Qingdao Daily, March 27, 2024, page 1
The AI Revolution: It's Just Beginning
"The rise of AI text-generated image technology and the advent of models such as ChatGPT have fueled enthusiasm in the field of generative AI. At the same time, AI models are shifting from a single mode to a multi-modality, which can understand and connect data from different modalities to achieve text-to-image, image-to-video, text-to-audio and other transformations. ”
The above description of the development trend of artificial intelligence does not come from an expert, but is generated by Baidu's large model Wenxin Yiyan.
Entering the spring of 2024, the popularity of large models continues. OpenAI has released Sora, a large video model for Wensheng, which can generate a 60-second video as long as you enter some descriptive prompts. The V3 version of the Suno model released by Suno can set the music style and content theme according to your preference, and generate 2 minutes of music in a few seconds. The KimiChat large model supports 2 million words of long text input, and it has set off a large model long text competition......
With the combination of large models and more practical application scenarios, artificial intelligence has quietly penetrated into all aspects of production and life while rapidly iterative and evolving, and has gradually become an indispensable part.
No one can doubt that we are going through a profound transformation. But at a time when eagerly embracing AI, it may be useful to heed Bill Gates' reminder that "we always overestimate what will change in the next two years and underestimate what will happen in the next decade." ”
It's just getting started. Find the right track outlet and embark on the train of the times, in this moment of change, artificial intelligence opportunities and challenges coexist.
More than just Wensheng videos
What can AI do?
In the past year, large models have emerged in various industries, and artificial intelligence has evolved faster than expected, but the vast majority of projects have focused on text, image, and video generation. This has also created a stereotype, as if AI is only good at writing poems and paintings.
In fact, the real point of artificial intelligence transformation has never been these, but the implementation of more real industries: changing the operation mode of many fields such as biomedicine, industrial manufacturing, financial technology, smart mining, and scientific research.
But as with the proliferation of all new technologies, the evolution of AI cannot be carried out at a uniform pace. At first, the application of ChatGPT attracted social attention, but AI technology must reach a new inflection point for more industrial applications, which often requires bringing together several different developments and the ability to integrate them to make the overall product experience more engaging than before.
Of course, there are good reasons to believe that AI is approaching a leap forward in many application scenarios.
Not long ago, KoBold, a unicorn company invested by Bill Gates, announced the discovery of giant copper reserves in Zambia. It took just over a year for KoBold to go from investment to discovery, and this efficiency lies in the use of artificial intelligence to process drilling data to optimize copper exploration. Traditional prospecting relies on extensive field operations, geophysical exploration, and geological analysis, which can take several years or more. Artificial intelligence, on the other hand, is not affected by luck, and through big data analysis and deep learning, it can accurately find precious mineral veins hidden underground.
In China, artificial intelligence is also being deeply integrated with manufacturing, biomedicine, energy, transportation and other industries.
Kaos COSMOPlat has independently developed the Tianzhi industrial model, which covers intelligent question answering, control code generation, database query, and decision-making assistance, HUAWEI CLOUD Pangu drug molecule model helps pharmaceutical companies open a new model of AI-assisted drug research and development, greatly improving the efficiency of new drug research and development, and Shanghai Artificial Intelligence Laboratory has jointly released an upgraded version of the artificial intelligence weather forecasting model "Fengwu", which uses artificial intelligence to model and forecast medium-term weather at the 10 km level......
Taking the Tianzhi industrial model as an example, artificial intelligence converts engineers' industrial experience and industrial knowledge into quantifiable data and indicators, which can be used in the form of robots, applets, apps, etc., which can form feedback on related problems within seconds, and realize the inclusive application of industrial knowledge.
Of course, all changes will not happen overnight, and the evolution and diffusion of the entire artificial intelligence technology will undergo continuous wave-like changes in different sectors and industries, which is likely to be an important feature of technological and social changes in the next two or three decades.
When change comes, there is no chance to be lost, but people who are down-to-earth will also find that opportunities are always there.
Large models are not the whole story
Big models aren't all there is to AI.
If it wasn't the "big manufacturers" who made phenomenal products like ChatGPT, others could lie flat. Opportunities in the AI industry are everywhere, with applications at the top and underlying infrastructure at the bottom.
At present, Chinese artificial intelligence enterprises have successively launched more than 300 large models, including technology giants such as Huawei, Baidu, Alibaba, and iFLYTEK, as well as a considerable number of small and medium-sized startups. Behind the "100 model war", there are extremely high R&D costs and technical barriers, a training can cost millions of dollars, and chips are "hard to find". Even OpenAI, which is leading the industry, has not solved the commercialization dilemma and needs frequent "blood transfusions" from the giant Microsoft.
In the crowded artificial intelligence track, it seems that everyone has to "roll" to the end on the road of general large models. But in fact, the artificial intelligence industry is a vast blue ocean, as long as you choose the appropriate industrial division of labor roles, and explore more subdivided scenarios such as office, education, e-commerce, and medical care, you can form a differentiated competitive advantage. Baidu founder Robin Li's view is representative: "Don't roll the model, roll the application, only the application directly creates value." ”
Mobile phones are a hot field for the current large-scale model landing. Artificial intelligence can provide a better experience in mobile phone photography, video production, intelligent search, document writing, etc., The mobile phone industry with slowing sales will take "AI" as the starting point, and Apple, which has given up making cars, will concentrate resources on generative artificial intelligence is an important signal, Samsung, Xiaomi, Honor, and OPPO are all focusing on the layout of artificial intelligence.
Another hot area is embodied intelligence, which is intelligent robots with physical entities that can interact with the environment in real time through sensors and actuators, execute commands, or make decisions. At present, the robot trained by the large model has been able to skillfully complete simple tasks such as grabbing eggs, folding shirts, and watering flowers, and the all-round housework robot in the future has come into reality.
No matter what kind of large model, its computing and storage capabilities rely on chips, sensors, and computing infrastructure, and are inseparable from the support of electricity. Hardware innovation, new energy and other fields closely related to artificial intelligence are also expected to usher in an explosive period.
Relevant data shows that the demand for computing power is growing at a rate of more than 10 times per year, and more hyperscale data centers are needed to match AI application scenarios. By 2025, the share of AI business in global data center electricity consumption will increase from 2% to 10%.
Deeper changes are on the way
Looking back at the long history of history, every scientific and technological revolution has had a huge impact on the course of human history. At present, emerging technologies represented by artificial intelligence are reshaping human society, and a series of challenges such as traditional job replacement, data privacy and security, intellectual property protection, and artificial intelligence ethics are roaring.
Many people's perception of artificial intelligence may still be stuck in the anxious confrontation between individuals and the industry. Instead of worrying about AI, think of AI as a capability multiplier.
In the medical field, artificial intelligence is used for medical imaging diagnosis, personalized treatment, patient monitoring, etc., in the transportation field, autonomous driving can improve traffic efficiency and reduce traffic congestion, in the industrial field, artificial intelligence can drive automated production, predictive maintenance, quality inspection, etc., and in the field of education, artificial intelligence can carry out personalized teaching, online learning, etc...... Artificial intelligence is changing thousands of industries, but it is also changing the work itself.
It should be noted that at this stage, artificial intelligence is still in the early stage of development and cannot completely replace human work in every field. For a long time, artificial intelligence will only play a supporting role, making people's work more efficient and greatly shortening the production cycle of products.
Taking AI mapping as an example, the expressiveness of drawing largely depends on the accuracy of the imported model and the ability of the operator to give instructions, and AI mapping has become an essential skill for practitioners, and the demand for related training is increasing. Artificial intelligence has also sprouted new professions, and large models are "fed" with massive data, and artificial intelligence trainers who screen, annotate, and correct errors on relevant text data have emerged.
As Kai-Fu Lee, Chairman of Sinovation Works, said, "AI models are a historical opportunity that must not be missed. Because it's going to be the biggest platform revolution ever, it's going to be 10 times bigger than Windows or Android, it's going to rewrite every application, it's going to refactor human work, it's going to amplify the ingenuity of creative people 10 times or more. ”
The artificial intelligence industry is approaching a new critical point, and we must firmly seize the strategic opportunity of industrial development and strive to win the future in the great changes. (Qingdao Daily/Guanhai News reporter Zhou Xiaofeng)