<h2>Introduction</h2> <p>Artificial intelligence (AI) has become a crucial aspect of modern technology, with various countries and companies investing heavily in its development. Recently, China has been making significant strides in the AI landscape, with its open-weights AI strategy gaining attention worldwide. This approach involves releasing AI models openly, allowing for permissionless use and modification, which is in stark contrast to the locked-down business practices of American companies.</p> <h2>The Moat in AI Models</h2> <p>AI models, as a product in themselves, have very little moat beyond what amounts to brand loyalty and superficial switching costs. Instead, the moat is in the enterprise services that sit around them: the deals and contracts, connectivity with enterprise systems, and quality of life features in an enterprise context. This means that companies can make deals to lock their customers in, but in practice, there's very little long-term technical incentive to use one vendor over another.</p> <h3>Switching Between Models</h3> <p>It's easy to switch between AI models, particularly in the engineering world, where models are accessed via API. For instance, someone could be using ChatGPT today and Claude tomorrow, with very little impact on their workflows. This flexibility allows companies to pick the best model for their needs and change models and vendors if another one becomes better.</p> <h2>China's Open-Weights AI Strategy</h2> <p>China's open-weights AI strategy involves releasing AI models openly, which turns a US-created compute disadvantage into a distribution advantage. This approach commoditizes the layer where American companies make money and creates a far more effective global ecosystem than could be established through locked-in, centralized services.</p> <h3>Benefits of Open Weights Models</h3> <p>Open weights models are not open source, but they are portable and permissionless. This means that companies can host them where they want, experiment with them, alter them, and tweak them to fit their use case. Open technologies can be used permissionlessly and are therefore at the center of more innovation.</p> <h2>Challenges and Implementation Details</h2> <p>The US government has placed export controls on GPUs, and there are strong regulations that prevent sharing certain kinds of data with Chinese servers. This has limited Chinese companies' ability to provide global-scale centralized services, at least not in the same way as American companies like OpenAI and Anthropic.</p> <h3>Overcoming Challenges</h3> <p>Despite these challenges, Chinese companies have been able to train models using sufficient compute resources. The open-weights AI strategy has allowed them to release these models openly, creating a more effective global ecosystem. This approach has already shown significant benefits, with ecosystem benefits across various sectors, from manufacturing to scientific research.</p> <h2>Case Studies and Examples</h2> <p>Moonshot and Alibaba have unveiled models that claim to go toe-to-toe with the best from OpenAI and Anthropic at a fraction of the cost. The rapid-fire releases suggest America's lead at the AI frontier is increasingly tight, just as the technology is becoming central to national security, economic power, and geopolitical influence.</p> <h3>Adoption and Impact</h3> <p>Even without these new capabilities, the strategy has already been working. a16z partner Martin Casado noted in the Economist that there's an 80% chance that any given startup is using Chinese models, and Chinese models are poised to take the lead. This has significant implications for the global AI landscape, with China's open-weights AI strategy potentially disrupting America's AI dominance.</p> <h2>Conclusion</h2> <p>In conclusion, China's open-weights AI strategy is a winning approach to global AI dominance. By releasing AI models openly, China is turning a compute disadvantage into a distribution advantage, creating a more effective global ecosystem, and commoditizing the layer where American companies make money. As the AI landscape continues to evolve, it's essential to consider the implications of this approach and how it may shape the future of AI development and adoption.</p> <pre><code>import torch import torch.nn as nn

Example code for using open weights models

class OpenWeightsModel(nn.Module): def init(self): super(OpenWeightsModel, self).init() self.fc1 = nn.Linear(784, 128) self.fc2 = nn.Linear(128, 10)

def forward(self, x):
    x = torch.relu(self.fc1(x))
    x = self.fc2(x)
    return x

model = OpenWeightsModel() </code></pre>