zhu jiaming: various types of embodied intelligence will become the new protagonists of future economic activities
2024-09-09
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
on the afternoon of september 6, the paper research institute held a seminar on "digital china 2029: embracing the next generation of information technology". four guests, zhu jiaming, shi dan, peng wensheng, and zheng lei, gave keynote speeches. the picture shows the roundtable dialogue session.
today's theme "digital china 2029" is a very forward-looking topic, focusing on how to achieve the deep integration of the real economy and digital technology and intelligent technology, and what impact such integration will have on human economic and social life. therefore, this is a topic of the times.
what i will share with you today is "how artificial intelligence will affect and change the future economy", which is divided into four sub-questions: 1. artificial intelligence technology is in a period of continuously accelerating development; 2. artificial intelligence changes the basic model of the real economy; 3. the main ways in which artificial intelligence affects the future economy; 4. the risks and challenges of artificial intelligence in the process of creating the future economic form.
after 9 months, mr. zhu jiaming was a guest at the shangbao building again. this time, he gave a lecture on "how artificial intelligence changes and affects the future economy", bringing the most cutting-edge changes and overall change observations
artificial intelligence technology is in a period of continuous acceleration of development
it is now september 2024. the past nine months have undoubtedly been a period of accelerated development of artificial intelligence technology, with its technological iteration and innovation cycle shortened from "years" to "months" and even to "weeks". the "emergence" of innovative artificial intelligence technologies is basically one wave after another. this is a historical phenomenon that has never happened since the industrial revolution. at this moment, discussing "moore's law", which is updated every 18 to 24 months, is an outdated topic. in the 20 months since openai released chatgpt chat software on november 30, 2022, all major global technology companies have been wrestling with big models, running and competing on the new track of generative artificial intelligence, which has brought about earth-shaking changes in the artificial intelligence ecosystem. at present, the expansion of artificial intelligence technology is mainly concentrated in the following aspects:
* the emergence of autonomous agentic ai, ushering in the era of embodied intelligence "baby boom"
1. the scale of large models continues to expand.the "big" of the big model continues to rise. the big model "llama 3.1" developed by mate has more than 400 billion parameters. this almost astronomical revolution in parameters has gradually increased from billions, tens of billions, and hundreds of billions.
2. the influence of open source big models has expanded dramatically.big models are divided into open source and closed source. meta, which was facebook before october 2021, is becoming a leader in the open source of big models. their open source speed and scale are far beyond everyone's imagination. in the past 10 months, they have 350 million users due to open source. what does this number mean? everyone still remembers the shocking report in february 2023 that openai's monthly active users surged to 100 million in two months, making it the fastest growing consumer application in history.
3. the performance of hardware chips supporting artificial intelligence continues to break through.after the rise of machine learning and deep learning, nvidia has been leading the gpu (graphics processing unit) for 9 years and has taken the lead; google launched tpu (tensor processing unit) in 2017, and launched the sixth-generation tpu "trillium" in june this year; in 2024, npu (neural network processing unit) that simulates human neurons will be released. these hardware are constantly optimized in terms of memory, computing power, and energy consumption ratio, and are progressing in coordination with artificial intelligence.
the latest situation is that groq, founded in 2016, launched lpu, which has high-speed reasoning performance 10 to 100 times that of gpu and tpu, but with significantly reduced energy consumption. the emergence of this innovative enterprise is challenging the leading companies in the field of ai processors. groq successfully raised another 680 million in august and its current market value is 2.8 billion.
4. the emergence of autonomous agentic ai.on march 26, 2024, andrew ng of stanford university (the most authoritative international scholar in the field of artificial intelligence and machine learning) first mentioned it in a speech at the sequoia capital artificial intelligence summit. recently, he gave a related speech at the snowflake summit. agentic ai has become a hot topic among technical people. the biggest feature of this type of intelligent agent is that it is proactive, autonomous, and adaptable. it is a form of intelligence that does not need to be influenced or controlled by people. therefore, when planning the macro-scenario of artificial intelligence today, we will find that there are two types of intelligence, one is ai agent (ai agent) that serves people, and the other is agentic ai that directly represents the intelligent agent. the former serves people, and the latter is no longer a traditional artificial intelligence tool that serves people, or a narrow artificial intelligence tool.
5. the era of embodied intelligence “baby boom” has arrived.along with agentic ai, what is even more shocking is the emergence of embodied intelligence with a compound growth rate of 7%. what kind of big scenario will we face: suddenly, we have entered the "baby boom" period of embodied intelligence, just like the "baby boom" after world war ii. it is now developing at a growth rate of tens of thousands or hundreds of thousands. i believe that in three to five years, it will become the main body that affects the economy. therefore, in the next economic development, humans still have a say and occupy a dominant position, but the influence is no longer unique and absolute. the greater transformation is how people understand and adapt to various artificial intelligence entities. this is the cognitive premise that we must understand when discussing the impact of ai on the economy.
*it is expected to leapfrog chips and use biological storage, and it is expected to train large models on the moon
6. there is a third space: ai space.in february this year, the video software sora developed by open ai pushed the multimodality of large models to the extreme. fei-fei li from stanford university proposed that a third space has emerged - the ai space created by physical search engines.
7. the emergence of "dendritic colloids" for biological storage.at the end of august, a professor at north carolina state university in the united states published an article in nature magazine, which stated that a storage method called "dendritic colloid" has been discovered based on dna storage and computing engines. in the past, we believed that information could only be stored in chips constructed of silicon-based materials, but the current "dendritic fiber" storage can store dna files more stably and for a longer period of time. at 4 degrees celsius and -18 degrees celsius, the half-life is about 6,000 years and 2 million years. this system includes multiple image files encoded into dna adsorbed on colloidal substrate particles, which can erase and replace data to achieve lossless file access.
8. large space training models can or solve energy needs.training large models requires a lot of resources, including computing power, which has always challenged the energy limit. at present, the us company lumen orbit has planned to launch its first satellite to the moon next year, and then launch a larger iteration every year until the total server power reaches the gigawatt scale, so that the operation of the artificial intelligence computing power cluster will no longer be constrained by the energy supply cost and heat dissipation on the ground, and large models can be trained on the moon. according to the current training trend of large models, in 2027, we need gw power cluster support. for example, the power consumption of the gpt-6 model directly exceeds the energy supply of the largest power plant in the united states. and lumen orbit is not the only company committed to sending data into space orbit. the eu-funded axiom space has planned to build an orbital data center on axiom's space capsule and launch it in 2026-2027. many companies are exploring the operation of space computing power clusters.
therefore, artificial intelligence is constantly changing the world on a weekly basis, making the world's technological changes more profound and drastic.
artificial intelligence changes the basic model of real economy
under this premise, artificial intelligence changes the real economy mainly through two modes.
*new companies emerge from ai technology breakthroughs, which is innovation from 0 to 1
the first model: ai development itself has industrial genes, and each step of its advancement naturally leads to the emergence and development of new industries. that is, ai directly creates new industries and sectors, forming new economic sectors and new enterprises. the second model is that ai transforms traditional industries, sectors and sectors.
there are many cases of the first type of model. openai is a new company related to large models. it is itself a new industry. it has the basic characteristics of the industry - the industry's technology, the industry's direction, the industry's business model, and typical products that are extremely innovative. nvidia's h100 gpu is the main tool for training ai models and costs about $30,000. training a large ai model with 1.8 trillion parameters requires about 2,000 nvidia blackwell gpus. some institutions estimate that the global generative artificial intelligence market will be worth $13.71 billion in 2023. by 2025, the generative artificial intelligence market valuation is expected to reach $22.12 billion. the market is expected to grow at a compound annual growth rate (cagr) of 27.02%. in 2022, the adoption rate of large-scale generative artificial intelligence was 23%. it is expected that by 2025, the adoption rate will reach 46%.
another example is the avant-garde "dendritic colloid" storage, which combines artificial intelligence chip hardware with biology. this field is infinitely broad. the so-called innovation is most important to go from 0 to 1, and artificial intelligence is the most important driving force for this kind of innovation in the world today.
*ai transforms enterprises: computing power and algorithms will be improved, and productivity will increase
the second type of model, ai, is also used in many cases to transform old industries. for example, in the gaming industry, the popular game "black myth: wukong" in the past month or two has used a lot of artificial intelligence technology. traditional labor-intensive, technology-intensive, and capital-intensive industries must be changed to intelligent-intensive. this is a fundamental change of great importance. for example, siemens has implemented ai-driven predictive maintenance in its factories, which has significantly reduced unplanned downtime. by predicting potential failures, they can solve problems before they interrupt production, thereby improving overall efficiency. fanuc, a leading industrial robot company, uses ai to enable its robots to learn from their own experience, improve their performance over time, and contribute to more efficient production processes. google's deepmind ai has been used to reduce energy consumption in data centers by optimizing cooling systems. similar ai-driven energy management systems are now being applied to manufacturing to achieve more sustainable and cost-effective production processes.
no industry can reject artificial intelligence, because the greatest contribution of artificial intelligence is to ensure that the labor productivity of traditional industries can be improved, and to ensure that the computing power and algorithms of any industry, enterprise and economic organization can be improved. this is a big trend, and this trend has just begun. artificial intelligence, especially embodied intelligence, will have a comprehensive impact and penetrate into the scope of life, affecting everyone's daily life, and it will become more and more intense and drastic.
if there is a third model, it is the integrated development of the first and second models, combining new artificial intelligence technology with traditional transformation product industries.
in general, artificial intelligence will shape an unprecedented economic ecology. so far, it has not been deduced by theory, but is entirely based on empirical summary.
the main ways in which artificial intelligence will affect the future economy
specifically speaking about how artificial intelligence transforms the real economy, if we make a simple summary based on the current actual situation, it mainly focuses on the following aspects.
*people and various intelligent entities become the main body of economic activities
1. artificial intelligence has changed the subject of economic activities. natural persons used to be the subject of economic activities, but now they are no longer the only subject. various intelligent entities, including robots, have become the subject of economic activities together with humans. this is a very obvious change, and this change will develop further. in ten or even thirty years, we will see a significant increase in the proportion of embodied intelligence. we don't know when a fundamental turning point will occur. but i believe it will only be 20 to 30 years.
2. changes in economic entities lead to changes in people themselves, and humans are becoming more and more intelligent. humans are surrounded by a variety of artificial intelligence tools, various wearable devices, and various brain interfaces, and people are no longer the same as they used to be. the development of artificial intelligence will change people themselves: the combination of artificial intelligence and genetic engineering, brain interfaces, and extended reality technology will help humans experience the virtual world, or simulate the digital world, and change people's cognition and experience of the world.
3. artificial intelligence has changed the space and time of human activities. because artificial intelligence is not limited by the so-called traditional geographical space, nor is it limited by newton's laws in time. we have seen musk's mars immigration plan, many of which are beyond the reach of humans, and the technology relied on to achieve such a goal is artificial intelligence. because of artificial intelligence, humans will be able to explore the universe more easily. humans can also create the physical world and create forms that simulate the physical world through various means of artificial intelligence. because of artificial intelligence, the space of human activities will surpass the traditional spatial state based on geography so far. similarly, humans, or the subject of the intelligent era, will also break the traditional time model. humans will enter a new era in which space and time can be expanded and intersected.
*the economic cycle is fuzzy and depends on the structural, systematic and clustered innovation of ai
4. artificial intelligence has changed the economic cycle. in the industrial age, economic activities have obvious business cycles. the medium cycle has a greater impact on economic operations, with an average length of 8 years. however, in the transition period towards the information economy and the intelligent economy, the short cycle will become blurred and the medium cycle will become disordered. the economic cycle is increasingly determined by the structural, systematic and clustered innovations of the artificial intelligence technology revolution. simply put, the world entering the 2020s is likely to rely on artificial intelligence innovation to achieve a long cycle of prosperity.
5. artificial intelligence has changed the form of products. in the future, products that can be sold well will not only have high technical content, but also high intelligence content. human beings will enter an era where they need to continuously produce intelligent products. all economic activities today need to be based on intelligent input and intelligent output as basic characteristics.
6. artificial intelligence has changed the economic distribution system. the distribution method will determine the relationship between production and consumption. in the era of artificial intelligence, natural persons, that is, traditional humans, must significantly reduce their working hours and work intensity. the basic characteristics of the era of artificial intelligence are that the government and the state guarantee the basic income of the people. it is very likely that humans will enter a historical period of redefining labor, labor force and distribution. with the guarantee of basic income for the people, they can have a wider range of work or employment options. this change completely subverts the traditional employment system since the industrial age.
*changes in economic structure and the relationship between production and consumption have led to changes in national strength
7. artificial intelligence will reshape and adjust the infrastructure of economic development. the traditional economic infrastructure is mainly energy and transportation, and later computing power and algorithms. now it is not enough. now, artificial intelligence, especially the big model of artificial intelligence, will be incorporated into the infrastructure of human economic activities. it will change all aspects of the infrastructure: energy will become smart energy; transportation will become smart transportation; cloud computing will become intelligent computing; computing power will become intelligent computing power. this is a revolutionary change - the era of infrastructure without artificial intelligence has completely ended.
8. artificial intelligence has changed the relationship between production and consumption. in traditional economic activities, production is production and consumption is consumption; in the economic activities of the supply chain and value chain in the intelligent era, a lot of production is consumption, and a lot of consumption is production. the activities of generative artificial intelligence are proof of the dual functions of production and consumption.
9. artificial intelligence will change the world and the relationship between countries. traditionally, the dual economy mainly refers to agricultural countries and industrial countries, developed market economies and emerging market economies, industrially developed countries and industrially underdeveloped countries. from now on, the development level of countries in the world depends to a large extent on the development of the artificial intelligence industry and the artificial intelligence economy. in this wave of artificial intelligence, too many countries and too many people have no way and no possibility to keep up with such a wave due to lack of talent, technology, and capital.
risks and challenges in creating the future economy
human loss of control, corporate monopoly, class restructuring, and global governance are easier said than done
any new technology that brings convenience is inevitably accompanied by risks and challenges, and the same is true for current generative artificial intelligence and big models.
first, ai faces the danger of awakening its own consciousness and losing control of it. this is reflected in the fact that humans do not fully understand the internal operating mechanism of the big model. for example, more people believe that artificial intelligence has self-conscious reasoning and even psychological activities, but how to prove it is a new challenge. to date, the input and output processes of the big model are still "black boxes"; the agi (general artificial intelligence) stage is accelerating, and the "dawn" of asi (super artificial intelligence) has appeared. therefore, there is a challenge of how to adapt to the "co-intelligence" of human intelligence and artificial intelligence.
secondly, large ai companies strengthen their natural monopoly on ai development. if these companies fail to represent the public interest of humanity, they will directly or indirectly affect the ideological, economic, political and social fields of the entire society through their monopoly on ai technology.
third, artificial intelligence will indirectly disrupt the existing balance of the existing social structure, forming an artificial intelligence divide similar to the digital divide, and thus triggering the possibility of reconstructing social classes.
fourth, the international community's cooperation in ai governance may fail. at the "future summit" at the 75th anniversary of the founding of the united nations to be held on september 22, the united nations will submit the "global digital compact". this compact has been proposed since april and has been revised to the third version. governments are currently negotiating and consulting on it. to this end, the united nations has passed two resolutions, one is a resolution led by china on strengthening ai capacity building and conducting international cooperation, and the other is a resolution proposed by the united states on promoting safe, reliable and inclusive ai systems. however, due to different geopolitical and national governance positions, this disagreement has led to an impact on the global governance process. at the same time, since only a few countries can participate in ai governance, and most countries do not have the ability to participate, there is also an objective gap between northern and southern countries and between developed and developing countries. therefore, it will affect the global balanced development of ai.
in summary, this wave of artificial intelligence is a truly subversive historical trend. its impact on the economy is both microscopic and macroscopic; both immediate and long-term. its impact on enterprises, families, countries, and the entire human civilization is not only to promote transformation, but also to fundamentally change. its impact can be compared to that of the invention of electricity and the energy revolution, because we have entered and are entering artificial intelligence. it is not just that artificial intelligence helps humans, but the development of artificial intelligence will continue to push humans to adapt to artificial intelligence. such a moment is just around the corner.
i hope everyone will imagine artificial intelligence and the profound changes it will bring to economic life and social structure with an extraordinary mentality.
compiled by: li nian
this is the keynote speech delivered by zhu jiaming (economist, chairman of the academic and technical committee of digital finance research institute of hengqin guangdong-macao deep cooperation zone) at the "digital china 2029: embracing the next generation of information technology" seminar held by the paper research institute on the afternoon of september 6.
author: zhu jiaming
text: zhu jiaming photo: li nian editor: li nian responsible editor: li nian
please indicate the source when reprinting this article.
