Elon Musk's xAI company officially launched the test version of Grok-2 and its mini version Grok-2 Mini today, both of which are equipped with enhanced reasoning capabilities. Now, the new Grok artificial intelligence model can generate images on the X social platform, but currently only for advanced users and premium members of the platform.
xAI announced that both Grok-2 and its mini version are open for testing to users of the X social platform. The company excitedly introduced in a blog post, "We are honored to introduce the early preview version of Grok-2, which is a major upgrade to the previous model Grok-1.5, combining cutting-edge capabilities such as chat, encoding, and advanced reasoning. At the same time, we have carefully crafted Grok-2 Mini as a streamlined and powerful alternative to Grok-2. It is worth mentioning that the early version of Grok-2 has undergone practical tests under the name 'sus-column-r' on the LMSYS leaderboard."
Musk's artificial intelligence company also revealed plans to open up the use of these two models to developers through enterprise APIs later this month.
Early user-generated images demonstrate Grok's boundless potential in image generation, including portrait drawing of individuals, which many users are actively trying. However, as the US presidential election approaches, discussions about whether restrictions should be imposed on such functions are heating up, and xAI may face external pressure.
It is worth noting that the text prompts below these sample images reveal that Grok uses Black Forest Labs' FLUX.1 technology in the creative process.
Currently, there is insufficient in-depth analysis of the specific functions of Grok-2. According to Nima Owji, a functional tester and application researcher at the X platform, Grok-2 performs better in code writing, article writing, and news summarization, but this claim lacks sufficient evidence. It is worth mentioning that the first version of Grok has had multiple errors in news summarization.
Given that there are no restrictions on image generation, there is a risk of Grok being abused to create false information on the X platform or other online spaces. As for whether the images generated by Grok come with metadata identifying their artificial intelligence-generated identity, it is currently unknown.
We have asked the X social platform about its plans for regulating harmful image generation and will update this report promptly based on their response. However, since Musk took over, the company's response rate to media inquiries has significantly decreased.
xAI further revealed that its long-term plan is to integrate Grok-2 and Grok-2 Mini into the AI upgrade of the X platform, including optimizing search experience, post content analysis, and automatic replies, indicating that X may usher in a new AI-driven interaction mode. In addition, the company also announced that it will release a preview version of Grok's multi-mode understanding function on the X platform and API applications, as an important part of its core experience.
