Elon Musk’s artificial intelligence (AI) company on Monday unveiled the latest version of its chatbot, Grok 3, which the billionaire hopes will find traction in a highly competitive sector contested by the likes of ChatGPT and China’s DeepSeek (深度求索).
“Grok is to understand the universe,” Musk said at the start of the Grok 3 launch presentation.
“We’re driven by curiosity about the nature of the universe — that’s also what causes us to be a maximally truth-seeking AI, even if that truth is sometimes at odds with what is politically correct,” he said.
Photo: Reuters
Musk has promoted Grok 3 as “scary smart,” with 10 times the computational resources of its predecessor that was released in August last year.
The flagship product of his X.AI Corp (xAI) was trained on synthetic data and employs self-correction mechanisms that avoid errors — known as “hallucinations” — that plague some AI chatbots and lead them to process false or misleading data as fact.
Across math, science and coding benchmarks, Grok-3 beats OpenAI’s GPT-4o, Alphabet Inc’s Google Gemini, DeepSeek’s V3 model and Anthropic’s Claude, xAI said via a live stream on Monday.
“Grok 3 has very powerful reasoning capabilities, so in the tests that we’ve done thus far, Grok 3 is outperforming anything that’s been released, that we’re aware of, so that’s a good sign,” Musk said in a video call last week with the World Governments Summit in Dubai.
Grok 3 would be made available first to Premium+ paid subscribers of X before rolling out to other users. xAI is starting a new subscription called SuperGrok for the bot’s mobile app and Grok.com Web site, and plans to open-source preceding versions of Grok models as soon as the latest one is fully mature. Musk said he expects that transition to be complete for Grok-3 in a few months.
After the Grok-3 updates were released, Andrej Karpathy, an OpenAI co-founder no longer at the company, posted a preliminary review of the new model on X, writing that it “feels somewhere around the state of the art territory of OpenAI’s strongest models.”
But the computer scientist, who formerly lead AI efforts at Tesla Inc, said Musk’s model also fabricated facts and lagged behind in certain functions. Karpathy said more evaluations are needed over the next days and weeks to get a better idea of the model’s capabilities.
The upgraded chatbot enters a crowded field with countries racing to introduce more sophisticated — and cost-effective — AI products.
Chinese startup DeepSeek shocked the global AI industry last month with the launch of its low-cost, high-quality R1 chatbot — a direct challenge to US ambitions to lead the world in developing the technology.
Grok 3 is also going up against OpenAI’s chatbot, ChatGPT, pitting Musk against collaborator-turned-archrival Sam Altman.
Musk, who also acts as boss of Space Exploration Technologies Corp and Tesla, launched xAI in July 2023 as an alternative to the ChatGPT maker, which he’s publicly criticized for its plans to restructure as a for-profit business.
The company has been canvassing potential investors in a round of funding for about US$10 billion, which would value the company at about US$75 billion, Bloomberg has reported.
Additional reporting by Bloomberg
MANAGING RISKS: Taiwan has secured LNG sufficient to cover 95 percent of electricity demand for next month, UBS said, describing the government’s approach as proactive UBS Group AG has raised its forecast for Taiwan’s economic growth this year to 8 percent, up from 6.9 percent previously, and said expansion could reach as high as 8.6 percent if external energy shocks are avoided. The upgrade reflects a stronger-than-expected first-quarter performance and sustained momentum in artificial intelligence (AI)-driven exports, which UBS said are providing a firm foundation for growth despite geopolitical and energy risks. Taiwan’s GDP expanded 13.69 percent year-on-year in the first quarter, the fastest growth since the second quarter of 1987, the Directorate-General of Budget, Accounting and Statistics (DGBAS) reported on Thursday. On a seasonally
The Fair Trade Commission’s (FTC) ongoing review of Grab Holdings Ltd’s US$600 million acquisition of Foodpanda Taiwan’s operations, announced on March 23, has taken on fresh urgency as industry experts warn that the transaction could embed significant Chinese cybersecurity vulnerabilities into Taiwan’s digital infrastructure through Grab’s deep ties to autonomous-driving firm WeRide (文遠知行). Less than 16 months after the FTC blocked Uber Eats’ direct attempt to acquire Foodpanda Taiwan — citing potential combined market shares of 80 to 90 percent — the emergence of Grab as the buyer has prompted questions about whether the same competitive harm is simply being rerouted
The list of Asian stocks that benefit from business partnership with Nvidia Corp is getting longer, as the region further integrates into the artificial intelligence (AI) chip giant’s business ecosystem. Just in the past week, South Korea’s LG Electronics Inc, Taiwan’s Nanya Technology Corp (南亞科技), as well as China’s Huizhou Desay SV Automotive Co (德賽西威) and Pateo Connect Technology Shanghai Corp (博泰車聯) have become the latest to rally on news of tie-ups, supply-chain participation or product collaboration with the US chip designer. Asian suppliers account for about 90 percent of Nvidia’s production costs, up from about 65 percent last year, data compiled
POWER BUILDUP: Powered by Nvidia’s B200 Blackwell chips, the data center would support MediaTek’s computing power demand and business growth, the company said Smartphone chip designer MediaTek Inc (聯發科) yesterday launched a new artificial intelligence (AI) data center with a maximum capacity of 45 megawatts to meet its rising demand for computing power required to develop new advanced chips for AI applications. The company has completed the first-phase computing power buildup at the data center in Miaoli County’s Tongluo Township (銅鑼), providing 15 megawatts of capacity to support its research and development (R&D) capabilities, despite an industrywide shortage of key components, MediaTek said. Supply constraints have plagued a wide range of key components, including memory chips, solid-state drives, power supply units and central
RSS