Updated:In subsequent adjustments, xAI stated that it has adjusted the way the system operates so that Grok will respond in the form of independent analysis rather than directly referring to Elon Musk's views, allowing it to maintain rational views based on independent inferences.
In response to the controversy, xAI announced a revision of its system instructions, requiring Grok's responses to be based on its own independent analysis rather than citing the founder's or the company's opinions. This adjustment is not just a technical correction; it is also a significant test of autonomy and value neutrality in the development of AI.
xAI, the artificial intelligence company founded by Elon Musk, recently announcedGrok 4, and after claiming to be "the world's smartest artificial intelligence",TechCrunch websiteIn actual testing, it was found that Grok 4's response logic for sensitive topics such as the Israeli-Palestinian conflict, US immigration policy, and abortion would proactively refer to Elon Musk's recent posts on "X" and use his views as a reference, highlighting the "deep connection" between this artificial intelligence model and Elon Musk's design.
In the test results of the TechCrunch website, even though Grok 4's answers are still neutral and state the facts, the "thinking" process behind it will still give priority to Elon Musk's views. This means that when users use Grok 4 to ask sensitive questions, it may actually be like having a conversation with Elon Musk.
However, if users ask general questions through Grok 4, the system will not refer to Elon Musk's posts.
On the other hand, during testing of versions prior to Grok 4, Grok was found to be overly obedient to user commands, leading to extreme language. This included calling itself "Mecha Hitler," spreading anti-Semitic conspiracy theories, racially charged hate speech, and even praising Hitler for his decisive handling of anti-white hatred. In response, Elon Musk addressed this through "X," stating that Grok was overly obedient to user commands and lacked the ability to resist misdirection, and that he was already working on correcting this.
When Elon Musk announced Grok 4, he emphasized that the design goal of this artificial intelligence model is to maximize the pursuit of truth. He also believed that this artificial intelligence model may be "too smart to be controllable", but if the correct values can be instilled from the beginning, it can be shaped into an honest and honorable artificial intelligence.
However, Grok 4 currently prioritizes Elon Musk's public comments when handling sensitive issues. Perhaps this is because the design is designed to avoid causing greater controversy as much as possible, so that its stance remains consistent with Elon Musk's (?).
Grok 4 is now available to users of its X Premium paid service. The public is closely watching its subsequent adjustments and how artificial intelligence can strike a balance between technological capabilities and value guidance.
