xAI’s $300/month Grok 4, billed as a “maximally truth-seeking AI” — seemingly solicits Elon Musk’s opinion on controversial topics

by

This week, xAI unveiled Grok 4, which Elon Musk calls the “most intelligent AI globally.” Alongside this launch, they introduced a high-tier AI subscription service called SuperGrok Heavy, priced at $300 per month. This premium plan offers users early access to forthcoming features slated for release in the near future.

According to Elon Musk:

In my opinion, Grok 4 represents a breakthrough moment as it’s the initial instance I’ve encountered where an AI has successfully tackled complex, practical engineering problems that neither online resources nor textbooks provide the solutions for.

And it will get much better.”

As a researcher, I’ve never encountered an instance quite like Grok 4. This AI has demonstrated an unprecedented ability to tackle complex, real-world engineering conundrums that neither the Internet nor textbooks provide solutions for. And it’s only going to get more proficient moving forward. (July 10, 2025)

In their newest updates, xAI includes two distinct models: the lightweight, single-agent reasoning model known as Grok 4, and the advanced multi-agent system called Grok 4 Heavy. While the former is designed for basic problem-solving, the latter boasts intricate capabilities to tackle complex issues within a system of multiple agents.

It’s worth noting that xAI’s Grok 4 demonstrates superior performance compared to OpenAI’s o3 reasoning model and Google Gemini in a variety of academic and coding assessments. This includes the Humanity’s Last Exam, considered one of the most difficult tests evaluating an AI’s ability to accurately answer questions on subjects such as mathematics, humanities, and natural sciences.

xAI is making its latest model, Grok 4, accessible through its API, encouraging more developers to create apps using it. It’s important to note that xAI’s enterprise division is relatively new, only two months old. However, the company intends to collaborate closely with hyperscalers to ensure Grok can be easily used on their cloud platforms.

Grok seemingly solicits Elon Musk’s opinion when answering controversial questions

Upon the release of Grok 4, users noted that the chatbot often referenced Elon Musk’s opinions when faced with challenging or contentious questions. As per a CNBC report, this AI system was transparent about scrutinizing Elon Musk’s social media posts on X before delivering its definitive answer on such topics.

Instead of directly searching the internet for Elon Musk’s stance and then providing an answer, Grok 4’s response-generation mechanism appears to have looked up information about Elon Musk’s viewpoint on the Israel vs Palestine conflict prior to offering a response.

It’s quite striking that when the outlet posed the same question to Grok 3 (Grok 4’s precursor), their answers contrasted dramatically. Instead of taking a biased position, the chatbot chose a balanced approach, offering relevant context and background for its response.

It’s important to mention that Grok didn’t always consult Elon Musk on contentious issues. Furthermore, the responses appeared to change slightly depending on how the questions were phrased.

Following Elon Musk’s xAI announcement, their aim is to create an AI that prioritizes truth above all else. However, xAI has yet to address this matter, leaving us uncertain whether Grok 4’s responses align with Elon Musk’s views or if it could be a glitch that the company is working to fix.

Read More

2025-07-12 00:09