Summary: On July 6, the World Artificial Intelligence Conference kicked off in Shanghai, with the theme of "Intelligent Connected World, Generating the Future", focusing on the development of general artificial intelligence and discussing the new future of the industry. At the opening ceremony, Tesla CEO Elon Musk and Huawei Rotating Chairman Ken Hu and other guests delivered speeches on the R&D progress, technology application, and future prospects of artificial intelligence. Below are some of the speeches compiled by the Observer.com, which have not been reviewed by the speakers.
Elon Musk: Secretary Chen Jining, friends in Shanghai, hello!
Artificial intelligence will play a profound role in the development of human civilization in the future. Now, digital computing power is exploding, and one of the key indicators is the ratio of digital, machine, and biological computing. In other words, what is the ratio between the computational intensity of computers and the computational intensity of humans? Every year, this proportion is getting higher and higher, which means that the gap between computing power between machines and humans is widening further.
At the opening ceremony, Musk delivered a speech in the form of a video, with video screenshots
Compared with machine intelligence, human intelligence will account for an increasingly low proportion of all intelligence, which is a fundamental and profound change, the impact of which is still difficult to fully understand, I can only say that this is one of the most far-reaching periods in human history.
I believe Tesla will play an important role in this process. Tesla's Optimus robot (Chinese called "Optimus Prime") is still in the early stages of development, but in the future, we will have a lot of robots.
Now it seems that in the next stage, the number of robots will exceed that of humans. They have much more computing power, which seems to be the trend. This can have both positive and negative effects. The positive impact is that we will enter a very rich post-scarcity era, with no more resource shortages. Humans can get what they want instantly, because the large number of robots will produce much higher productivity than the current human-dominated order. This is a very big change, so we have to be very careful to make sure that the end result is good for humanity.
Tesla's robots focus on doing boring, repetitive, dangerous work and helping humans do things they don't want to do, which is what we're aiming for. I don't want to be overly optimistic that Tesla's role must be important, but Tesla is very interested in self-driving technology and is willing to share this technology license with other automakers. Autonomous driving can completely bid farewell to the boring driving process, which is a technology that is ahead of the times and will greatly increase the use of cars. Under normal circumstances, a family car is used for about 10-12 hours a week, most of the time it is parked in the parking lot; But fully autonomous vehicles may be used for 50-60 hours per week, which is 5 times the usage rate of non-fully autonomous vehicles. At Tesla, we want to provide technology like this, which is why we're licensing fully autonomous driving technology to other automakers.
Tesla's humanoid robot Optimus
Currently, we are testing on U.S. roads, and our technology is already very close to fully autonomous driving without human intervention. When I drive a Tesla car with the latest version of the full self-driving system, there is little manual control required to get from one destination to another. My personal guess is that Tesla may be able to achieve Level 4 or Level 5 full self-driving capability later this year. I've made wrong predictions before, but I think the predictions at this moment are closer than ever to reality.
We need to be very cautious and worry about this kind of deep and comprehensive AI, especially fully autonomous vehicles. For example, in our case, fully autonomous driving with limited AI is challenging. But it is believed that this problem can be solved soon, and it is possible to achieve fully autonomous driving probably later this year. Although there is no 100% guarantee, the trend is. However, this limited AI is completely different from General Artificial Intelligence (AGI), which is very different. It is difficult to define this type of artificial intelligence and classify it as a certain type.
This is not what Tesla is currently specializing in, and other companies are working on it. But I think that's also an area we should focus on. Especially when deep artificial intelligence is equivalent to the collaborative computing power of tens of thousands, or even hundreds of thousands, millions of state-of-the-art computers, such superintelligence capabilities are more powerful than humans, which creates risks and concerns. It may have a very positive future, but there is also a probability of some negative effects, and we need to be as sure as possible that positive prospects can occur and avoid negative factors.
I have always admired the wisdom and drive of Chinese people. Once China is determined to do something, it will do it well, in all fields and industries. Of course, this also includes the artificial intelligence industry. Therefore, I believe that China will have a strong AI research and development capability, which is my prediction.
Thank you very much for joining us today and for inviting me to express my views online, and I hope my speech will not be too boring. Thank you again to friends in Shanghai, thank you to Secretary Chen Jining, and look forward to participating in the next artificial intelligence conference offline.
Ken Hu: The biggest difference between this year's conference and previous editions is that we are in a completely new context. The emergence of ChatGPT at the end of last year pushed artificial intelligence to a new outlet. Artificial intelligence, especially artificial intelligence in general, is currently the hottest topic.
Ken Hu delivered a speech on the theme of "Win-win Artificial Intelligence New Era" at the conference, video screenshot
In the pavilion of the World Artificial Intelligence Conference, many achievements were exhibited around the research of large models and their applications. This shows that the development of artificial intelligence is bringing us new expectations. The direction of future artificial intelligence development has become clearer, and in the near future, artificial intelligence, especially general artificial intelligence, will help us rewrite everything around us. When the direction is clear, path design is critical. Today, taking the opportunity of this conference, I would like to report to you some of Huawei's current thoughts on the future development of artificial intelligence.
Huawei's core consideration now is that in the next stage, we must fully promote the "deep" to "real" artificial intelligence, and there are two starting points: on the one hand, we must deepen computing power and create a strong computing power foundation to support the development of China's artificial intelligence cause. On the other hand, it is combined with large models, from general large models to industry large models of research and innovation, so that artificial intelligence can serve thousands of industries and scientific research.
In the development of artificial intelligence, computing power is the foundation. At present, the mainland is facing considerable challenges in terms of the availability and cost of computing power. Over the years, Huawei has made breakthroughs in root technology by focusing on the root technology of Kunpeng and Ascend. At present, what we are doing is supporting the construction of China's computing power base through architecture innovation, ecological development, and flexible co-construction.
First, we improve computing efficiency through architectural innovation.
On the one hand, at the computing node level, we have launched a revolutionary peer-to-peer flat architecture, which breaks through the performance bottleneck that traditional CPU-centric heterogeneous computing may bring, improves the entire computing bandwidth, reduces latency, and improves node performance by 30%.
On the other hand, at the data center level, we launched Ascend's AI computing cluster in 2019, which brings together the comprehensive advantages of computing, storage, and network through the cluster method, which is equivalent to designing and managing the AI data center as a supercomputer, so that its performance has been greatly improved.
At present, the largest AI computing cluster built in China is the second phase of the "Pengcheng Cloud Brain" in Shenzhen, and it is expected that the scale of the third phase will reach the level of 16,000P (160 billion billion times per second) by 2024. At the same time, Huawei has also deployed several kilocards in the Ulanqab computing center, and it has been measured that efficiency can be improved by more than 10% through clustering with the same computing power.
Huawei Ulanqab Computing Center
Secondly, the development of computing power, ecology is the key means, but also often difficult to overcome bottlenecks. Four years ago, Huawei put forward the strategy of "open hardware, open software, enabling partners, and developing talents" around the development of the entire computing industry. Over the past four years, our cooperation with various partners has achieved certain results. For example, the hardware side has further adhered to the openness, and more diversified modules and boards have been launched this year. At present, more than 30 hardware partners have launched hundreds of artificial intelligence hardware based on Ascend AI, which can meet the needs of scenarios in different industries. At the same time, in terms of software, Huawei has also made great achievements in strengthening basic software through open source, especially innovation for current large models.
Huawei provides a full-process enablement platform to better support scientific research institutions and enterprise customers. At present, our ecosystem has incubated more than 20 basic large models, and also adapted to more than ten mainstream large models in the industry. According to statistics, half of the current large models in China are supported by Ascend AI computing power. Although we have made a lot of achievements in the past four years, there is no shortcut to ecological development, we must move forward step by step, and hope that more companies will join in the future to do a good job in the ecology.
To develop computing power, there is another consideration. The means and modes of development should be diversified and flexible. Therefore, combined with the actual situation in China, we adopt a variety of modes for computing power construction.
On the one hand, in terms of urban computing power infrastructure construction, we have supported local governments, including Shanghai, to build 25 city-level artificial intelligence computing centers. On the other hand, quite a few large enterprises and leading enterprises, such as China Mobile, iFLYTEK, China Southern Power Grid, etc., have the appeal of building their own artificial intelligence computing power centers, and we actively cooperate to help these enterprises establish their own artificial intelligence computing power centers.
At the same time, the demand for artificial intelligence in small and medium-sized enterprises is extremely strong. Faced with the computing power needs of small and medium-sized enterprises, we use cloud services to provide services on HUAWEI CLOUD. Combine a variety of means to truly meet the needs of computing power in this way, and do not let it become a bottleneck in the development of artificial intelligence.
While deeply cultivating computing power, it is also necessary to let artificial intelligence truly enter thousands of industries. On the one hand, we must continue to improve the ability of general large models, and on the other hand, we must build industry models on this basis.
Huawei has introduced a new three-layer large model structure. At the lowest level, the benchmark is the general large model, which we call the basic large model. The basic large model is equivalent to "reading ten thousand books", that is, doing a good job of learning a large number of basic knowledge. On this layer, creating industry models and scene models can be called "traveling thousands of miles". From "reading 10,000 books" to "traveling 10,000 miles", there are still many challenges to overcome, and the key point is to fully match and integrate the knowledge of all walks of life with the big model, and Huawei is working hard with partners in various industries.
At present, Huawei's large model has supported the implementation of more than 400 application scenarios around more than 10 industries such as finance and manufacturing, and hopes to work with more industries in the future to further expand the application scope and further "deepen" and "realize".
In addition to making artificial intelligence serve thousands of industries, it must also serve scientific research. We found that AI can promote the combination of AI with basic disciplines, such as molecular dynamics and biology, by learning massive historical data and scientific knowledge, and encoding mathematical equations into large models, so as to help us discover more scientific laws.
At present, Huawei's Pangu model has released the Pangu Scientific Computing Model, which includes the drug molecule model, the Pangu meteorological model, and the wave model. We worked with scientists and achieved good results.
On July 6, HUAWEI CLOUD Pangu was unveiled at the 2023 World Artificial Intelligence Conference. Pictured from Jiupai Finance
For example, in the field of meteorological research, the Pangu model can complete the forecast of the global meteorological situation in one hour to seven days in a few seconds, which is both fast and accurate. It requires more than 40 years of meteorological data, combined with meteorological industry knowledge for continuous training and correction. The same work is being carried out in multiple scientific research fields, and we hope that through our efforts, we will truly focus on AI, bring more new ideas, new methods and new supplies to scientists and scientists, and inject new momentum into human development.
We are fortunate to have witnessed several waves of technological revolutions, from the Internet to mobility, to cloud computing, to artificial intelligence, each round of transformation has had a profound impact on society. There is no doubt that standing today, general artificial intelligence is giving us infinite imagination space and is leading us into the next golden decade. We hope to join hands with you to innovate together, so that artificial intelligence can better serve thousands of industries, better serve scientific research, and let us win the new era of artificial intelligence together.
This article is an exclusive manuscript of Observer.com, the content of the article is purely the author's personal opinion, does not represent the platform's views, unauthorized reproduction, otherwise legal responsibility will be pursued. Follow the observer network WeChat guanchacn and read interesting articles every day.