Elon Musk’s AI startup, xAI, has introduced the latest version of its chatbot, Grok-2, which is now available on X for users subscribed to the Premium ($8/month) or Premium+ ($16/month) plans. This upgraded model, still in beta, is said to be a major advancement over its predecessor, Grok-1.5, and includes new features like image generation, making it more competitive with similar offerings from OpenAI, Google, and other tech giants.
What’s Happening & Why This Matters
Grok-2 is designed to be more intuitive, versatile, and capable across various tasks, from answering questions to assisting with writing and coding. According to xAI, the model aims to deliver a “state-of-the-art AI assistant” experience, leveraging real-time information from X posts. The company also released a smaller, faster variant called Grok-2 mini, which offers a balance between speed and answer quality.
Grok-2 has already made a strong showing in the AI community. It ranks third on the LMSYS leaderboard, a rating system run by the University of California, Berkeley, where it competes against other major AI models like OpenAI’s GPT-4, Amazon-backed Anthropic’s Claude 3.5, Google’s Gemini Advanced, and Meta’s Llama 3. This ranking is impressive, given that Grok-2 is relatively new to the field.
However, Grok has faced criticism for generating inaccurate information, particularly in the context of elections. Concerns were raised by several Secretaries of State in the U.S., who urged Musk to redirect election-related queries to CanIVote.org to prevent misinformation. xAI acknowledged these issues, noting that Grok-2 includes updates aimed at improving the model’s accuracy and reliability.
TF Summary: What’s Next
With the rapid release cycle of Grok models—Grok-1.5 launched in April 2024, and now Grok-2 in August 2024 — Musk has announced that Grok-3 is already in the works and expected later this year. As Grok-2 learns and *hopefully* improves, its performance in real-world applications and its ability to provide accurate information will be closely watched by users and critics alike.