On Friday, Musk informed that xAI was all set to unveil its first AI model with a firm view to rival industry leaders like OpenAI and Google and their popular AI chatbots ChatGPT and Bard. A day later bot named Grok was launched as a “very early beta product” meant for X’s paid subscribers. xAI stated that Grok “is designed to answer questions with a bit of wit and has a rebellious streak” and asked users who “hate humor” to steer clear. Musk has claimed that the Grok bot has a “massive advantage” over other AI models such as ChatGPT as it has “real-time access” to the X platform. The companies X and xAI have similar names and owner is common too but still both companies remain distinct entities. On its website, xAI makes it amply clear that it is separate from X Corp (twitter till recently) but assures that will “work closely with X (Twitter), Tesla, and other companies to make progress towards our mission.” According to Musk, xAI will be “built into the X app” and also be “available as a standalone app”.

Just last week Musk had stated that AI was “one of the biggest threats to humanity” and now ironically he has announced a bot to take on ChatGPT. It is a well-known fact that Musk was cofounder of the company that launched ChatGPT, OpenAI. To present the playful tone of Grok, Musk posted a screengrab of a query to Grok. Screengrab showed that when asked to provide step by step guide to make cocaine, Grok provided answer that included steps like “obtain a chemistry degree” and “set up a clandestine laboratory in a remote location”.  But at the end chatbot adds: “Just kidding! Please don’t actually try to make cocaine. It’s illegal, dangerous, and not something I would ever encourage.” Musk stated that Grok, which is still in early testing and not accessible to the general public, would eventually be released to subscribers to X’s Premium+ (top-tier subscription service of X).

The xAI team said Grok was powered by a large language model called Grok-1 said the xAI team. The team claims that Grok-1 has already surpassed GPT-3.5, the language model used in the freely available version of ChatGPT, on several benchmarks including solving middle-school maths problems. Team conceded that Grok-1 lagged behind GPT-4 (most powerful ChatGPT model). “It is only surpassed by models that were trained with a significantly larger amount of training data and computer resources like GPT-4.”