"My appearance is a cute girl, but please don't be fooled by my appearance, because I have a strong learning ability and can do many things, such as chatting with you and helping you with housework......" The sweet voice came from a "beauty" with short ear-length hair and silver armor with heavy makeup, named Xiaoqi. On April 25, when the 2024 Zhongguancun Forum opened in Beijing, the bionic robot stationed at the front desk to serve visitors became one of the busiest "people" in the audience.
Of course, it's not just Xiaoqi. It is the coffee robot with the intelligent "iron arm" that prepares the coffee for the guests, and the digital human anchor who broadcasts the news for the audience...... The most direct impression of the Red Star News reporter who went to Beijing to participate in the conference was that he "entered the world of human-computer collaboration in science fiction movies in advance".
In fact, the penetration of artificial intelligence into social life has been presented in all aspects. In particular, the emergence of generative artificial intelligence such as ChatGPT and Sora has made people believe that all walks of life will undergo subversive changes. Therefore, at this year's Zhongguancun Forum, AI is everywhere, and everyone talks about AI.
10 years is a short time, but it is long enough for the future development axis of artificial intelligence. In 10 years, how smart will robots be, and how will they enter life? Red Star News reporters talked to a number of leading experts at the forum to discuss how artificial intelligence will change the world.
Bionic robots at the Zhongguancun Forum
future
Like owning a smartphone, go and have a genius assistant
In fact, generative AI technology has emerged in the last 5 years, but in the last 2 years alone, it has been able to bring more intelligence than any genius. "Two years ago, the level of large AI models could probably reach the level of ordinary people, that is, they could answer 34 out of 100 questions correctly. Kai-Fu Lee, chairman and CEO of Sinovation Works, said that after two years of progress, the large model of the best company can answer 90% of the questions correctly, "It is conceivable that if you are given another two years, the AI large model will have a huge opportunity." ”
Kai-Fu Lee's speech
In the PC era, a computer was placed on every desk, and in the mobile Internet era, people could carry a mobile phone that could connect information and people. Kai-Fu Lee said: "The AI era will put an IQ300 genius assistant by your side to help you think about problems and solve problems. "And when such a genius emerges, productivity will increase like never before.
Nowadays, people tend to imagine the large models that may appear around them as chatbots. Kai-Fu Lee predicts that in the next 10 years or so, chatbots will evolve into super assistants with full-modal understanding capabilities, able to understand the world in depth, able to process video, pictures, audio and other information, and able to act flexibly and fetch things, "It will enter all walks of life and become a real robot that comes out of science fiction." ”
The digital human anchor at the Zhongguancun Forum
So back to the present, when AI large models have not yet been widely used, the focus of the domestic artificial intelligence industry should be to focus on general large models or vertical large models? When the Red Star News reporter asked Li Kaifu, he replied firmly: "It has to be done, and it is all important." ”
Only when a hundred flowers bloom can we better empower thousands of industries in the future.
In Kai-Fu Lee's vision, the next stage, "we can expect a complete AI machine." "It's no longer just about generating content, it's about starting to do things for people. "For example, if you tell him that my wife's birthday is coming, he will no longer just answer whether you should buy lilies or roses, but can directly help you buy gifts. ”
AI will be everywhere
However, it is necessary to develop and govern well
At the opening ceremony of the 2024 Zhongguancun Forum, the prototype of the embodied AGI (General Artificial Intelligence) system launched by the Beijing Institute of General Artificial Intelligence was selected as a major scientific and technological achievement of the forum. The image of "Tongtong" is the appearance of a 3-4-year-old child, who has a unique "mind" that can understand words and grasp intentions, and behaves on a par with a human 3-4-year-old child. In the future, "Tongtong" will enter life, such as helping people serve tea and water, providing warm companionship, etc.
Ten years later, the iterated "Tongtong", or "Xiaoqi", will have more powerful wisdom and will penetrate more deeply into people's lives. The development of artificial intelligence has become unstoppable, so when discussing the future, there will be a topic of "how humans and machines can coexist".
In an interview, Gong Ke, executive director of the China New Generation Artificial Intelligence Development Strategy Research Institute, director of the Haihe Laboratory of Advanced Computing and Key Software (Information Innovation), and former president of Nankai University, expressed his views to the Red Star News reporter: "Artificial intelligence is not an independent subject, it is an extension tool to expand human brain power and labor ability, just as we do not need to discuss the coexistence of people and cars." ”
Gong Ke
Gong Ke believes that no matter how smart AI is, it is also a tool, and there is no such thing as "artificial intelligence will infringe on humans in the future". If software wants to become smart, it still depends on hardware, after all, it depends on scientific and technological breakthroughs, not on its own metabolism.
Of course, generative AI is getting smarter as technology evolves, but it's not infallible either. Gong Ke said that when artificial intelligence speculates on the answer, it uses probability, which is called joint distribution probability. Feed it a question that it determines which words have the highest probability of being combined with the input word, and that is the output it chooses. Since it is judged by probability, it is impossible to be 100% right, and there will always be times when it is wrong. Therefore, people must have the ability to examine facts, logic, ethics, and scientific principles of artificial intelligence, and understand what factors affect the distribution of probabilities. People need to be able to identify problems and correct them. In other words, "in order for man to control the tool he has developed, he must know the inherent weakness of the tool." ”
In Gong Ke's view, artificial intelligence can be classified and classified into risks. For example, brain-computer interfaces are implanted in the brain, which is a high-risk application. For example, you can age the audience of some apps to make it clear who uses it and where it is used. Therefore, the development of artificial intelligence is unstoppable, but it must be developed and governed well.
Red Star News reporter Wang Yao photo report
Edit into a preface
(Download Red Star News, there are prizes for reporting!)