Why DeepSeek-R1 had to be open source, and why it can't beat OpenAI



Chinese AI company DeepSeek

released the inference models 'DeepSeek-R1-Zero' and 'DeepSeek-R1' on January 20, 2025. These models offer performance equivalent to OpenAI's inference model o1, but have significantly lower training costs than o1. They are also open-sourced under the MIT license. AI technology company Lago explains why DeepSeek decided to open-source its AI models.

Lago Blog - Why DeepSeek had to be open-source (and why it won't defeat OpenAI)
https://www.getlago.com/blog/deepseek-open-source



DeepSeek-R1, released by DeepSeek on January 20, 2025, is reported to have performance equal to or better than O1 while requiring only 4% of the training cost compared to O1.




'They achieved this result using a new training method that is more efficient than what OpenAI, Anthropic, and other well-funded competitors use to train their own models,' Lago said. 'Just as Coca-Cola doesn't open-source its recipes, market-leading innovative products should take advantage of them. Why did DeepSeek open-source DeepSeek-R1?'

According to Lago, DeepSeek open-sourced its AI model because it is a Chinese company. In fact, the United States restricts exports of AI chips to China due to concerns that AI-related applications could be diverted for military purposes by hostile countries, including China. Given these factors, the DeepSeek-R1 will inevitably be met with skepticism and apprehension in the Western market. Therefore, by open-sourcing its AI model , DeepSeek hopes to gain the trust of users and developers. Regarding the open-sourcing of such models, Lago stated, 'Open-sourcing is not just a technical move, but also a cultural move to solidify a foothold in the market.'

DeepSeek is also subject to US export restrictions on high-performance AI chips, which means it must maximize results with fewer resources. Therefore, DeepSeek needed to find a more efficient way to train its models.

How did DeepSeek surpass OpenAI's O1 at 3% the cost? - GIGAZINE



On the other hand, companies like OpenAI, Meta, and Google, which have abundant funding, computing resources, and the world's largest distribution, can simply invest in building models without needing to find efficient ways to train them. Furthermore, building large-scale language models more easily and cheaply would mean losing the advantages of these companies, such as their deep funding.

Additionally, in recent years, large-scale language models on the level of GPT-4 have been released one after another. In this situation, OpenAI is facing a challenge from users who wonder whether it is worth paying such a large fee for the OpenAI API compared to models available in open source. Lago points out, 'I don't know whether the AI app I'm using uses LLaMa, GPT, Claude, or Mistral. Why should I pay so much for the OpenAI API when end users don't know the difference?'

In addition, while open source software is low-cost and highly customizable, it requires maintenance and technical knowledge. Proprietary software , on the other hand, is smooth to use but lacks flexibility. Therefore, infrastructure needs to be constantly customized. Open source databases, which allow engineers to audit and modify the code, have become popular in recent years.

The same can be said for large-scale language models. Because the development of AI products requires a high degree of customization, open source models like DeepSeek-R1 are more suitable than proprietary models like o1. 'In infrastructure, open source ultimately wins,' Lago said.

However, according to Lago, it is not naive to expect OpenAI, which has led the AI industry, to become obsolete. OpenAI develops various AI products that serve as the basis for other AI models, such as GPT-4 and o1. OpenAI CEO Sam Altman said, 'DeepSeek's R1 is an impressive model, especially in terms of the features it offers for the price. Clearly, we can deliver models that are better than R1, and it's really exciting to have new competitors! We have several releases coming up.'

OpenAI CEO Sam Altman says, 'DeepSeek-R1 is great, but we plan to release even more amazing AI models' - GIGAZINE



Still, Lago said, 'DeepSeek-R1 may wake up existing, well-funded AI companies and encourage them to find more efficient training methods.'

Related Posts:

in AI,   Software, Posted by log1r_ut