DeepSeek Uncovered: The Open-Source AI Challenger from the East

This tech and meet session by Dimitri Casier at HoWest was all about the rise of DeepSeek and what makes it different from other LLMs. Dimitri Casier is a lecturer at HoWest for courses in both the software engineering path and the AI path.

The session started off by Dimitri explaining DeepSeek’s open-source nature and what this means for the LLM. The impact of making the LLM open source was that there are improvements that can be made through crowdsourcing. This also means that the LLM can be run locally by any private user, meaning that all data sent to this privately running chatbot will stay locally on your own workstation also allowing you to “train” it to your own needs. This was demonstrated in the session so the audience could also see the thinking this chatbot does, which is also one of the factors to DeepSeek’s fast rise.

In the session the geo-political impacts were also explained to us. Because DeepSeek was created by a Chinese company and thus also had a need for hardware to support the AI development, prices of graphics processing units or GPUs were raised and there was even a trade ban from the USA to China on certain high tier GPUs. This need for more processing power also played a part in the open sourcing of the LLM.

I had personally used DeepSeek before following this session but decided to stop using it for many things because I didn’t like that it was Chinese owned, implying that my data may be sold to their government. After following this session I downloaded Ollama and was able to run it locally, making me be able to use this LLM along with other ones depending on what task I’m trying to achieve. The session did teach me that I can always try looking further at what resources are available for me to use but also gave me some more knowledge about AI which will be very useful considering the currently evolving IT landscape.

Photos