top of page
Website.png

DeepSeek: AI's Sputnik Moment

Peter Gill.png

Peter Gill

Managing Director

Q. Why should I care about DeepSeek?
A: DeepSeek is a great example of how AI can be developed in a cost-effective, ethical, and accessible way. Its technology has the potential to revolutionize industries like healthcare, education, and finance, making advanced AI tools available to more people around the world. Whether you’re a tech enthusiast or just someone curious about the future, DeepSeek is a company worth watching.

Q. How has DeepSeek impacted the AI landscape globally?

A: DeepSeek’s advancements have been described as a "Sputnik moment" for the AI industry, sparking geopolitical competition between China and the U.S. Their cost-effective strategies have also prompted major tech firms to engage in price wars, and their models match or surpass leading models from companies like OpenAI.

Q. Who founded DeepSeek, and what’s their background?

A: DeepSeek was founded by Liang Wenfeng, a visionary entrepreneur with a background in cognitive neuroscience and electronics from Zhejiang University. He previously co-founded High-Flyer, a successful Chinese hedge fund specialising in AI-driven quantitative trading.

Q. What makes DeepSeek’s AI models unique?
A: DeepSeek’s models are multimodal, meaning they can simultaneously process text, images, and audio. This capability allows them to handle complex tasks like analysing medical scans alongside written reports. The models are also equipped with dynamic real-time learning, unlike static models such as GPT-3, and demonstrate human-like reasoning. DeepSeek supports over fifty languages.

Q. What are some of DeepSeek’s notable technological advancements?
A: DeepSeek has developed cutting-edge techniques like Multi-head Latent Attention and Mixture-of-Experts (MoE) architectures. These innovations enhance model efficiency and reduce computational power usage by approximately 30%. Additionally, the company employs a patented "Fog Learning" method, which anonymises data across devices for secure decentralised training.

Q. What role does open-source play in DeepSeek’s strategy?
A: DeepSeek is committed to an open-source approach, making its models freely available under MIT licences. This decision promotes transparency, collaboration, and innovation within the global AI community.

Q. Has DeepSeek developed any proprietary hardware?
A: Yes, DeepSeek has created its own custom hardware, such as "DragonSilicon" AI chips. This innovation reduces reliance on external suppliers like NVIDIA and enables more cost-effective model training.

Q. Has DeepSeek made any groundbreaking contributions to AI research?
A: Yes, as well as DragonSilicon hardware, there is a patented decentralized training method that anonymizes data across edge devices like smartphones called 'Fog Learning'. In conjunction with Tsinghua University DeepSeek is exploring quantum computing for faster inference. Ancient Text Training on classical Chinese literature and Sanskrit manuscripts enables unique historical analysis.

Q. What are some of DeepSeek’s global achievements?
A: DeepSeek’s AI assistant app quickly became the most downloaded free app on Apple’s iOS App Store in the United States, surpassing ChatGPT. The release of its models even caused a significant market disruption, leading to a 17% drop in NVIDIA’s stock value (the largest ever one-day drop in company value).

Q. What are some of DeepSeek’s partnerships and collaborations?
A: DeepSeek has established partnerships with leading organisations, including Tsinghua University for quantum computing research, AMD for GPU collaborations, and major tech companies like AWS, Stripe and Toyota for AI integrations.

Q. Does DeepSeek have any symbolic identity or branding?
A: Yes, the company’s mascot is a mythical qilin (Chinese unicorn), representing ethical AI and wisdom. This symbol underscores DeepSeek’s commitment to creating AI that aligns with moral and ethical principles.

Q: How successful has DeepSeek become?
A. Their success has been so significant that it's been described as a "Sputnik moment" in AI development - referring to how it's sparked new competition between China and the United States in AI technology.


Q. How does DeepSeek keep its costs down?
A: DeepSeek has found clever ways to be cost-effective. They use something called "sparse neural networks" that need about 30% less computing power than traditional methods. They've also developed their own computer chips (called "DragonSilicon") and use innovative training methods that help them build powerful AI systems for much less money than their competitors. For example, they reportedly created one of their advanced AI models for about $5.58 million, which is remarkably low in the industry.

Q. What’s next for DeepSeek?

A: With its focus on foundational research and innovative advancements, DeepSeek is poised to remain a key player in the AI industry, driving both technological progress and ethical responsibility on a global scale.

bottom of page