Deepseek - An Update on the AI model from an education and business perspective
Written by James Adams
It seems that everywhere in the news at the moment we are hearing about how a Chinese startup, Deepseek, has completely disrupted the AI market, knocking $600bn off of the value of chip manufacturer NVIDIA. As a provider of focussed courses on AI, both with our Digital Marketing with AI Course and our Digital Innovator course, we wanted to give our perspective on the news around Deepseek, and what it means for the industry and our view on the news, the model, and what it means for education and business.
What’s the story with Deepseek?
Deepseek (and their Large Language Model named R1) has disrupted the AI market by producing a model that can produce similar efficacy results as other models such as OpenAI’s ChatGPT, Anthropic’s Claude and Google’s Gemini. The main disruption has come (and this is what has caused the issue with NVIDIA’s share price) in that the model is extremely cheap to run and use compared to its competitors. This is in part due to the architecture of the model (how it computes the request you place into it), the costs into training and refining the model, and the processors and the power consumption of those processors. Effectively, Deepseek are saying that they can run the R1 model on a fleet of mopeds, instead of the thousands upon thousands of Porches it takes to run competitors.
Some will question the timing of a Chinese company and the disruption, this given that the US has just announced $500bn investment into the industry.
Should I try Deepseek?
Technology should be about exploration, whether it is a new AI model, a new framework or just reading documentation to build a new project, of course we recommend giving it a look and trying it for yourself. However, we would recommend going into the experience knowing some critical elements about the model, ones that you should be aware of before using the model.
China’s AI Regulation
China has some quite deep and specific AI regulations. Some may argue that this leads to censorship and whilst we teach and users should always review the bias’ that are inherent to any large language model, rumours of focused censorship around specific prompts make us promote even more vigilance from the user when using this model.
Terms of Service
As with all free platforms, it is worth their individual terms of service. We have taken a look over Deepseeks and whilst this isn’t legal advice, would recommend (as with all use of AI in your business)
- Review Data Handling Policies: Ensure compliance with privacy laws before submitting sensitive data.
- Prepare for Service Changes: Have a contingency plan if DeepSeek modifies or terminates services.
- Assess Content Risks: Avoid full reliance on AI-generated outputs without verification.
- Consider Legal Implications: If disputes arise, resolving them in China may be challenging.
- Understand Ownership Rights: Clarify how AI-generated content can be used commercially.
AI technology keeps on developing at a pace, and no sooner had Deepseek released R1, Alibaba has released its Qwen 2.5 model. One thing is for sure, this technology is not standing still and as soon as one solution is deployed, another comes along to keep developers busy.
About the Author
James has 8 years with Fortune 200 US firm ITW, experience of managing projects in China, USA, and throughout Europe. James has worked with companies such as Tesco, Vauxhall, ITW, Serco, McDonalds. James has experience in supporting start-up and scale up companies such as Readingmate, Gorilla Juice and Harvest London. James completed his MBA at the University of East Anglia in 2018.
