Elon Musk has recently introduced a new generative AI chatbot called Grok AI. Grok AI was developed by xAI, a generative AI venture established by Musk to enhance the understanding of AI models about the true nature of the universe in order to make them safer. The team describes Grok as an AI modeled after the “Hitchhiker’s Guide to the Galaxy” and designed to answer a wide range of questions and even suggest new questions to ask. They also mention that Grok has a bit of wit and a rebellious streak.
One interesting aspect of Grok AI is its access to Twitter, specifically the now X platform, which allows it to learn from the content posted on the platform. The xAI team claims that Grok can answer spicy questions that are often rejected by other AI systems. However, they caution that Grok AI is still in the early beta stage and has only been trained for two months. They expect the model to rapidly improve over time.
Grok-1, the engine behind Grok AI, is capable of various natural language processing tasks such as question-answering, information retrieval, creative writing, and coding assistance. The xAI team emphasizes the importance of human review to ensure the accuracy of Grok-1’s work, as it lacks the capability to independently search the web. The team also warns that despite having access to external information sources, the model can still generate incorrect or false information.
To develop Grok-1, the xAI team initially trained a prototype large language model called Grok-0, which had 33 billion parameters. Grok-0 performed well, comparable to Meta’s LLaMA 2 model, which was trained with 70 billion parameters. Grok-1 was then trained and further improved based on Grok-0.
Benchmark tests were conducted to evaluate Grok-1’s performance against other models such as PaLM 2, Claude 2, Inflection-1, LLaMA 2, GPT-3.5, and GPT-4. Grok-1 outperformed all models except GPT-4, which had access to a significantly larger amount of training data and compute resources.
Currently, Grok AI is being offered to a limited number of users in the US. Interested users can sign up for a waitlist, and the xAI team plans to release the chatbot to all X users with a subscription.
It’s worth noting that the development of AI safety remains a major concern. Despite Musk’s involvement in xAI and Grok AI, he has previously expressed concerns about AI evolution potentially leading to an extinction event. However, he has also included the director of the Center for AI Safety, Dan Hendrycks, as part of xAI’s advisory team. In addition, several countries, including China, the US, EU, Australia, India, Germany, and France, have signed the Bletchley Declaration, aiming to oversee the safe advancement of AI. OpenAI, the creator of ChatGPT, has also formed a team to prevent potential threats from frontier AI models. The US government, under President Joe Biden’s administration, has issued an executive order to establish rules and oversight measures for AI while allowing its growth.