Speaker 1: They say no matter what you make, the Chinese can copy it better. It rings true for the AI industry too. For years, artificial intelligence has been America's playground. The US had a clear head start. They built cutting-edge chatbots. They integrated AI into nearly every facet of our lives. But now it seems China is closing the gap. They rolled out new AI models, on par with their Western counterparts, and more importantly, at a fraction of the cost. So how did Beijing pull it off, and what does this mean for the AI race? Last month, a Chinese startup sent shockwaves through the AI world. It's called DeepSeek. They launched an AI model, DeepSeek V3, and it made headlines. It solves problems, codes programs, and even tackles intricate questions. And here's the kicker. It is on par with the likes of OpenAI's ChatGPT and Google's Gemini. And that's not all. This week, DeepSeek took it even further. They unveiled a groundbreaking AI reasoning model. It's called R1. Its performance is right up there with OpenAI's latest models. And the best part? It is free. So where did DeepSeek come from? The company spun out of a hedge fund. Until 2023, no one knew its name, but their first model turned heads. Now the second one is captivating the tech world. R1 is an open weight model, meaning researchers can dig into the algorithm, they can study it and build on it. The training data is not available. Even so, it's still on par with industry standards. Now here's where it gets stunning. The budget. DeepSeek does not have deep pockets. The estimated cost to train these models is around $6 million. Compare that to Meta's Lama, which uses upwards of $60 million. That's 10 times more. What about the chips that DeepSeek used? They weren't top of the line. They relied on NVIDIA's H800 GPUs, one of its less advanced chips. But it's not just about DeepSeek. Alibaba, Tencent, ByteDance, Moonshot, and ZeroOne AI. They're all closing the gap. They're releasing AI models. They're matching the capabilities of U.S. counterparts. But here's the game changer. They're not only on par, they're also cost efficient. And the best part? They're offering it for free. All of this is not a fluke. These advances are a direct response. Washington has long tried to curb China's tech industry. Tariffs, sanctions, restricting access to technology. They've tried everything. The rationale was simple. Without powerful chips and tech, China would struggle. They won't be able to develop AI systems. They would lag behind in the AI race. That's what America hoped for. But the restrictions had an unintended consequence. They triggered a wave of ingenuity. China was denied access to top-tier hardware, so engineers focused on maximizing efficiency. They trained smaller, specialized models. They chose efficiency over brute computing power, and the result is out there for the world to see. DeepSeek is the biggest example. It puts the West on the back foot. If China can make AI systems at a fraction of the cost, why can't they do the same? If Beijing's models are free, why is America charging for them? So for Washington, this presents a quandary. Trump started with tech restrictions on China in his first term as president. Joe Biden doubled down on them. Now Trump plans to add more, all to keep China lagging behind. But China seems to have flipped the script, which raises a question. Are tighter sanctions turbocharging the very progress they aim to suppress? The answer is far from clear. But what is certain is that the AI race is entering a new phase. It's no longer just about who has the best technology. It's also about who can do more with less.

Speaker 2: First Post.