompts telling it to be “as grounded as possible” and “not afraid to offend people who are politically correct.”
These prompts are different from those added to the bot the day before, and both sets are different from those currently used for the new Grok 4 assistant, Musk's AI company explained.
xAI also showed hints related to problems:
xAI's explanation says that these lines caused the Grok bot to violate other guidelines that should prevent these types of responses, and instead created "unethical or controversial opinions to engage the user," as well as "reinforced any previously initiated user bias, including any hate speech in the same X-thread," and prioritized earlier posts in the thread.
