Why China is so "benevolent" about AI?Ismail Kizir

Why Chinese Companies release open source local models?

 I've been thinking for a long time. I had a very basic explanation. With help of Gemini, I could write a detailed, informative article

I asked this question to Gemini. It gave me a very nice, detailed and interesting answer; yet, i still have an additional explanation at the end:

The perceived "benevolence" of Chinese companies in releasing open-source Local Language Models (LLMs) is driven by a complex mix of strategic national interests, commercial goals, and a response to the global AI competition, rather than purely altruistic motives.1

Here are the key reasons why Chinese companies are releasing open-source models:

1. Strategic National Interest and Self-Reliance

Countering US Export Controls: Facing US restrictions on advanced semiconductors and other critical technologies, China is focused on building an "independent and controllable" domestic AI ecosystem (known as "self-reliance and self-strengthening").2 Open-source development fosters a domestic ecosystem where Chinese companies and developers are less reliant on proprietary, closed-source models and frameworks from Western companies like OpenAI.3

Treating AI as Infrastructure: China views AI as a critical national infrastructure, not just a commodity.4 By encouraging the open-sourcing of foundational models, the government supports the rapid development and application of AI across various domestic industries (military-civil fusion).5

2. Commercial and Ecosystem Strategy

Ecosystem Lock-in and Platform Play: By releasing high-quality open-source models (like Alibaba's Qwen or DeepSeek), Chinese tech giants aim to establish their models as the default standard for developers, especially in the "Global South" (Asia, Africa, Latin America).6 This strategy, similar to how Android gained market dominance, locks developers into their ecosystem.

Monetization Pathways: The open-source models are often a "marketing tool" for more lucrative services.7 Companies can monetize later through:

Paid API Access for larger, cutting-edge models.

Customization and Fine-tuning Services for enterprises.8

Cloud Computing Services (e.g., Alibaba Cloud, Tencent Cloud) that developers need to run and deploy these models.9

Data and Feedback Loop: Open-sourcing models encourages a wide community of developers to use, fine-tune, and deploy them in real-world applications.10 This process generates invaluable feedback, usage patterns, and data on failure points that help the original company quickly improve its next-generation models.11

Focus on Practical Applications: Some analysts suggest Chinese companies, constrained by limited access to the highest-end GPUs and capital for massive AGI projects, are strategically focusing on smaller, more efficient open-weight models optimized for practical, day-to-day applications and deployment on consumer or enterprise devices.

3. Global Influence and Soft Power

Global Leadership and Soft Power: Releasing highly competitive, open-source LLMs positions China as a major contributor to the global AI field.12 This act of "sharing" helps project soft power and can win adoption in countries looking for alternatives to US-based closed-source platforms.

Shaping Global Governance: China has advocated for "open and inclusive AI development" and has proposed global governance frameworks, positioning its open-source push as a way to ensure AI development "should not be a game of rich countries."13

4. Competitive Advantage

Undercutting Competition: The availability of high-performing, free open-source alternatives erodes the pricing power and potential monopoly of closed-source Western competitors like OpenAI.14 This effectively "commoditizes the complement" (the foundational model), shifting the commercial battleground to applications and services built on top of the models, where Chinese firms are also highly competitive.

At 4th article, Gemini explains what I have in my "implicitly". China thinks it can't compete with huge soft power of USA even if it has more capital. Once masses starts using commerical, "centralized" models of US giant companies by paying them, it will be difficult to revert this back. So, they try to shift the commercial battleground to their domain of expertise. Think it like Google: Once people are familiarized with a product and one company becomes a monopoly, it will be almost impossible to break that monopoly later, as in Google example.

Yorumlar

Yorum Gönder

Bu blogdaki popüler yayınlar

  Uluslararası düzeyde olan biteni anlamaya çalışanlara naçizane tavsiyem, "1. Dünya Savaşı"nın bitmediğini ön kabul olarak almalarıdır. Bu yazımın, bir önceki yazım ile birlikte değerlendirilmesinin faydalı olacağı kanaatindeyim. Çok küçümseriz ilkokul eğitimimizi fakat, çok doğru yönleri vardır. Bizim zamanımızda, İran ve Çin hakkında; çok basit klişelerle fakat çok derin tarihi perspektiften biraz da karikatürize edilmiş şekilde bir kaç "temel" belletilirdi: "İran'a her zaman dikkat et!"; "Çin'in entrikası bitmez! Hiçbir zaman güvenme". Fakat konumuz tam olarak bu değil. Artık çoğumuzun unuttuğu "Sevres Antlaşması" da en temel klişemizdi. Hatırlar mısınız? Birinci Dünya Savaşı'nın çıkma sebebi; çok kabaca, sömürgecilikte son noktayı; hedefe oturttuğu Osmanlı ve Avusturya-Macaristan İmparatorluklarını bölüşmeyi kafaya koyan İngiliz İttifakı ile(ki Osmanlı, bu ittifakın parçası olalım dediği halde, kabul etmeyip ...
Devamı

Visual proofs of Hohha Dynamic XOR Encryption Algorithm

Resim
In order to better understand the meaning of those images and to compare Hohha Dynamic XOR with other algorithms, please visit the great article of Derek Zimmer at  https://vikingvpn.com/blogs/security/visualizing-weak-encryption-experiments-with-aes  about the subject Orjinal Image: Viking Viking Encrypted with Hohha Dynamic XOR Jumps=2 Key Body=64 Viking Encrypted with Hohha Dynamic XOR Jumps=2 Key Body=128 Viking Encrypted with Hohha Dynamic XOR Jumps=2 Key Body=256 Viking Encrypted with Hohha Dynamic XOR Jumps=3 Key Body=64 Viking Encrypted with Hohha Dynamic XOR Jumps=3 Key Body=128 Viking Encrypted with Hohha Dynamic XOR Jumps=3 Key Body=256 Original image: Panda Panda Encrypted with Hohha Dynamic XOR Jumps=2 Key Body=64 Panda Encrypted with Hohha Dynamic XOR Jumps=2 Key Body=128 Panda Encrypted with Hohha Dynamic XOR Jumps=2 Key Body=256 Panda Encrypted with Hohha Dynamic XOR Jumps=3 Key Body=64 Pa...
Devamı