LATEST NEWS:

Grok 3 launches, users share their experience with Elon Musk's 'smartest' chatbot

Grok 3 launches, users share their experience with Elon Musk's 'smartest' chatbot

American billionaire Elon Musk's xAI startup launched its latest chatbot, Grok3, claiming to be the "smartest artificial intelligence on earth."

After the live event for the launch of Grok, several experts shared their experiences with the chatbot in early access and most of them seemed impressed with the chatbot's capability.

American computer scientist Lex Fridman shared that his mind was "blown" after using Grok 3 extensively, reports the Telegraph.


"I should use Grok 3 extensively (at first). My mind is blown, very impressive model? Congratulations to Elon and the team who brought it to life," he wrote on X (formerly Twitter).

In response to Musk's comment on whether he's tried out the voice mode yet, Fridman said: "Not yet. Just the main model and DeepSearch & Think modes. Really impressed with the depth and nuance of the responses, plus how intuitive the interface is. Great job."

Amjad Masad, CEO of Repli, a platform that allows users to write, run and share code, called it a “cutting-edge model.”

"Grok 3 appears to be a more advanced borderline model. This is a huge achievement, especially considering how late in the game they started," he wrote on X.

Andrej Karpathy, CEO of Eureka Labs and former director of AI at Tesla, shared his initial impressions of the Grok 3 model at X, noting that it demonstrates near-state-of-the-art performance, especially with the “Thinking” feature enabled.

He praised its ability to solve complex coding tasks, such as generating a Settlers of Catan-style hex grid, which many rival models fail to do. The Grok 3 also answered math-based questions accurately and performed well on puzzles like tic-tac-toe tables.

However, he struggled with more abstract challenges, like decoding hidden messages in Unicode or generating SVG images of pelicans on bicycles.

“He solved some tic-tacs I gave him with a pretty nice/clean chain of thought (many SOTA models often fail these!). So I raised the difficulty and asked him to generate 3 “tricky” tic-tacs, which he failed at (generating meaningless tables/texts), but then so did o1 pro,” he added.

The model's "DeepSearch" feature seemed to impress Karpathy with its ability to retrieve information on trending topics, like Apple launch rumors or Bryan Johnson's choice of toothpaste.

But it occasionally hallucinated sources and failed on some specific questions, putting it roughly on par with Perplexity's DeepResearch but behind OpenAI's equivalent offering.

Karpathy concluded that Grok 3 approaches the capabilities of OpenAI's o1-pro model and outperforms competitors like DeepSeek-R1 and Gemini 2.0 Flash Thinking — an impressive feat for a team that has only been operational for about a year.

Shivon Zilis, who is the CEO of Neuralink and the mother of Musk's two children, shared her first impressions hours before the launch was announced and called it "the most unexpectedly rewarding hours" of her life.

She said the chatbot helped her delve into topics like electricity in biology and quantum entanglement, answering all her questions – no matter how “dumb” they felt – without judgment.

Zilis appreciated how Grok 3 gently corrected her mistakes and kept the atmosphere upbeat, which made her feel more confident about exploring topics she would want to learn for years.

Grok3 is now rolling out to X Premium+ users and will be rolled out to other users gradually. Musk, during that event, warned users that the AI ​​chatbot is still in the development stages and will have more improvements and features, including voice mode. /Telegraph/