On June 21st, Zhu Xiaohu, a partner at Jinsha River Venture Capital, stated that this generation of AI could be another long-term opportunity, similar to the PC and mobile internet, with a decade-long cycle. If the iteration speed of large models slows down, there will be more opportunities for application innovation. He believes that starting from this year, there may be a lot of opportunities in the AIGC application end. In five years, there may no longer be independent large model companies, only AI application companies or cloud services. Zhu Xiaohu pointed out that in the past 20 years, there have been two ten-year cycles, one for the PC internet and one for the mobile internet, both with a cycle of ten years. This generation of AI could be another long-term opportunity, similar to the PC and mobile internet, with a decade-long cycle. In addition, since GPT came out in 2022, the iteration speed has been fast in the past one or two years. From GPT-3 to GPT-4, it has been 15 months since GPT-4 was released. GPT-5 has not been released yet and has been delayed. Last night, Anthropic just launched Claude 3.5, which has significantly surpassed GPT-4 in many scenarios. Zhu Xiaohu said that if GPT-5 cannot be released by the end of the year, he believes that the stock prices of Open AI and NVIDIA may plummet. But this is good news for Chinese entrepreneurs. Because if the iteration speed of large models slows down, there will be more opportunities for application innovation. In the past, many investors were hesitant to invest in AIGC applications because the rapid evolution of large models could disrupt many startup companies' application scenarios. But if the iteration speed of large models slows down, there will be more opportunities for application innovation. Zhu Xiaohu believes that starting from this year, there may be a lot of opportunities in the AIGC application end. He believes that in the AI era, application companies also have a lot of opportunities, and may even create more value than companies in the previous era. In the United States, this is already very obvious. Currently, first-line large model companies have almost all allied with giants, while second-line large model companies can only sell themselves. The situation in China may be similar. Zhu Xiaohu asserts that in five years, there may no longer be independent large model companies, only AI application companies or cloud service companies. Today's large model companies are hard to exist independently because they do not have their own business models. Zhu Xiaohu found that the United States has a relatively rational attitude towards the implementation of AI applications. Their expectations for the speed of implementation are lower because today's large models still have many errors and low controllability of results. Our domestic large models or open-source models like Llama2 may be equivalent to the level of a vocational school graduate. But after high-quality vertical data training, their performance may be better than that of untrained graduates from top universities. Therefore, for today's Chinese entrepreneurs, ensuring the reliability of delivery is the most important consideration for AI entrepreneurs. Apple launched Apple Intelligence two weeks ago, redefining AI, where the core is not in technology but in user experience. Technology will never be a barrier, especially in China where there are many smart people, once someone develops a technology, hundreds or thousands of people will soon be able to replicate it. Therefore, in China, the core is never technology, but user experience. For Chinese entrepreneurs, Zhu Xiaohu believes that the most critical thing is not to invest too much in basic technology research and development. Because large models are already very powerful and are still being continuously iterated. He believes there are many opportunities for optimization in vertical scenarios. And, do not blindly believe in large investments, as doing AI does not require a lot of funds or computing power. The United States may pursue 100% AI generation, but this is too difficult. It is more suitable for Chinese entrepreneurs to generate 80% with AI and use 20% human work to complement. Zhu Xiaohu points out that one should never underestimate any vertical scenario, as many vertical scenarios contain opportunities. During the 618 period, AI assistants that automatically reply to consumer questions are very useful, the scenario may seem simple but it is very practical. In China, for entrepreneurs, one must find these targeted scenarios. Investment does not need to be large, with two or three people, a standard product may be developed in two or three months, and quickly sold to customers, who will be very satisfied. Therefore, entrepreneurs must find scenarios that make customers ecstatic.