# DeepSeek: The AI Disruptor You Should Be Watching

> Source: https://agi.co.uk/emerging-trends-and-technologies/
> Author: Damon Segal
> Published: 2023-07-08T01:00:00+00:00
> Modified: 2025-02-03T10:26:08+00:00

DeepSeek, a Chinese AI startup, is shaking up the industry with its efficient, open-source AI model. Learn why it matters, how it compares to GPT-4, and the risks involved.

DeepSeek, a Chinese AI startup, has made waves with its DeepSeek-R1 model, shaking up the tech world and affecting Nvidia’s stock. But what makes it stand out, and should we care? The short answer is yes. Here’s why.



### What Makes DeepSeek Different?


Unlike resource-heavy models like OpenAI’s GPT-4, DeepSeek prioritises efficiency and cost-effectiveness. It achieves high performance with lower computational demands, thanks to software-driven optimisation and the "Mixture of Experts (MoE)" technique. This approach activates only the necessary computing blocks for a task, saving energy and cutting costs.


DeepSeek’s AI is also open-source, unlike many Western counterparts. Sharing its models and training details fosters global collaboration and innovation, making AI more accessible.



### What Can It Do?


DeepSeek-R1 is built for reasoning, excelling in tasks like maths, coding, and general knowledge. Its benchmark results are impressive:



- 
**Math and Coding** – Scores 79.8% on AIME tests, 97.3% on MATH-500, and outperforms 96.3% of human programmers on Codeforces.




- 
**General Knowledge** – 90.8% accuracy on MMLU tests.






Its potential applications span multiple industries:



- 
**Healthcare** – Assisting in diagnosis, treatment planning, and drug discovery.




- 
**Finance** – Enhancing fraud detection and risk analysis.




- 
**Software Development** – Code generation, debugging, and optimisation.




- 
**Research** – Accelerating AI innovation and data analysis.







### How Does It Compare?


Compared to GPT-4 (OpenAI), Bard (Google), and LLaMA (Meta), DeepSeek stands out for its:



- 
**Lower Development Costs** – Built on a fraction of the budget.




- 
**Optimised Training** – Focused on efficiency over brute-force computing.




- 
**Open-Source Model** – Encourages broader AI adoption and transparency.







### Why It Matters


DeepSeek’s emergence has major implications:



- 
**Democratising AI** – Lower costs make advanced AI more accessible to smaller businesses and researchers.




- 
**Challenging Big Tech** – Western AI leaders now face increased competition.




- 
**China’s AI Ambitions** – Strengthens China’s position in global AI leadership.







### The Concerns


Not everything about DeepSeek is positive. Being a Chinese company, concerns exist around:



- 
**Bias and Censorship** – Its AI may align with Chinese government policies.




- 
**Security Risks** – Reports suggest vulnerabilities in generating harmful content.




- 
**Ethical Use** – Potential for surveillance and misinformation.







### Final Thoughts


DeepSeek is shaking up the AI world, offering a cheaper, more efficient alternative to traditional models. While it raises ethical and security questions, its potential to democratise AI is undeniable. This is one to watch.
