Elon Musk never ceases to amaze. From his state-of-the-art Teslas to backing a memecoin crypto like DOGE to the acquisition of Twitter, his ideas are more eclectic than his portfolio.
In November, he announced his generative AI chatbot- Grok. xAI, with Elon Musk at its helm, has revealed the expansion of its generative AI-based chatbot, Grok AI, into India and 46 additional countries, encompassing Pakistan, Australia, Canada, New Zealand, Singapore, Sri Lanka, and others. Earlier this week, Grok was rolled out to Premium+ subscribers of X(formerly Twitter) in the United States.
Like in the U.S., Grok AI is exclusively introduced for X’s Premium+ subscribers in India, and those interested can sign up for the service. In India, access to Grok requires a Premium+ subscription, priced at Rs 1,300 per month or Rs 13,600 per year.
What is Grok AI?
Grok is a chatbot inspired by “The Hitchhiker’s Guide to the Galaxy.” It’s here not just to answer questions but also to throw in some suggestions. But his answers are not like the run-of-the-mill machine-like. Grok likes to throw in some wit and humor to spice up the conversations.
Grok isn’t shy about taking on “spicy” questions, and it does it with a bit of wit and a touch of rebellion.
Why choose Grok as a name, you may ask? It should not come as a surprise, given Elon has named his son X Æ A-Xii Musk. The term Grok comes from a sci-fi novel, “Stranger in a Strange Land.” In tech lingo, it’s like saying Grok is super good at understanding things, almost like it has a sixth sense for it.
Elon Musk says xAI’s Grok is extremely “curious,” especially about the “truth.” According to Musk, Grok needs a lot of computing power, and it’s built to be a handy tool for regular folks and businesses. It’s supposed to be the best among all the language models out there. Grok is like ChatGPT but with a twist. It can pick up real-time information through the X platform (formerly Twitter). Also, Grok isn’t shy about taking on “spicy” questions, and it does it with a bit of wit and a touch of rebellion. Like xAI warns, “Please don’t use it if you hate humor!”
Speaking of humour, Elon Musk shared a tweet where a user asked Grok for a guide on making cocaine. Grok gave a funny method, suggesting in step 4 to “Start cooking and hope you don’t blow yourself up or get arrested.”
How come Grok is coming up with such cool answers? Grok-1, the powerhouse driving Grok, exhibits remarkable performance in machine learning benchmarks, surpassing its peers in the same computing class. Notably, Grok-1 outperforms models like ChatGPT-3.5 and Inflection-1, trailing behind only larger models like GPT-4. Elon Musk recently shared a Twitter image highlighting how Grok effectively incorporates current information into its responses, showcasing an edge over typical GPT models.
Breaking Down the AI Chatbot Competition
While ChatGPT remains widely used globally, Google Bard and Grok emerge as formidable competitors in the AI chatbot arena. Google recently introduced Gemini AI, a novel model with the ability to produce outputs in diverse formats such as images, video, and audio, especially in its most advanced iteration.
ChatGPT, across its versions, excels at tailoring responses to a wide array of questions, thanks to its GPT-based training. This enables it to understand the nuances of questions, even when presented in simple human language.
Whereas Grok takes a unique approach by enabling branched conversations, allowing users to smoothly switch between different topics. What sets it apart is the ability for users to tweak any part of Grok’s output as if it were the original response. Built on xAI’s Grok-1, it outperforms GPT-3.5 and comparable AI models, according to the company. In this landscape, each AI player brings its strengths to the conversation, shaping the future of conversational AI.
What Sets Grok AI Apart?
xAI says Grok will answer “spicy” questions that most other AI systems reject. This feature is designed to make conversations more interesting and relatable, demonstrating Grok’s ability to adapt to different tones. It maintains that Grok is set to redefine the outlook of language models, introducing a unique blend of curiosity and truth-seeking capabilities.
According to xAI, Grok aims to help humanity understand and acquire knowledge. To achieve its rather broad vision, Grok is powered by the Grok-1 LLM. The initial Grok-0 prototype, trained with 33 billion parameters, is as capable as Meta’s LLaMA 2, which supports 70 billion parameters, claims xAI.
Grok-1 shines in benchmarks:
- 63.2% in HumanEval for coding, showing it’s good at generating (Python) code.
- Scoring 73% in MMLU means it understands the language well.
- In math word problems (GSM8k), it gets 62.9%, better than GPT-3.5 and LLaMa 2 but not as high as Palm 2, Claude 2, and GPT-4.
Overall, it’s strong in coding, language, and math for middle-class problems. While it’s not at GPT-4’s level yet, the improvement from Grok-0 to Grok-1 indicates progress in capabilities.
In a similar test, Grok-1 scored a C grade (59%) in the 2023 Hungarian national high school maths finals, outperforming Claude 2 (55%), but GPT-4 scored higher with a B grade (68%).
| Benchmark | Grok-0 (33B) | LLaMa 2 70B | Inflection-1 | GPT-3.5 | Grok-1 | Palm 2 | Claude 2 | GPT-4 |
| GSM8k | 56.8% 8-shot |
56.8% 8-shot |
62.9% 8-shot |
57.1% 8-shot |
62.9% 8-shot |
80.7% 8-shot |
88.0% 8-shot |
92.0% 8-shot |
| MMLU | 65.7% 5-shot |
68.9% 5-shot |
72.7% 5-shot |
70.0% 5-shot |
73.0% 5-shot |
78.0% 5-shot |
75.0% 5-shot + CoT |
86.4% 5-shot |
| HumanEval | 39.7% 0-shot |
29.9% 0-shot |
35.4% 0-shot |
48.1% 0-shot |
63.2% 0-shot |
– | 70% 0-shot |
67% 0-shot |
| MATH | 15.7% 4-shot |
13.5% 4-shot |
16.0% 4-shot |
23.5% 4-shot |
23.9% 4-shot |
34.6% 4-shot |
– | 42.5% 4-shot |
Source: x.AI
These results clearly show that Grok-1 is more capable than OpenAI’s GPT-3.5, although it falls short of the latest GPT-4 model.
Looking into the Future
Future developments in Grok are set to unveil thrilling features, including upcoming image and audio recognition, as well as image generation. Besides these, Elon Musk has dropped hints about a potential rendezvous between Grok and Tesla vehicles! Musk is optimistic, declaring, “Provided our vehicle AI computer can run the model, Tesla will probably have the most usable inference compute on Earth.”
Wrapping Up
Grok has a big opportunity in India, where the government is proactively embracing AI while being cautious. Prime Minister Narendra Modi, at the GPAI Summit in New Delhi, highlighted India’s enthusiasm for AI innovation, emphasizing the government’s commitment to “AI for all.” Grok could play a crucial role in realizing this vision.
This initiative has the potential to reshape the AI landscape, bringing benefits like better understanding, increased safety, ethical development, and collaborative open-source efforts. However, it’s important to address concerns such as privacy, safety, ethics, accessibility, and potential disruptions.
As Prime Minister Modi rightly pointed out, “AI is transformative, but it’s up to us to make it more transparent.” Eventually Grok AI’s success will depend on how well it handles these challenges, contributing to a safer, more transparent, and ethical AI landscape.
____________
Written By: Poorva Sagar
