一、文章原文
PASSAGE · 同色高亮 = 定位词P1On January 20th, 2025, a Chinese tech company named DeepSeek created a new AI called DeepSeek-R1. This smart program can solve maths problems, write code, and answer questions as well as top models such as Open AI's GPT-40, but it costs much less to build. The team spent only $5.58 million (1/70 of Open AI's cost), used 2,048 computer chips and completed the project in just two months. This shows that good ideas can beat big spending.
P2DeepSeek-R1 learns by trying many times and improving, just like students practise maths. Other AIs need many prepared examples, but R1 gets better by itself. Besides, a simple version of R1 can even run on low-cost computers, helping schools and small companies.
P3DeepSeek-R1 is free to use and modify. On its first day online, over 19,000 programmers worldwide started projects to study it. A German scientist said experiments with R1 cost $10 instead of $300 with other AIs. Even a top scientist at Meta, a famous US tech company, said, "This kind of free AI could change the world."
P4US tech companies are worried. Meta plans to spend $65 billion more on AI research. Experts say China is catching up fast. A report in Nature magazine called R1 "a big step forward", proving that smart ideas beat expensive tools.
P5As a DeepSeek engineer said, "We want AI to be a friend, not something only the rich can use." With R1, China shows that great things come from creativity and hard work, not just money.