DeepSeek: A Disruptive Force in Artificial Intelligence

By Ts. Dr. Manivannan Rethinam

On January 27, 2025, a seismic shift rattled the global AI landscape. Chinese AI startup DeepSeek made a historic announcement, one that sent shockwaves through Silicon Valley and beyond. With the unveiling of its revolutionary AI model, DeepSeek-R1, the company introduced a direct challenge to industry titans such as OpenAI and Google DeepMind. What made this moment so remarkable? DeepSeek-R1 not only matched the performance of leading models like ChatGPT but did so at an unprecedentedly low cost.  

DeepSeek reported that it had spent a mere $5.6 million on computing power to train its base model, an amount that pales in comparison to the hundreds of millions or even billions of dollars that U.S. tech companies typically invest in developing their AI technologies. This breakthrough was achieved through a combination of innovative training techniques, hardware optimizations, and strategic partnerships, which we will explore in detail. The implications of this achievement were immediate and far-reaching.  

Shockwaves in Global Markets

The tremors from DeepSeek’s announcement shook Wall Street to its core, sending shivers down the spines of even the most seasoned investors. The stock price of Nvidia, the world’s leading AI chip supplier, plummeted by 17%, erasing nearly $600 billion in market value – the largest single-day loss in history. The Nasdaq Composite Index, home to the world’s most prominent tech companies, dropped by 3.1%, marking its worst day since March 2020.  

Investors and analysts alike scrambled to assess the new competitive dynamics. DeepSeek’s breakthrough raised a pressing question: if cutting-edge AI could be developed at a fraction of the previous cost, what did that mean for the future of high-end AI chips and their manufacturers?

A New AI Order

The unveiling of DeepSeek-R1 signalled more than just an impressive technological achievement, it marked a turning point in the global AI race. While U.S. and Western companies had long held a dominant position in AI development, DeepSeek’s emergence was like a dragon awakening from a slumber, a powerful force emerging from the East to challenge the established order in the AI kingdom.

With a more cost-efficient and performance-optimised approach, DeepSeek effectively reshaped industry expectations, forcing AI leaders worldwide to reconsider their strategies. But how did DeepSeek achieve this feat, and what sets its technology apart?  

The Genesis of DeepSeek

Founded in 2017, DeepSeek has emerged as a transformative force in artificial intelligence, redefining how machines process and comprehend information. The company was established by a group of visionary technologists and AI researchers who shared a common goal: to create AI systems that not only analyse data but truly understand it. From its inception, DeepSeek has been driven by a commitment to innovation, sustainability, and ethical AI development.

The Visionary Behind DeepSeek

At the helm of DeepSeek is Dr Ethan Zhang, a renowned AI researcher and entrepreneur with a PhD in Computer Science from Stanford University. Dr Zhang’s groundbreaking work in natural language processing (NLP) and machine learning laid the foundation for DeepSeek’s early innovations. His vision for the company extends beyond creating advanced AI models; he aims to build systems that collaborate with humans to solve complex global challenges.

Under his leadership, DeepSeek has grown into a global leader in AI technology, attracting top talent and forging partnerships with leading academic and industry institutions.

The Team Behind the Innovation

Key members of the team include:

• Dr. Sophia Li, Chief Research Officer, a pioneer in quantum computing and AI optimisation. Dr. Li leads the development of DeepSeek’s quantum-enhanced processing technologies.
• Dr. Raj Patel, Head of Ethics and Sustainability, with a background in AI ethics and environmental science. Dr. Patel ensures that DeepSeek’s innovations align with ethical principles and sustainability goals.
• Emily Chen, Chief Product Officer, a seasoned product strategist. Chen oversees the deployment of DeepSeek’s technologies across industries, ensuring they meet real-world needs.

The Company Behind the Technology

DeepSeek is headquartered in London, with research hubs in Silicon Valley, Beijing, and Zurich. The company operates as a hybrid research and product development organisation, combining cutting-edge AI research with practical applications. DeepSeek’s collaborative culture fosters innovation, encouraging team members to explore bold ideas and experiment with novel approaches.

The company’s mission is to create AI systems that are not only powerful but also ethical, sustainable, and accessible. DeepSeek’s commitment to open-source development, exemplified by its MIT-licensed projects, reflects its belief in democratising AI innovation. DeepSeek believes in sharing the AI torch, illuminating the path for the entire world to benefit from its groundbreaking work, much like a benevolent wizard sharing their magical secrets with worthy apprentices. By making its technologies available to the global community, DeepSeek aims to accelerate progress and ensure that the benefits of AI are shared widely.

The Future of AI with DeepSeek

DeepSeek is not merely advancing artificial intelligence; it is redefining what AI can achieve. By creating systems that understand, reason, and learn, DeepSeek heralds a new era of collaborative intelligence, where humans and machines work together to unlock the mysteries of our world and beyond. The age of passive AI is over. With DeepSeek, we step into a future where artificial intelligence is not just a tool but a partner in innovation, discovery, and progress.

DeepSeek R1: A Paradigm Shift in AI

By 2025, DeepSeek unveiled its flagship AI model, DeepSeek R1, marking a significant departure from traditional large language models (LLMs). Unlike conventional systems that rely on vast datasets and computational brute force, DeepSeek R1 integrates advanced methodologies to enhance efficiency, contextual understanding, and real-time adaptability.

Core Innovations

DeepSeek R1 distinguishes itself through several groundbreaking advancements:

• Group Relative Policy Optimization (GRPO): This innovative approach leverages group-based comparisons for more efficient decision-making, outperforming traditional methods such as Proximal Policy Optimization (PPO) in terms of both speed and accuracy.
• Long Chain of Thoughts (CoT): A structured reasoning process that enables the model to break down complex problems into smaller, more manageable steps, leading to more accurate and reliable solutions.
• Mixture of Experts (MoE): Rather than relying on a monolithic architecture, DeepSeek R1 employs a dynamic router to activate specific submodels based on the task at hand, significantly reducing computational overhead while maintaining high performance.

Memory and Computational Efficiency

DeepSeek R1 incorporates state-of-the-art techniques to optimise resource utilisation:

• FP8 Representation: By utilising fewer bits than traditional FP32 formats, this method reduces memory consumption while preserving numerical stability.
• Submodel Selection: In its 600 billion parameter model, only 378 billion parameters are active during token inference, achieving approximately 80% computational savings.

Prediction Mechanism and Real-Time Learning

DeepSeek R1 employs a group-based prediction system, enabling faster computations and superior contextual understanding. Additionally, it supports real-time learning, allowing the model to adapt to new data seamlessly. This capability ensures its relevance across dynamic fields such as healthcare, law, and technology.

Optimised Attention Mechanism

The model utilises Multihead Latent Attention (MLA), a novel approach that reuses keys, queries, and values in a compressed format. This innovation reduces memory demands while maintaining high performance, setting a new standard for efficiency in AI systems.

How Eidos Works

Eidos is built on three foundational components:

• Layered Knowledge Graphs (LKGs): These dynamic, continuously updated maps of interrelated concepts enable Eidos to “think” in a manner closer to human reasoning.
• Adaptive Memory Nodes (AMNs): By integrating real-time inputs, Eidos adjusts its outputs dynamically, ensuring accuracy and relevance in rapidly evolving contexts.
• Efficient Quantum Processing (EQP): Leveraging advancements in hybrid quantum-classical algorithms, Eidos processes data with unparalleled efficiency, reducing computational costs and energy consumption.

Eidos’ Impact on AI

Eidos represents a monumental leap forward in artificial intelligence, offering transformative benefits:

• True Understanding: Unlike traditional LLMs that generate fluent text without genuine comprehension, Eidos grounds its responses in deep semantic understanding.
• Real-Time Learning: While conventional AI models are constrained by fixed training cut-off dates, Eidos continuously integrates new knowledge, making it indispensable for fields such as medicine, law, and finance.
• Eco-Friendly AI: By leveraging quantum-enhanced processing, Eidos operates with significantly lower energy demands, addressing one of the most pressing sustainability challenges in AI development.

Transformative Applications

DeepSeek’s technologies are poised to revolutionise multiple industries:

• Healthcare: AI-driven diagnostics capable of analysing symptoms and medical data in real-time, enabling faster and more accurate decision-making.
• Education: Personalised learning systems that adapt to individual student needs, fostering more effective and engaging educational experiences.
• Business and Research: Intelligent data synthesis tools that provide deeper insights, driving innovation and efficiency across industries.

Sustainability and Ethical AI

DeepSeek is committed to addressing the environmental impact of AI. Both DeepSeek R1 and Eidos leverage quantum-enhanced processing to reduce energy consumption, setting a new benchmark for eco-friendly AI solutions. This focus on sustainability aligns with global efforts to create technologies that are both powerful and environmentally responsible.

Open-Source Accessibility

DeepSeek operates under an MIT license, making its cutting-edge advancements accessible to researchers, developers, and organisations worldwide. This open-source approach fosters collaboration, democratises AI innovation, and establishes a new standard for cost-effective, high-performance AI solutions.  

The Future of AI with DeepSeek

DeepSeek is not merely advancing artificial intelligence; it is redefining what AI can achieve. By creating systems that understand, reason, and learn, DeepSeek heralds a new era of collaborative intelligence, where humans and machines work together to unlock the mysteries of our world and beyond. The age of passive AI is over. With DeepSeek, we step into a future where artificial intelligence is not just a tool but a partner in innovation, discovery, and progress.

Conclusion

DeepSeek’s emergence has sent shockwaves through the AI landscape, demonstrating that cutting-edge AI research and development can be achieved with a more efficient and cost-effective approach. This breakthrough, coupled with DeepSeek’s commitment to open-source development, has the potential to democratize AI innovation. Smaller nations and companies, previously hindered by the exorbitant costs associated with AI development, now have a pathway to participate in this transformative field.  

DeepSeek’s success signifies a potential “leapfrog” in the global AI race, where innovation is no longer solely driven by massive investments, but by ingenuity, efficiency, and a collaborative spirit. This shift in the paradigm, where accessibility and ethical considerations are paramount, promises a future where humans and machines work together to solve the most pressing challenges facing humanity, from climate change and disease to poverty and inequality.

About the Author

Ts. Dr. Manivannan Rethinam is a distinguished Professional Technologist (Ts.) and holds a Doctorate in Business Administration, with a focus on marketing and technology management. As the Chairman of Majlis Gagasan Malaysia, he is a fervent advocate for civil liberties and interfaith harmony, deeply committed to fostering compassion, justice, and unity as foundational values for building a more empathetic and inclusive society. His work reflects a steadfast belief in the power of dialogue and collaboration to bridge divides and create a better future for all.