Elon Musk is always adding new things to Grok. This time, though, he’s looking forward to hearing from people about how Grok can be improved.
He used the feedback as an excuse to ask people on his own social media site X to give him examples of how Grok, the AI helper made by xAI, could still be better. Musk asked for specific examples of where Grok “goes wrong today” in a post that has been seen over 26 million times.
He said it would help the team find problems faster after seeing how other AIs answered the same question. The odd call for public testing shows that xAI is looking for mistakes that users have pointed out so that Grok can get better promptly.
Latest updates: US Officials Request Travelers to Dress Respectfully and Avoid Pajamas
Musk’s ask for the real time feedbacks
Musk wrote, Please give examples of how Grok’s replies could be better. It would be helpful to show how another AI does it better. He said that some bugs from earlier in the week had been fixed, but the team wanted new cases to happen in real time.
There were thousands of replies to the post, with photos, side-by-side comparisons, and criticisms from people who had tested Grok against ChatGPT, Claude, and Google’s AI tools. Following up, Musk said briefly, We will not rest until Grok work is perfect.
Grok reacts by making mistakes of its own
In an interesting change of events, Grok responded directly to Musk’s post and told itself what it had done wrong. The AI pointed to a question about a decision made by a federal court about Donald Trump’s use of the National Guard in Los Angeles. Grok said it didn’t get the details of the permanent stay, but ChatGPT did a better job of answering the question and used words from Judge Breyer’s order.
Grok also talked about another time it got the source of a trick movie shared online wrong. The bot said that Claude usually does better in situations like this because it checks the video information before replying. “Working on deeper source checks!” was the last thing the AI said in its message. —a sign that xAI wants Grok to be honest about its mistakes and learn from them.
Users bring up more problems
Many people answered with their own cases of flaws. One user said Grok has trouble following instructions exactly, pointing out that when asked to answer a question with just one word, the model gave a full text, which Google’s AI did right. Others pointed out that Grok sometimes explains simple questions too much, which makes it take longer to use for quick information.
Some users said that Grok sometimes loses track of where the conversation is going in longer threads, especially when talks start with more than one suggestion. Another user pointed out that ChatGPT is better at staying relevant over long conversations.
What will Grok do next?
Musk’s public request and Grok’s self-criticism are both signs that xAI is training and improving itself with live user input. As users continue to share cases and similarities, the answers are likely to influence upcoming Model changes as xAI tries to catch up with other AI systems.
