|
|
"After six months of AI immersion, it's hard to find a more appropriate term than ""stress response"" to describe the state of everyone in the tech industry today—nervous, exciting, and pressured.
A ""stress response"" refers to a series of reactions that a biological organism generates to maintain its homeostasis when encountering external environmental stress or threats. It is a natural reaction that organisms have to adapt to their surroundings and ensure survival. This response can be temporary or long-term.
On July 26th, the official OpenAI Twitter account announced that the Android version of ChatGPT was available for download in the United States, India, Bangladesh, and Brazil, with plans to expand to more countries in the near future. ChatGPT is expanding its reach, gaining more users and stronger user engagement, and the wave of generative AI continues to rise.
In early July, at the 2023 World Artificial Intelligence Conference (WAIC) in Shanghai, a technician from a big-model startup was moving around the exhibition hall. She was looking for a cost-effective domestic chip solution for her company's large-model training.
""We have 1,000 A100s, but it's not enough,"" she told Huxiu.
The A100 is a high-end GPU from NVIDIA and the hardware foundation for ChatGPT's growth. Some publicly available data shows that OpenAI used approximately 25,000 NVIDIA GPUs in the process of training the GPT series models. As a result, evaluating how many A100 GPUs you can get your hands on has become a customary practice in this industry.
Where can you find GPUs? Where can you find affordable computing power? These questions are just a microcosm of the many questions asked at the 2023 WAIC conference.
For everyone who has been ""stressed"" over the past six months, finding more answers about AI at this ""grand event"" is almost a universal desire.
A technician from a chip exhibitor told Huxiu that during the days of the WAIC conference, many product managers came to their ""big model"" booth. They were looking for a product definition for their company's big model business.
According to a research report titled ""Chinese Artificial Intelligence Big Model Map Research Report"" released by the Chinese Academy of Science and Technology Information Institute on May 28, as of the end of May, China had released 79 large models with parameters exceeding one billion. In the following two months, there were a series of AI large models released, such as Alibaba Cloud's Tongyi Wanshang, Huawei Cloud's Pangu 3.0, and Youdao's ""Ziyue."" Incomplete statistics show that there are now over 100 domestic AI large models.
The rush of domestic companies to release AI large models is the best reflection of this ""stress response."" This ""response"" is causing anxiety and spreading to almost everyone in the industry, from the CEOs of internet giants to researchers at AI research institutions, from venture capital fund partners to founders of AI companies, and even legal practitioners in AI-related fields, as well as regulators in data and network security.
For those outside the industry, this may be just a temporary celebration, but in today's world, how many people dare to say they are not involved in AI?
AI is ushering in a new era, and everything is worth reshaping with large models. More and more people are beginning to think about the consequences of technology diffusion.
Funds Pour In, the Flywheel Is Turning
In the month after the birth of ChatGPT, Li Zhifei, the founder of Momenta, made two trips to Silicon Valley. He talked about big models with everyone he met, and he told Huxiu that this is his last ""All in.""
In 2012, Li Zhifei founded Momenta, an AI company that focuses on speech interaction and combines software and hardware. The company experienced the ups and downs of two AI waves in China. During the last wave, Momenta's valuation reached unicorn levels for a time, but it later experienced a period of decline. It wasn't until the emergence of ChatGPT that the long-silent AI industry was given a lifeline.
""Investment money is pouring in"" is the consensus in the industry when it comes to big models.
This has been the industry consensus over the past six months when discussing large models. Qiji Forum founder Lu Qi believes that AI large models are a ""flywheel,"" and the future will be an era where models are everywhere. ""This flywheel has started,"" and the biggest driving force is capital.
In early July, data from the business information platform Crunchbase showed that companies categorized as AI raised $25 billion in the first half of 2023, accounting for 18% of global financing. While this number is lower than the $29 billion raised in the first half of 2022, total global financing in all industries in the first half of 2023 decreased by 51% year-on-year. This suggests that AI's share of financing in the global financing total has nearly doubled. Crunchbase wrote in its report, ""If it weren't for the AI boom sparked by ChatGPT, financing in 2023 would have been even lower.""
So far, the largest financing deal in the AI industry in 2023 was Microsoft's $10 billion investment in OpenAI in January.
Huxiu's public data statistics show that among U.S. startups in the large model space, Inflection AI may become the second-largest startup in terms of financing in the field of artificial intelligence, second only to OpenAI. Following it are Anthropic ($1.5 billion), Cohere ($445 million), Adept ($415 million), Runway ($195.5 million), Character.ai ($150 million), and Stability AI (about $100 million).
In China, there were 456 public investment and financing events in the domestic artificial intelligence industry in the first half of 2023. This compares to 731, 526, 353, 631, and 648 events in the five years from 2018 to 2022.
Another event that triggered the flywheel is ChatGPT's release of the API interface. When OpenAI first opened the ChatGPT API interface in March, there was almost a consensus both inside and outside the AI industry: the industry was about to change. As more applications integrate with large models, a denser forest is growing on top of AI.
""Building large models and building applications should be separate,"" and investors always have a keen sense of smell. To Source Code Capital's Executive Director Chen Runze, AI is a logic similar to the division of labor in semiconductors. After the prosperity of AI large models, a wave of prosperity in AI applications will soon follow.
At the beginning of this year, Chen Runze and his team began to look for companies in China that use large models for applications. However, he found that there were very few such companies. While a lot of capital has flowed into the artificial intelligence industry, tracing the flow of this money reveals that most of it is still concentrated in a few top companies.
""Even now, it's not easy to invest in one or two out of ten projects related to generative AI."" Besides Source Code Capital, Huxiu also spoke with several hard tech investors, and they all said that while they see many projects, truly reliable ones are few and far" |
|
Post time 6-10-2023 12:39:35
