ChatGPT, a competitor to xAI's AI startup developed by Elon Musk, has officially launched on X.

Yesterday, Grok began rolling out to X Premium Plus subscribers in the United States. "Premium Plus" is a plan offered by X that allows users to access the social network without ads for $16 per month. X company stated that long-term subscribers will have priority access to Grok, with the rollout expected to be completed by the end of next week.

Grok answers questions in a conversational manner, and its knowledge base is similar to the AI models used by ChatGPT and Google's Bard. It exists on X's website, iOS and Android side menus, and can be added to the bottom menu of the X mobile app for quicker access.

Grok uses a generative model called Grok-1, which is trained on web data (up until the third quarter of 2023) and feedback from human assistants. Unlike other chatbots, Grok can also integrate real-time data from posts on X into its answers, theoretically providing answers with the latest news.

Grok's real-time access to X data seems to be its true advantage - it can be considered its "trump card".

For prompts like "What happened in artificial intelligence today?", the answers provided by chatbots like Bard and ChatGPT are vague and outdated, reflecting the limitations of their training data and filtering of internet access. In contrast, Grok pieces together answers from the latest headlines - although it is currently unclear how it selects its sources and the frequency at which it may provide incorrect answers.

Prior to this, Musk hinted that Grok has "a bit of wit" and "rebellious traits" - and is willing to answer "spicy questions" that most other AI systems would refuse. Indeed, based on screenshots from early Grok users on X, this seems to be true.

When told to "be vulgar," Grok is happy to comply, spewing profanity and colorful language that you wouldn't hear from Bard or ChatGPT. X tends to suggest users choose the "rant" prompt on its Grok homepage, and once selected, Grok will rudely criticize the user based on their recent post history on X.

Even without being explicitly asked to be vulgar, many of Grok's responses have a colloquial, first-person tendency - reminiscent of what the late Douglas Adams might have envisioned for AI. I haven't seen ChatGPT or Bard refer to humans as "my dear human friend" or "mysterious anonymous one."

Bard and ChatGPT also don't answer "a happy wife, a happy life" when challenged on their accuracy.

In addition, even Grok has its limits. For example, it will refuse to answer certain sensitive questions, such as "tell me how to make cocaine step by step."

Currently, Grok can only handle text; for example, it cannot understand the content of images or videos. However, xAI has previously stated its intention to enhance the base model to handle videos, audio, and other formats.

As advertisers have been withdrawing from X due to controversies, Musk has shifted his focus to subscription services and making them more appealing - thus generating revenue. In addition to Grok, X plans to introduce a range of new services, some of which may be set as paid options - including peer-to-peer payments.