tech

Recently, xAI, an American artificial intelligence company founded by Tesla and SpaceX's founder Elon Musk, has released a significant upgrade of its large model (chatbot) Grok, named Grok-2.

The model is currently in the testing phase, but xAI believes it represents a major advancement compared to its predecessor, Grok-1.5. In addition, xAI has also released a smaller version concurrently, named Grok-2 mini.

The launch of these two new models demonstrates xAI's rapid development in the field of AI. It has also sparked new discussions in the industry about the impact of AI technology on information dissemination and social media.

Grok-2 is now available to paid users on the social platform X, who need to subscribe to the Premium ($8 per month) or Premium+ ($16 per month) service to use it.

Advertisement

xAI publicly stated that compared to its predecessor, Grok-2 has made significant improvements in intuitiveness, controllability, and diversity, showing excellent performance in answering questions, collaborative writing, and solving programming tasks.The new version of Grok has not only made progress in text generation but also added image generation capabilities, bringing it closer to the products of competitors like OpenAI. This new feature greatly expands the application scope of Grok, making it a more comprehensive AI assistant.

It has been less than a day since its release, and netizens are scrambling to show off their "masterpieces" online.

Due to fewer creative restrictions on Grok-2 (especially political restrictions), many people's works are related to American political figures such as Trump, Biden, and Harris. Of course, the works also include Musk, this "traffic star."

The following are representative works created by netizens. For example, Trump and Musk riding unicorns to fight zombies.Armed Pikachu, provoking Nintendo's "strongest legal department on the planet".

 

Photographs so realistic they can deceive the eye.

 

xAI stated to the media that users can utilize Grok-2 and Grok-2 mini through the newly designed interface on the X platform, and experience a more intuitive and friendly interaction.

 

Its goal is to make Grok-2 the most advanced AI assistant experience, capable of extracting real-time information from posts on the X platform, providing users with the latest and most relevant content.

 

It is worth noting that, according to the ranking in the Large Model Systems (LMSYS) chatbot arena, operated by the University of California, Berkeley, Grok-2 has surpassed many other well-known AI company models, including some models from OpenAI, Anthropic, Google, and Meta.On the LMSYS leaderboard, Grok-2, under the codename "sus-column-r," ranks third, just behind OpenAI's GPT-4o and Gemini-1.5 Pro, but above GPT-4o mini, Claude 3.5, Gemini Advanced, and Llama 3.

This ranking result is surprising because, as a relatively new chatbot, Grok is already on par with the products of industry giants.

However, despite Grok's excellent performance in the LMSYS ranking, this may not fully evaluate its capabilities. Although the LMSYS ranking system is quite popular in the AI industry, it is not good at evaluating the ability to report news, which happens to be an important application for Grok.

In fact, in terms of news reporting, Grok has repeatedly made mistakes before, especially in election-related information.

This has raised concerns about its accuracy and even prompted five state secretaries to urge Musk to redirect U.S. election-related inquiries to CanIVote.org to prevent the chatbot from hallucinating.For these concerns, Musk himself did not directly respond.

However, in the release announcement of Grok-2, xAI particularly emphasized the improvements of the model in providing accurate and truthful information, claiming that it can "correctly identify missing information, reason event sequences, and discard irrelevant posts".

These improvements are aimed at enhancing the reliability and accuracy of Grok, especially in dealing with news and current affairs information.

xAI also announced an important plan, "to open Grok-2 and Grok-2 mini to developers through its enterprise application programming interface (API, Application Programming Interface) later this month."

This move will enable more developers to utilize the capabilities of Grok, which may promote the emergence of more innovative applications.The company also stated to the media that it will more extensively deploy AI functionalities driven by Grok-2 and Grok-2 mini on the X platform, including improved search functions, post analysis, and reply functionalities, among others. This could imply that the X platform will introduce AI replies, further enhancing the user experience.

However, the image generation capabilities of Grok-2 have also raised some new concerns. Widely circulated images suggest that Grok seems to have no restrictions when it comes to creating images of political figures, and many users are leveraging this for their creations.

A realistic situation to consider is the upcoming U.S. presidential election; therefore, the company may be asked to limit related functionalities.

It is currently unclear whether the images generated by Grok have embedded any metadata to indicate that they are AI-generated. This could lead to the spread of misleading or false information, especially during politically sensitive times.

Since its initial release in November 2023, xAI has rapidly introduced multiple new versions. Grok-1.5 was launched in April 2024, and now we have Grok-2. This means that within less than a year, Grok has undergone three upgrades.Previously, Musk had told the media that Grok-3 would be released later this year. This rapid iteration strategy demonstrates xAI's ambition and commitment in the field of AI, and also reflects the rapid development of AI technology.

Although xAI and Musk have highly praised the capabilities of Grok-2, some industry experts point out that the differences between large models may be minimal, and these differences may further diminish as they exhaust the available data.

This perspective also reminds us to maintain objectivity and caution when assessing the performance of AI models.

Currently, ChatGPT is positioned more as a productivity tool, and the release of Grok-2 has opened up new possibilities for AI applications on social media platforms.

By integrating AI assistants into the X platform, users may experience smarter and more personalized social media interactions. This could include more accurate content recommendations, smarter search functions, and AI-assisted content creation.This innovative social platform interaction method, in addition to making the content richer and more interactive, will also have a profound impact on aspects such as news dissemination and the formation of public opinion.

Comments