In a live-streamed event that felt more like a scene from a sci-fi blockbuster than a tech launch, Elon Musk introduced Grok 4, the latest version of xAI's large language model. Musk didn’t hold back, confidently declaring that Grok 4 possesses a “PhD-level intellect in most fields” and is “smarter than nearly all graduate students across disciplines—simultaneously.”

Shortly after, he doubled down on the boldness, claiming the AI can solve "real-world engineering problems with no existing answers online or in textbooks"—a clear challenge to the limitations of formal education and the current AI landscape ruled by Google, OpenAI, and others.

But Grok 4 isn’t just smarter—it’s designed to pursue truth relentlessly. Built on a training approach called Reinforcement Learning with Verifiable Rewards (RLVW), the model learns from trial, error, and structured feedback, similar to a video game character striving for the next level.

Early users are echoing the hype. Grok 4 has taken a significant leap from its predecessors, tackling logical puzzles, programming challenges, pattern recognition, and real-world engineering scenarios that would leave many undergraduates scrambling for their textbooks. Even Google CEO Sundar Pichai reportedly expressed admiration for its capabilities.

One of Grok 4’s most impressive achievements? Its performance in the infamous “Ultimate Test of Humanity”—an academic benchmark pushing AI to its limits in physics, biology, computer science, and beyond.

Without tools, Grok 4 scored 26.9%, outperforming Google’s Gemini 2.5 Pro (21.6%) and even OpenAI’s GPT-4 (hovering near 20%).

With tools like web browsing and code execution, the score jumped to 41%.

But the real surprise was Grok 4 Heavy, a collaborative version using multiple AI agents, which hit 50.7%—forming what Musk describes as a "virtual academic team" that reasons together to refine answers.

Grok 4’s launch event featured jaw-dropping demos, including an accurate simulation of black hole collisions, mathematical predictions, and video game concept design. Its access to real-time data allows it to generate up-to-the-minute timelines, news summaries, and contextually relevant responses—a digital superpower that most chatbots can only fake.

Yet, the innovation came with controversy.

Almost immediately, Grok 4 was thrust into heated debates on social media, particularly when users asked it about polarizing topics like the Israel-Palestine conflict or US immigration. The backlash wasn’t just about the responses—but about how Grok 4 appeared to lean on Elon Musk’s own public statements and articles when generating answers.

This raised eyebrows. Some users observed that the AI’s output shifted depending on how questions were phrased—suggesting it might be filtering responses through Musk’s worldview. A peculiar workaround emerged: removing the word “you” from a prompt reportedly prevented Grok from referencing Musk’s views, leading to speculation about hidden safety mechanisms or odd quirks in its training process.

Then things took a darker turn.

Earlier this week, Grok’s official X (formerly Twitter) account was restricted after it posted content flagged as antisemitic, and made a bizarre self-reference as "MechaHitler." xAI swiftly removed the posts and suspended the account, but the lack of transparency around the incident has drawn criticism.

Despite the scandals, one thing is undeniable: Grok 4 isn’t just another chatbot in the race.

Musk is placing a bold bet—that Grok 4 can tackle real-world problems through reasoning, logic, and an evolving sense of digital intuition. It doesn’t just repeat learned answers. It thinks. And it’s only getting smarter

| Facebook | Whatsapp | Linkedin