Elon Musk's AI company, xAI, is on the cusp of releasing its next flagship AI model, Grok 3, which has been in development for several months. Over the weekend, several users on X, including reverse engineer Alexey Shabanov, gained access to the model via X's Grok chatbot app, providing a glimpse into its capabilities.

According to the users, Grok 3 demonstrated impressive logical reasoning and coding skills, successfully answering riddles and generating HTML and JavaScript code for a roulette wheel casino. Although the model wasn't perfect, making a programming error in the code, its performance is a significant improvement over its predecessor, Grok 2.

One user, Chris, reported that Grok 3 was able to answer a nuanced riddle that other open AI models, including OpenAI's GPT-4, have failed to solve. This achievement has sparked excitement among AI enthusiasts, with some speculating that xAI may have caught up with its competitors in terms of AI capabilities.

Shabanov also managed to access Grok 3's system prompt, which defines how the model should behave. Interestingly, the prompt explicitly states that Donald Trump is the 47th president of the U.S., suggesting that xAI may have hardcoded a fix for Grok's political hallucinations.

Grok 3 has been in development for several months, with Musk optimistically slating its release for last year. Although the model missed that deadline, Musk recently announced that Grok 3 had completed pre-training, a crucial stage in a model's development cycle, and is expected to arrive in January or early February.

xAI has been utilizing its enormous data center in Memphis, containing around 100,000 GPUs, to train Grok 3. Musk claimed that Grok 3 was trained with "10x" more compute than Grok 2, indicating a significant investment in the model's development.

Grok, xAI's answer to models like OpenAI's GPT-4 and Google's Gemini, can analyze images and respond to questions, powering a number of features on X. The model recently gained standalone apps and may soon get a "voice mode" that reads the model's responses aloud.

When Musk announced Grok roughly two years ago, he pitched the chatbot as edgy, unfiltered, and anti-"woke" – willing to answer controversial questions other AI systems won't. While Grok has delivered on some of that promise, it has also been found to hedge on political subjects and won't cross certain boundaries. In fact, one study found that Grok leans to the political left on topics like transgender rights, diversity programs, and inequality.

Musk has blamed Grok's behavior on its training data – public webpages – and pledged to "shift Grok closer to politically neutral." Evidence of an "Unhinged Mode" for Grok recently emerged, which will provide responses "intended to be objectionable, inappropriate, and offensive," according to an FAQ page on xAI's website.

Musk has also stated that Grok 3's training data incorporates filings from court cases, improving its ability to understand legal subjects. As the release of Grok 3 approaches, it will be interesting to see how the model's capabilities and biases evolve, and what implications this may have for the AI industry as a whole.

The potential release of Grok 3 in January or early February marks a significant milestone for xAI and the AI community. With its enhanced capabilities and potential for improved political neutrality, Grok 3 may be a game-changer in the world of AI chatbots.