Elon Musk’s artificial intelligence startup, xAI, has launched the latest version of its chatbot, Grok-2. It announced that the beta mode of the new AI assistant will be available for paying X users (formerly Twitter).
xAI released Grok-2 to the public on August 13. It advertised its latest AI model as an improved version of the Grok 1.5 release, which it unveiled in April. The new model comes in two large language model (LLM) versions: Grok-2 and the less powerful Grok-2 mini.
xAI calls its Grok-2 the foremost chatbot model
The latest release is the third iteration of the xAI chatbot since November 2023, when the startup released Grok-1. The company has described this release as a significant improvment to its predecessors.
Independent reviews came back with mixed results in a comparison of Grok-2, ChatGPT-4o, and Gemini 1.5 pro. The comparison was done by Chatbot Arena, a platform for assessing LLMs created by the Large Model Systems Organization (LMSYS Org.)
According to LMSYS Org, Grok-2 matched GPT-4o in certain aspects, excelling in coding, maths, and hard prompts. However, it ranked third overall in all AI capabilities, below GPT-4o and Gemini 1.5.
Wharton Professor Ethan Mollick also acknowledged the capacity of Grok-2 but noted that no AI models have outperformed OpenAI’s GPT-4. He said:
“There are now five GPT-4 class models: GPT-4o, Claude 3.5, Gemini 1.5, Llama 3.1, and now Grok 2. All of the labs are saying there is room left for continued giant improvements, but we haven’t seen any models truly leap above GPT-4… yet.”
The new chatbot is currently exclusively available to X Premium and Premium+ users. However, the startup plans to make its application programming interface (API) available to developers by the end of August.
Grok users can now use it to generate images
One notable improvement in the Grok-2 release is the ability to generate images. Compared to other AI image generators, such as Midjouney and DALL-E, it has fewer restrictions on the images it can generate. Users have already used it to create politically themed images.
The ability to generate images helps Grok-2 in its comparison with other chatbots, as almost all the top models, such as OpenAI’s ChatGPT, Google Gemini, and Meta’s Llama, also create images. Musk even reposted a screenshot of Grok generating a George Washington portrait based on a prompt.
— Elon Musk (@elonmusk) August 14, 2024
The new model highlights xAI’s rate of development as the startup seems to have made up ground in its race to catch up with its competitors. In the two years since it launched, it has become one of the top five AI developers.
However, Musk’s ambitions appear to be set higher. The startup recently shelved plans to use Oracle to build its data center because the tech giant raised concerns about the short timelines and power availability. xAI now plans to build the data center on its own.
Despite the fast-paced developments for its AI startup, there are still major concerns about Grok, particularly its risks, as it appears to have no restrictions. Users have shared images showing that the chatbot generated images of copyrighted brands, products, and politicians, raising concerns that it might flout copyright rules and be used for malicious purposes.
X is facing a lawsuit from a European group, NOYB, over its use of users’ data to train Grok without their consent. Although the social media platform said it has already stopped the practice after it was challenged by Irish regulators, NOYB is proceeding with its lawsuit.
