Before the Spring Festival, the large-scale ChatGPT model launched by the artificial intelligence organization OpenAI caused quite a stir in the China AI industry.
There is a wave of artificial intelligence every five or six years. The last wave of AlphaGO shocked everyone, and this wave is ChatGPT. But these two times, everyone's mentality was very different. When Google's AI defeated the world champion of chess, everyone watched it as news, but this time many people experienced it as consumers.
In addition to lamenting the fantastic experience of ChatGPT, the Chinese industry is also rapidly reflecting and taking action: "Where are our gaps? How will we face up to and improve the shortcomings? What are the opportunities for Chinese practitioners?"
Baidu Will Launch the Chinese Version of ChatGPT
It is reported that Baidu will launch an artificial intelligence tool similar to ChatGPT in China in March. According to insiders, the software will be released as standard application software but may later be incorporated into Baidu's search engine. It is reported that Baidu also plans to establish a separate website for users to use the tool.
Even though there is no official response from Baidu, the capital market has taken the lead in seeing the value brought by the "Chinese version of ChatGPT". Some commentators pointed out that once the Baidu version of ChatGPT is completed, Baidu will have both ChatGPT-like technology and search market advantages, and it will become China's "OpenAI+Google".
The Carnival of Capital VS the Gap of Reality
After the launch of ChatGPT, a senior person from Baidu said that he was "not interested" in talking about ChatGPT; a founder of an artificial intelligence company noted that in the face of ChatGPT's fantastic performance, he was itchy and confused, and had insomnia. He admits that there is still a long way to go in terms of the scale and effect of the model.
Someone asked the same question to a large model of a Chinese manufacturer and ChatGPT simultaneously. ChatGPT far surpassed the Chinese model in terms of logic and completeness of the answers. Moreover, in terms of reply speed, ChatGPT is also ahead.
Le Cheng, CEO of Tebcut Technology, who is engaged in digital human research and development, believes that no large-scale model in the world can compete with ChatGPT. The consensus in the industry is that the gap is more than two years. Let's talk about something other than overtaking in curves; it is more important to catch up as soon as possible.
Although some artificial intelligence veterans believe that China and the United States are "equal" in terms of the technology involved in ChatGPT, Liu Qun, the chief scientist of speech semantics at Huawei's Noah's Ark Laboratory, admitted frankly in an open discussion that there is still a technical gap between China & US. One of them is the gap in the basic model itself. China has trained many trillion models or hundreds of billions of models, but the adequacy of the training is far from enough. "I estimate that until now, no model can eat as much data as ChatGPT."
The Next Thing To Catch for China: Pure Innovation and Long-termism
Huang Minlie, a tenured associate professor of the Department of Computer Science and Technology of Tsinghua University, mentioned that after GPT-3, all OpenAI models are not open source, but it provides API calls. OpenAI has done one thing right in this process: establish a flywheel between real user calls and model iterations. It attaches great importance to the call of real-world data and the iteration of these data to the model. Of course, in the process, it has supported many startups in the United States and established an ecosystem.
"Look at our large-scale model research in China. Company A trained one, and Company B also trained one. After the advertising and news coverage, no one will care about the user experience afterward. At least we haven't seen a better company yet to completely turn the flywheel of data and models. Therefore, I think this is difficult for us to catch up with ChatGPT." An industry insider said frankly.
In addition, people in the industry have mentioned the computing power problem. Due to issues such as GPU chips, China's domestic computing power has been stuck to a certain extent. Even domestic leading companies, compared with Google and others in terms of computing power, the gap is quite apparent.
In terms of data quality, there is still a clear gap between the Chinese data quality of the entire Internet compared to English. "We may find a way to complement the data between Chinese and English languages." Some people in the industry said.
In fact, from the perspective of principles and methods, the industry understands what they do, it is not the case that the United States can do but we can't. "Liang Jiaen, chairman of Unisound, said, like OpenAI and DeepMind, they may be the only two institutions in the industry, whether in innovation, investment, determination, or top talent pool, they are as consistently excellent. "We might see success, but it may already be many failed attempts. "
Some senior AI practitioners believe that OpenAI has made a very firm investment in the stage when there is no prospect and no apparent effect. On the contrary, China tends to follow quickly after a technological breakthrough. "The first thing people in China think about is how we can use it now, but not in the long-term; however, those American people will invest in it for a long time when there is no hope."
"This thing is worth learning. We need enough money and a group of passionate talents to continue accumulating strength in one direction. This is an essential condition. "Huang Minlie said.
Disclaimer: All the information on this website is provided on an “as is” and “as available” basis, and you agree to use such information entirely at your own risk. Monisight gives no warranty and accepts no responsibility or liability for the accuracy or completeness of the information and materials contained in this website. Under no circumstances will Monisight be held responsible or liable in any way for any claims, damages, losses, expenses, costs, or liabilities whatsoever (including, without limitation, any direct or indirect damages for loss of profits, business interruption, or loss of information) resulting or arising directly or indirectly from your use of or inability to use this website or any websites linked to it, or from your reliance on the information and material on this website, even if the Monisight has been advised of the possibility of such damages in advance.
Commentaires