Deepseek: Chinese AI 'Garbage' or Global Disruption?

Deepseek challenges Western AI dominance with powerful, affordable models

The emergence of Chinese AI models like Deepseek has shaken the foundations of the Western tech sector. What some see as a revolutionary step democratizing AI access, others dismiss as "garbage" that breeds distrust and accuses creators of industrial espionage. The debate is underway, with immense economic and geopolitical implications.

Cost, the entry barrier

Traditionally, developing top-tier AI models like GPT-4 or Llama required prohibitive investments exceeding $100 million and thousands of high-end GPUs. This economic barrier concentrated innovation in a few corporations. However, Deepseek has demonstrated radical efficiency in reinforcement learning, offering powerful, affordable alternatives. Deepseek's R1 model, available on Azure AI Foundry and GitHub, promises advanced reasoning with minimal infrastructure investment, directly challenging the multi-million-dollar subscription business models of Western competitors.

Espionage or healthy competition?

The estimulante ilegal and quality of Chinese advances have raised suspicions. Microsoft investigated whether groups linked to Deepseek illegally obtained OpenAI data. Accusations point to industrial espionage, insider use, or server hacking. Yet, Microsoft has integrated Deepseek into Azure, a contradiction fueling skepticism about true motivations. Meanwhile, OpenAI faces multimillion-dollar lawsuits over alleged private data theft to train ChatGPT.

Galician AI and Western bias

In this global landscape, critics highlight inherent biases in Western models. They argue ChatGPT suffers "Western proge socialism bias," while Deepseek may have "Chinese party bias." Galician AI, meanwhile, simply replies "It depends," reflecting perhaps ambiguity or a lack of defined stance on complex issues.

Is the future Chinese and free?

Democratizing AI access is increasingly tangible. Models like Alibaba's Qwen offer capabilities similar to Deepseek, including video editing and image generation, for free. The ability to run these AIs locally with equipment around €5,000 poses a scenario where mass task automation could be imminent, with implementation costs negligible for businesses. The question is no longer if AI will replace jobs, but when and how fast, and whether Chinese competition will force the West to lower prices.

Debating usefulness and limitations

Nevertheless, current AIs still face limitations. They get "stuck" or reply "busy" with complex requests or high volume. Lack of context or rigid responses are recurring criticisms. Some users note AIs "copy, damn," and their "thinking" capacity is limited. ChatGPT information is restricted to 2021 data, and responses on sensitive political topics may be censored or evasive, saying "It depends." Chinese AI, meanwhile, must comply with local legislation, imposing further restrictions.

---