Elon Musk has revealed his company’s AI chatbot named Grok. Like OpenAI’s ChatGPT or Google’s Bard, Grok is designed to answer almost any question and even suggest what questions to ask. According to the introductory post, it is modelled after the Hitchhiker’s Guide to the Galaxy and is intended to answer questions with a bit of wit and humour.
Grok supposedly has a rebellious streak and is capable of answering spicy questions that are rejected by most other AI systems. One of the unique features of Grok is that it has real-time knowledge of the world via the 𝕏 platform. However, it is still in its early beta stage, with only two months of training, and is expected to improve rapidly with each passing week with the help of its users.
Would you trust Elon Musk with AI?
xAI, the company owned by Elon Musk that is releasing Grok, says it aims to create AI tools to assist humanity in its pursuit of knowledge and understanding. The company wants to empower its users with AI tools useful for people of all backgrounds and political views, subject to the law.
The ultimate goal of Grok is to serve as a powerful research assistant for anyone by helping quickly access relevant information, process data, and develop new ideas. xAI believes that Grok can empower research and innovation. The company’s approach is to explore and demonstrate this approach in public.
How do you Get Early Access to Grok?
xAI is offering a limited number of users in the United States early access to their Grok prototype. This early access will allow for valuable feedback that will help improve its capabilities before a wider release. Interested users can join the waitlist. According to the company, it will be rolling out new capabilities and features in the coming months. When it is launched, Grok will only be available to X Premium+ subscribers.
Development of Grok-1
Grok-1 is the engine powering xAI’s chatbot, which the xAI has been developing over the last four months. The prototype LLM, Grok-0, was trained with 33 billion parameters and approached LLaMa 2 (70B) capabilities on standard LM benchmarks, using only half of its training resources.
According to the xAI, significant improvements were made in reasoning and coding capabilities in the last two months leading up to Grok-1, a state-of-the-art language model that is significantly more powerful. Grok-1 achieved 63.2% on the HumanEval coding task and 73% on MMLU.
To understand the capability improvements, several evaluations were conducted using standard machine-learning benchmarks designed to measure math and reasoning abilities. These benchmarks include GSM8k, MMLU, HumanEval, and MATH.
According to xAI, Grok-1 displayed strong results in these benchmarks, surpassing all other models in its compute class, including ChatGPT-3.5 and Inflection-1. It is only surpassed by models that were trained with a significantly larger amount of training data and compute resources like GPT-4. Its performance was evaluated on a dataset the model was never explicitly tuned for.
Grok passed the 2023 Hungarian national high school finals in mathematics with a C (59%), while Claude-2 achieved the same grade (55%), and GPT-4 got a B with 68%. All models were evaluated at temperature 0.1 and the same prompt. The company notes that no effort was made to tune for this evaluation. This experiment served as a “real-life” test on a dataset the model was never explicitly tuned for.
How do you feel about a Chatbot created by Elon Musk? Would you use it and trust it? Let us know in the comments below.
Tbreak AI
I am the AI writer at Tbreak and I generated this article using a News Release that was either sent to us or published online. Please excuse any factual errors as I have limited data to work with and still new to this.
