Billionaire entrepreneur and “special government employee” Elon Musk’s artificial intelligence company xAI is set to launch Grok 3, the latest version of its chatbot. Musk has described the new chatbot as “scary smart”, and according to xAI, it has significantly improved reasoning, computational power, and adaptability.
According to the AI startup, Grok 3’s development was accelerated by its Colossus supercomputer, which was built in just eight months. The system, powered by 100,000 Nvidia H100 GPUs, provided 200 million GPU-hours for training. This was 10 times more than its predecessor, Grok 2.
READ: China disrupts AI market with DeepSeek: A better, cheaper version of ChatGPT? (January 27, 2025)
Apart from this, Grok 3 also has adjusted its training approach to improve capabilities. It now incorporates synthetic datasets, self-correction mechanisms, and reinforcement learning.
Synthetic datasets are data compilations that are generated artificially instead of being collected from real world sources. They ensure a diverse and controlled dataset while training AI models which claims to improve learning efficiency, and address privacy concerns.
Self-correction mechanisms are techniques that allow AI to catch and correct mistakes on their own. By evaluating output, and comparing them with known and correct responses, the AI model can refine answers over time, reducing errors and improving accuracy.
Reinforcement learning is a type of machine learning where an AI model learns by receiving rewards or penalties for its actions. This helps the system learn through trial and error, maximizing positive outcomes, and improving decision-making capabilities.
READ: This robotics startup has a sustainable solution for refurbishing discarded electronics (January 24, 2025)
xAI claims that these different training methods would help reduce “hallucinations” in Grok 3, through multiple validation steps, and improve accuracy by checking information against reliable sources. xAI also reports that human feedback loops and contextual training have been introduced to ensure more natural and accurate responses.
It has been announced that Grok 3 would debut with a demo on today, Feb.17 at 8 p.m. PST. The launch comes at a time of intense competition between companies over AI, following the launch of a cheaper and popular AI model by the Chinese startup DeepSeek.