Slovenčina 
Čeština 
Shqiptare 
Österreich 
België 
български 
Bosanski 
Українська 
Eesti 
Македонски 
Türkçe 
Deutsch 
Dansk 
Español 
Français 
Suomi 
Ελληνικά 
Hrvatski 
Magyar 
Schweiz 
Irish 
Italiano 
Lietuvių 
Latviešu 
Nederlands 
Norsk 
Polski 
Português 
Română 
Srpski 
Slovenščina 
Svenska 
What makes DeepSeek special is that it is, apparently, much cheaper. While industry leader OpenAI is reported to have spent more than $100 million on the development of GPT-4, DeepSeek developed its model from $6 million. This cost advantage was achieved by the use of high-end Nvidia A100 chips in combination with other relatively cheaper processors.
Experts argue that this figure is misleading because it does not account for broader research and development costs, data acquisition, personnel expenses, infrastructure costs, and hardware investments.
Performance and Features
Just like ChatGPT, DeepSeek works by providing answers to questions, and it is comparable to the o1 model of OpenAI in mathematics and coding. It is a ‘reasoning’ model that thinks and responds as human beings do, making it particularly useful for business and customer service purposes.
Performance compared Source: Artificial Analysis Graphic: Jasmine Cui and Joe Murphy / NBC News
The system is particularly good at demonstrating its reasoning process in solving a problem by breaking down its reasoning process in steps. It can also, like a human solving a puzzle, trace back through its logic and correct earlier steps if needed.
This transparent approach enables users to know how it comes up with its recommendations.
DeepSeek Raises Ethical Concerns
DeepSeek’s rise in the AI landscape is raising important questions about the consequences of its technology, especially given that it was trained on the backbone of expensive hardware that may have been developed to break the US export controls.
While the performance of its R1 model, which is capable of replicating the functionality of OpenAI’s o1 but at a fraction of the cost, is impressive, it also reveals the possibility of unmonitored or unethical practices in the development of AI in the rest of the world.
The company’s capacity to successfully incorporate Nvidia A100 chips, which were previously restricted by US regulations, is a red flag on the authenticity of the company and how such technologies can be used to subvert the international order.
Moreover, the conformity of DeepSeek’s responses to sensitive historical events to the Chinese government censorship is worrying, as it shows that the model could be used to support state-controlled narratives, which could limit the usefulness and credibility of the model globally.
Thus, as markets respond to this perceived threat to American technological supremacy, the wider implications of DeepSeek’s strategy—innovative yet morally sound—need to be thoroughly analysed.
