Last week, the visionary tech mogul Elon Musk and his xAI team actively unveiled their latest brainchild, Grok AI, to the world.
This revolutionary AI platform, which actively pulls its data from the rebranded X-app (formerly known as Twitter), is being hailed as a game-changer. People are saying that with its advanced features, Grok AI will actively compete with all existing AI platforms, including the likes of ChatGPT and Microsoft Copilot.
Indeed, the arrival of Grok AI signals the dawn of a new era in artificial intelligence.
Grok’s Powerful enhanced feature includes:
- Answering any question.
- Suggesting unformed queries.
- Offering real-time global knowledge via x-app.
According to the xAI team they described that:
“Grok is designed to answer questions with a bit of wit and has a rebellious streak, so please don’t use it if you hate humor!.”
ELON MUSK A MAN BEYOND DESCRIPTION:
Elon Musk, a man beyond description, embraces challenges as fuel for motivation.
He demonstrates the power of the human mind, proving his doubters wrong; time and again.
What sets him apart from other billionaires is not just his wealth, but his accessibility. He lives life on his own terms, being a devoted father, a CEO for the people, an advocate for free speech, and maintaining a sense of humor in the face of harsh criticism.
Elon has shown the world that it’s possible to be a billionaire, have fun, criticize, accept criticism, and still manage multiple companies.
These Companies Include the like’s of:
- Tesla (Car Company).
- SpaceX (Space Exploration Company).
- X-app (former twitter, A social networking app, and future everything app).
- Neuralink (Computer and Artificial intelligence powering the human brain).
- The boring company (geared towards creating underground tunneling system).
- Starlink (Private Satellite company, Providing affordable fast internet).
- xAI (Launching Grok X first official AI service and chatbot).
In addition, his numerous investments in companies such as SolarCity, Zip2, PayPal, Amazon, Hyperloop One, CNBC, Stripe, NeuroVigil, XPRIZE, and many others, are just a glimpse into the extraordinary life of this extraordinary man.
You might wonder how one man can manage such a vast business portfolio, attend podcast interviews, government meetings, manage his team, and still find time to post memes and troll online.
The answer is simple:
“will power, effective time management, having the right team to work with, and understanding the importance of leisure.”
CONTROVERSIES, UPHOLDING FREE SPEECH AND THE DANGERS OF A.I:
While some may accuse him of playing mind games, Elon remains focused on his dreams of uniting the world and upholding free speech.
He has warned about the potential dangers of AI, stating that if not properly directed, it could lead to the destruction of humanity.
Despite these concerns, he has shown his philanthropic side by donating over $100 million to the founders of CHATGPT (OpenAI), with the understanding that the app would be free and open source.
However, in March 2023, he expressed his disappointment with OPEN AI for making the services paid, while using data from X (former Twitter) to train their models, despite receiving free donation from him.
These said data were obtained on X-app for free, but the services are charged on ChatGPT 4, while the model of ChatGPT3.5 contains obsolete information dating back to 2022.
Users also face restrictions on the requests they can make, as certain queries result in ChatGPT stating its inability to provide answers. This goes against Elon’s main goal of access to information and free speech.
In response to this, he went back to the drawing board and created something new for his X-app users and citizens: “GROK AI”.
THE BIRTH OF GROK’S AND WHY ?
On November 4th Elon posted on his app (X-app) “Grok”, which garnered over 32-million views, since the post was made. Test users of the Grok have also, praised its power and superior capabilities compared to other platforms.
XAi team further emphasized that:
“A unique and fundamental advantage of Grok is that it has real-time knowledge of the world via the 𝕏 platform. It will also answer spicy questions that are rejected by most other AI systems.”
Despite being in the beta phase, Grok has demonstrated impressive results with just two months of data set training.
The XAi Team believes that the model will continue to improve week by week as more users utilize it.
The aim of Grok AI model according to the team is to:
- Gather feedback and ensure the AI model benefits maximally to all users.
- Empower researchers and inventors to gain access to relevant information’s, processing data, and providing new ideas on how they can go about their research.
In the XAi team opinion:
“We believe that it is important to design AI tools that are useful, to people of all backgrounds and political views, we also want to empower users with our AI tools, subject to the law.”
Undoubtedly, the organization actively gears towards exploring Grok and publicly demonstrates this approach, which helps in the pursuit of understanding.
ROADMAP TO Grok-1:
The XAi Team has been developing Grok-1, the frontier of Language Learning Models (LLM), for the last four months.
The team trained Grok-0, a prototype LLM with over 33 billion parameters, before Grok-1. It nearly matched LLaMA 2’s (70B) capabilities using only half its resources.
In the last two months, they added further functionalities and coding to Grok-0. This yielded significant capabilities and led to the creation of Grok-1.
“A state-of-the-art language model that is significantly more powerful, achieving 63.2% on the HumanEval coding task and 73% on MMLU.” –The team said
UNDERSTANDING THE CAPABILITIES OF GROK-1:
The team conducted a series of test prompts to gain a thorough understanding of Grok-1, using several machine learning algorithms as benchmarks.
These benchmarks included :
- GSM8k (which solves maths problems with prompts),
- MMLU (which involves multiple choice questions),
- HumanEval (which involves completion tasks using Python code), and
- MATH (which involves middle school and high school maths written in LaTeX).
Data curation revealed that Grok-1 strongly outperforms future models, including GPT3.5 and Infection-1. Despite the fact that the model used to train Grok is not enormous, it competes significantly with other models trained on much larger data sets.
Grok was created with a custom training and interference stack based on Kubernetes, Rust, and JAX.
The LLM training at xAI was susceptible to various GPU failures, and to counter this, custom distributed systems are used to identify and handle failures, by the developing team of Grok.
The team also adopted Rust, due to its high performance and rich ecosystem, for building reliable infrastructure. Rust programming ensures code modifications produce working programs that require minimal supervision.
The team is now preparing for a jump in model capabilities, requiring coordination on tens of thousands of accelerators, and has invited interested participants to Join.
Grok Early Access:
The xAi team is said to be actively engaging in the training and refinement of the Model to ensure it meets all necessary requirements.
The team invited a select group of users based in the U.S. to test the prototype, and their feedback will help its developers improve the model.
Although the Model has not been officially launched for public use, we at fixitgearware have proactively enrolled ourselves on the waiting list.
We are eagerly anticipating the opportunity to test the Grok AI and provide our valuable feedback to contribute to its development. We will certainly keep you updated on the results of our experience with the Model.
Elon announced that X-app premium subscribers will get Grok for free.
The team also promised future capabilities that is to be expected from the model, as they refine it into perfection. This features would be able to provide:
Scalable oversight with tool assistance:
With the essential feedback received from humans (users), the new model would able to bolster Grok in looking for external and different resources, while verifying these intermediate steps with external tools, and requesting feedback from users.
Integrating with formal verification to ensure safety, reliability, and grounding:
This will give Grok the ability to reason deeply about real world situations, clearer and more reliable situations. allowing the system to evaluate itself with human feedback or interaction with the real world.
In this context, the xAi developing team stated that:
“One major immediate goal of this approach is to give formal guarantees for code correctness, especially regarding formally verifiable aspects of AI safety.”
Long-context understanding, and retrieval:
Subsequent feature of the model would enable Grok to discover useful knowledge, and yielding intelligent system (capable of giving intelligent responses) .
The team also stated that the improvement will come with the enhancement of the robustness of Language Learning Models (LLMs), reward models, and monitoring systems.
This is to curb adversaries exploitation of AI system vulnerabilities during training and serving phases.
xAi team stated that newer models of Grok would be equipped with additional senses such as vision and audio, which Grok currently lacks. It is expected that enhancing Grok with these new senses would enable it to assist users more effectively and broaden its application to include real-time interactions and assistance.
At Fixitgearware, we firmly believe the impressive capabilities demonstrated by Grok, fuels competition, and its innovation serve as a testament to this belief. Also, this will not only sets a high benchmark in the market but also compels other models to enhance their performance in order to surpass Grok.
We anticipate that this competitive environment will lead to the introduction of more advanced features, thereby assisting all models in striving for market leadership.
Please do let us know in the comment section what are your thoughts about this.