Elon Musk 's AI startup xAI has released a new AI chatbot called Grok to rival the growing popularity of generative AI. Grok works similar to ChatGPT and answers questions posed by users based on information generated by a large language model.
Grok artificial intelligence from Musk-led xAI company is already used by a small number of users. However, Grok is claimed to outperform ChatGPT-3.5 on a variety of academic tests, including math and coding.
Grok is powered by a new engine called Grok-1, which provides real-time access to information via the X (Twitter) platform. This keeps him informed about current issues around the world. However, the "accuracy" of this information seems to be questionable. To evaluate the tool's capabilities, xAI, on the other hand, conducted a series of evaluations using standard machine learning metrics designed to measure math and reasoning abilities.
Grok outperformed all other models in its compute class, including ChatGPT-3.5 and Inflection-1. It was surpassed only by models trained with much larger amounts of training data and computational resources, such as GPT-4. In addition to these benchmarks, xAI also ranked Grok in the 2023 Hungarian national high school mathematics finals. While Grok passed the exam with a C (59%), Claude-2 achieved a similar grade (55%) in the same exam. OpenAI's GPT-4 managed to get a higher score (68%).
In its post shared on the platform, xAI states that Grok was designed to answer questions in a humorous way and therefore has a rebellious streak. It is also stated that the tool will also answer complex questions that are rejected by most other artificial intelligence systems .
However, Grok is actually a very early beta version. According to the company, it was created with only 2 months of training. The company is also working to implement more security measures to ensure Grok is not used with malicious intent. You'll need to get an X Premium Plus subscription for $16 per month to try Grok when it becomes more widely available, but it's currently only available to a limited number of users in the US.
