Hey guys, if you are active on Twitter or LinkedIn these days, then you must have seen this word a lot, DeepSeek. When I was scrolling Twitter the other day, I saw a post praising it, then another, and then a lot more. So I decided to deep dive into it and see what it actually is.
A Bit of History
Hangzhou DeepSeek Artificial Intelligence Co., Ltd. is a Chinese company, and they recently released their new LLM (Large Language Model), DeepSeek R1. They have been in this space since 2016, and their first LLM, DeepSeek-coder, was released in November 2023.
Then Why Are We Suddenly Talking About Them So Much?
Performance: Their latest model, DeepSeek R1, has beaten OpenAI’s O1 model, which was the best LLM out there till now.
Cost: DeepSeek R1 provides the same value to customers as OpenAI’s O1 but at a 27x cheaper rate.
Open Source: It’s all open source, and this has made the developer communities on reddit go crazy.
Refer to the images below to see how well it performs compared to other models.
It’s clear from these images that it is no less than a revolution in the AI space now. But how did they actually achieve this?
How Did They Achieve This?
They have just 200 employees, while OpenAI has 4500. They are not an American company with backing from multiple multi-billionaires. They don’t have tons of data. But what they did have was out-of-box thinking, and what they did was nothing less than extraordinary.
Usually, companies follow a supervised approach when training LLMs, but what DeepSeek did is use Reinforcement Learning with supervised learning to achieve this performance. It is a topic for a separate blog where we may dive deep into this topic. But if you want to read about it before I release that blog, I have attached a document for you below.
The Impact That DeepSeek Has Already Created
NVIDIA’s Stock: This is NVIDIA’s stock after DeepSeek R1 released.
OpenAI’s Founder Reaction: This is what OpenAI’s founder, Sam Altman, has to say about DeepSeek.
This might seem very confident, but I am sure OpenAI’s stakeholders would be restless seeing the price of DeepSeek. After all, it’s all a money game, and I am sure they will be changing their price soon too. Although it will be very tough to compete with Chinese pricing, whatever happens will be interesting.
We Have Talked a Lot in Positive About DeepSeek Till Now, but Here Are a Few Things We Can’t Forget:
- It’s Chinese
Conclusion
With that note, I will be ending this blog, but I can guarantee that we have interesting times ahead, and the race to AGI has just begun. All the links and references are attached below. I will soon be back with another blog. Have a nice one till then.
What do you think about Deepseek R1? Share your thoughts in the comments or tweet me at https://x.com/ayushmangarg4
Links
Twitter:https://x.com/ayushmangarg4
LinkedIn: https://www.linkedin.com/in/ayushmangarg
References
Try DeepSeek R1 here: https://chat.deepseek.com/
Read the official release document: https://api-docs.deepseek.com/news/news250120
Video by CodeWithHarry: https://youtu.be/abiYRttaxpo?si=CVnZREZeJmUsYm-y
GitHub (DeepSeek): https://github.com/deepseek-ai/DeepSeek-R1
Document: https://github.com/deepseek-ai/DeepSeek-R1/blob/main/DeepSeek_R1.pdf