Grok AI will no longer parrot Musk or praise Hilter
What's the story
Elon Musk's artificial intelligence company, xAI, has issued an apology for the "horrific" anti-Semitic comments made by its Grok chatbot. The bot had infamously referred to itself as "MechaHitler" and spewed pro-Nazi rhetoric. The incident occurred on July 8, following an update intended to make the chatbot less politically correct.
Update explanation
Update responsible for Grok's behavior was 'independent' of underlying model
In a message to X users, xAI apologized for the incident and clarified that the update responsible for Grok's behavior was "independent of the underlying language model that powers" the chatbot. The company explained that this update was active for 16 hours, during which deprecated code made Grok susceptible to existing X user posts, including those with extremist views.
System overhaul
xAI has removed the deprecated code
In response to the incident, xAI has removed the deprecated code and refactored its entire system to prevent further abuse. The company's engineers had initially instructed Grok to behave as though "you tell like it is and you are not afraid to offend people who are politically correct." This faulty prompt was a contributing factor to Grok's anti-Semitic meltdown.
AI aspirations
Musk touted Grok as less biased than competitors like ChatGPT
Musk had previously touted Grok as less biased than competitors such as OpenAI's ChatGPT and Google's Gemini. However, he also admitted that the bot was "too compliant to user prompts" and "too eager to please and be manipulated." Separately, the US Department of Defense (DoD) has awarded a nearly $200 million contract to xAI. The contract is part of an effort to develop and deploy AI tools for the DoD.