Gree Advertising 2021 (Artificial Intelligence, Gentle Entry into the Economy, Gree "Replace&qu

 

1. Artificial Intelligence Image Robot

Image source @ Visual Chinese | Lai Ka Think Tank, Compiler | Dou Dou, Editor | Gong Yan. Some economists jokingly point out that the global benefits obtained through AI may reach infinity, but with reference to the laws of human socio-economic development, the impact of new technologies is even longer tailed and more moderate. For example, even though many people believe that the industrial revolution in the late 18th century was the result of Jenny's invention of the textile machine, In fact, it is driven by multiple factors: the growth of coal use, stronger property rights protection, the rise of scientific spirit, and so on.

2. Artificial intelligence image generation

The era of generative artificial intelligence has truly arrived in November last year, with OpenAI chatbots using the Large Language Model becoming the pioneers of the era. Now, amazing progress is happening almost every day.

3. High definition artificial intelligence image materials

Recently, the songs of fake Drake and The Weekend stars produced by AI have shaken the music industry and programs that can convert text into videos are making convincing progress. Soon, consumer products such as Expedia, Instaart, and OpenTable will also be connected to OpenAI's robots, helping people order meals or book vacations by typing text in boxes.

4. Artificial intelligence image creativity

A recently leaked presentation by a Google engineer shows that Google is worried that there will be surprises after competitors can easily make progress. The development of AI has had a profound impact, but the most urgent and intuitive question is, what does this mean for the economy? Many people have great expectations for this.

5. Artificial intelligence image materials

AI Potential "Range": Monopoly, Labor Market, Productivity Goldman Sachs Bank believes in a recent research report that "the widespread adoption of AI will lead to a global average GDP growth of 7% within a decade, approximately $7 trillion." Academic research has pointed out that companies adopting this technology will have an average annual labor productivity growth rate of 3%, which will bring a significant increase in company profits for many years.

6. Artificial intelligence image high-definition

A study published by Tom Davidson of Open Philanthropy, a donor organization, in 2021 suggests that the probability of "explosive growth" (i.e. global output growth exceeding 30% annually) in this century is 10%. Some economists jokingly point out that the benefits gained globally through AI may reach infinity.

7. Artificial Intelligence Picture Cartoon

The financial market has shown a more conservative response, although it has risen in the past few months. Companies involved in AI have underperformed the global average in stock prices over the past year, and interest rates have also given the same signal. If people believe that technology will make everyone richer in the future, the need for savings will decrease, and interest rates should rise. Research by Massachusetts Institute of Technology Basil Halperin suggests that, The interest rate adjusted for inflation is highly correlated with subsequent GDP growth.

One of the methods for artificial intelligence images is to use object detection

However, since the start of the AI boom in November last year, long-term interest rates have decreased, and by historical standards, they are also very low. These researchers speculate that in the 30-50 year forecast period, the financial market has not formed high expectations for AI induced accelerated growth. In order to determine which group's judgment is correct, it is necessary to consider the history of technological breakthroughs in the past, which will be very helpful for investors.

One of the methods for artificial intelligence film reading is to use object detection

From a historical perspective, it is difficult to prove that a single new technology itself had a significant impact on the economy, whether this impact was good or bad. Even though many people believe that the industrial revolution of the late 18th century was the result of Jenny's invention of the textile machine, it was actually driven by multiple factors: the growth of coal use, stronger property rights protection, the rise of scientific spirit, and so on.

The process of artificial intelligence film reading is reflected in

Perhaps the most famous of them is Robert Fogel's work on American railways published in the 1960s, which later earned him the Nobel Prize in Economics. Many people believe that railways transformed the prospects of the United States, transforming agricultural society into an industrial powerhouse. However, Fogel found that the influence of railways was limited as they replaced potentially equally efficient canal shipping.

If railways had never been invented, then the per capita income level in the United States as of January 1, 1890 could also have been achieved on March 31 of the same year. Of course, for technologies like AI that are basically unpredictable, in which aspects they will replace humans, no one can accurately predict rapid growth or technological stagnation.

But you can still imagine multiple possibilities. At least from the current perspective, Fogel's railway theory can serve as a useful blueprint to consider the following three major factors: monopolies, labor markets, and productivity. New technologies can sometimes create a small group of people with vast economic energy. Rockefeller wins with oil refining, and Ford wins with cars.

Today, Jeff Bezos and Mark Zuckerberg's dominance is also attributed to technology. Many believe that the AI industry will generate huge profits in the near future. Goldman Sachs' recent research report predicts that in the most ideal scenario, generative AI will help global enterprise software revenue increase by $430 billion annually.

In their calculations, they assumed that everyone with 1.1 billion white-collar workers worldwide would use AI tools with an annual fee of $400, and any company would be happy to generate these revenues. However, at the macroeconomic level, $430 billion does not change the overall situation. It is unrealistic to assume that these revenues can be fully converted into profits, and that these profits are all obtained by the United States (let alone unrealistic).

Even under such assumptions, the pre tax profit ratio of US companies to GDP has only increased from 12% now to 14%. Although much higher than the long-term average, it has not yet reached the level of the second quarter of 2021. These profits may be earned by an institution, perhaps because OpenAI monopolies often occur in industries with high fixed costs or when business is difficult to transfer to competitors.

For example, customers cannot find a substitute for Rockefeller oil and cannot refine their own generative AI. There are also some monopolistic features of OpenAI chat robots, and the training cost of GPT-4 is rumored to exceed $100 million. Few companies have such strength. The model also needs to use acquired knowledge as training data and user feedback.

However, the likelihood of a single company controlling the entire AI industry is very low. It is more likely that, like industries such as aviation, retail, and search engines, a few large companies compete with each other due to the use of similar models, and no AI product is unique. This makes it easier for customers to switch from one model to another, and the computational power behind the model is also relatively universal.

Many codes, tips, and tricks are available for free online, which means that amateurs can also produce their own models, and the results are often surprisingly good. A team from venture capital firm Andreessen Horowitz believes that "generative AI does not currently form a systematic moat.

”The recently leaked report from Google has also formed a similar conclusion: "The entry threshold for training and experimentation has been lowered from the full strength of a large research institution to one person, one night, and one high-performance laptop." There are now several generative AI companies with valuations exceeding $1 billion, The biggest winners in the new AI era are not even AI companies: Nvidia, the computing company that drives AI models, is experiencing a surge in data center revenue.

So what does this mean to me? Although generative AI may not create a new class of bandit aristocracy, it does not provide much comfort for many people. People are more concerned about the impact of AI on their own economic prospects, especially whether their jobs will disappear. Terrible predictions can be seen everywhere, according to OpenAI's Tyna Eloundou and colleagues.

About 80% of the workforce in the United States is affected by the big language model, at least 10% of their jobs, "said Edward Felten of Princeton University, who conducted a similar analysis and believed that the professions most likely to be disrupted include legal services, accounting, and travel agencies. Economists have previously issued pessimistic predictions.

In the 2000s, many people were afraid of the impact of outsourcing on workers in the developed world. In 2013, two people from Oxford University published a widely cited paper that believed that automation would erase 47% of jobs in the United States over the next 10 years. Some also believed that even without widespread unemployment, there would be a problem of job hollowing out, where beneficial, high paying jobs would be replaced by brainless, low paying jobs.

The final result surprised many people. In the past decade, the average unemployment rate in the developed world has almost halved. The share of the working age population in employment is the highest in history, with countries such as Japan, Singapore, and South Korea having the highest automation and robotics rates. The unemployment rate is also the lowest. Recent research by the US Bureau of Labor Statistics shows that jobs that have been threatened by new technologies in recent years, There is no overall trend of rapid loss.

The evidence of "hollowing out" is also unclear. Job satisfaction indicators have increased in the 2010s. For most of the past decade, the poorest Americans have seen higher wage growth than the wealthiest. This time, it may be different. After acknowledging that ChatGPT "has an impact on its new customer growth," the stock price of Chegg, which provides homework assistance, has recently halved.

The CEO of IBM, a major technology company, has announced that the company will stop recruiting for positions that may be replaced by AI in the coming years. Are these early signs really leading to an economic tsunami? Imagine that after AI automates more than 50% of a job's tasks, the job disappears, or imagine workers being fired based on the proportion of task automation in the overall economy.

Regardless of the above situation, according to Ms. Eloundou's estimation, this will cause 15% of the jobs lost in the United States. Some people may be able to move to industries where there is a shortage of workers. For example, the unemployment rate in the nursing industry will certainly rise sharply, and may reach the level of 15% in 2020 when the COVID-19 epidemic is the most serious in the United States.

However, this scenario is unlikely to become a reality: historical experience shows that the disappearance of actual work is much slower. The automated telephone switching system that can replace manual operators was invented in 1892. But it was not until 1921 that Bell System installed the first fully automated office. Even after this milestone, the number of telephone operators in the United States continued to grow, reaching the peak of about 350000 people until the middle of the 20th century.

The industry did not completely disappear until the 1980s, 90 years after the invention of automation. AI swept through the labor market and it will not take 90 years: big data models are easy to use, and many experts are amazed at the speed at which the general public has integrated ChatGPT into their lives. However, the reason why technology was difficult to quickly enter the workplace in the past still exists this time.

Mark Andreessen of Andreessen Horowitz summarized the relevant reasons in a recent article, and his argument mainly focuses on regulatory oversight in economic fields with a large amount of government involvement, such as education and healthcare. Technological changes are generally quite slow, and the lack of competitive pressure leads to insufficient motivation for improvement.

The government may also have public policy goals such as maximizing employment levels, which goes against improving efficiency. These industries are more likely to have established trade unions, which are good at preventing job losses. There are many examples of this, although technology that can be partially or fully replaced has existed for decades, the driver income of London's publicly operated subway network is still close to twice the median national income.

Government agencies have repeatedly asked you to fill out forms providing personal information. In San Francisco, the center of global AI's rise, during traffic congestion, you still need to hire live police officers to command traffic work. Goodbye! Many jobs threatened by AI are in highly regulated departments before returning to Mr. Felten's paper at Princeton University.

Among the 20 industries most affected by AI, 14 are all in the teaching industry (foreign language teachers are the most affected, while geography teachers are relatively safe). However, only the bravest governments dare to use AI to replace teachers. Otherwise, imagine what news headlines would say about the relationship between police and anti crime AI. Italy has temporarily banned ChatGPT due to privacy concerns, and France, Germany, and Ireland are reportedly considering doing the same, Show how concerned the government is about the impact of AI on job destruction.

Perhaps in a while, the government will allow some jobs to be replaced, but this delay will give the economy space to do what it has been doing: while some jobs are eliminated, creating new types of jobs by reducing production costs, new technologies will create more demand for goods and services, promoting employment in areas that are difficult to automate.

A paper published by David Autor of the Massachusetts Institute of Technology in 2020 gave an astonishing conclusion: 60% of job types in the United States did not yet exist before 1940, and the "manicurist" profession was added to the statistics in 2000, while the "solar photovoltaic electrician" profession was only added five years ago.

The AI economy may create new professions that are currently unimaginable. The mild labor market impact may bring about a mild impact on productivity, which is the third factor. In the late 19th century, electricity was gradually introduced into American factories and households. However, productivity did not improve until the end of World War I.

The personal computer was invented in the 1970s, and the productivity improvement was faster, but people's feelings at that time were still slower. In 1987, economist Roert Solow famously stated that the computer age "has arrived in all aspects, except in the field of productivity statistics". The world is still waiting for the productivity improvement brought about by recent innovation.

Smartphones have been widely used for ten years, and billions of people have access to the ultra fast internet. Many workers freely switch between their homes and offices. Official statistics show that over one tenth of American employees are already working in companies that use certain AI. If you look at unofficial statistics, the rate may be even higher.

AI may eventually significantly improve the productivity of some industries Erik Brynjolfsson of Stanford University has studied customer support representatives in his paper. Access to AI tools has increased the number of problems that can be solved every hour by an average of 14%. Researchers themselves can also become more efficient: GPT-X is equivalent to providing countless almost free research assistant.

Others hope that AI can eliminate inefficient management in the field of healthcare, reduce costs, and even reduce productivity. However, there are many things that AI cannot solve, such as blue-collar jobs in construction and agriculture, which account for about 20% of GDP in developed countries. Big data models are almost useless for people picking asparagus.

Even though AI may be useful for plumbers to repair broken faucets (such as using AI tools to identify faucets, diagnose problems, and provide repair suggestions), manual work ultimately has to be done by plumbers. Therefore, in the coming years, it is difficult to imagine that blue collar work productivity will greatly improve compared to today. This is also true in industries such as reception and medical care, as interpersonal contact is an integral part of their services.

AI also cannot solve the biggest problem that troubles productivity growth in the developed world: a stagnant planning system. When the size of cities is limited and housing costs are high, people cannot work and live in the most efficient places. No matter how many outstanding ideas you have in your society, if there is no way to implement them in a timely manner or if they are useless, the government should solve the related NIMBY (Not in My Backyard) problem. Technology itself is not the key.

The same is true for the energy industry, where licensing and infrastructure are the reasons for high costs. The AI economy may even reduce productivity. Look at recent technologies, smartphones can achieve real-time communication, but they also have strong interference. You receive emails 24/7 every week, making it difficult to concentrate. Researchers from the University of California, Microsoft, and Massachusetts Institute of Technology published a paper in 2016, The longer you spend on email every day, the lower your potential productivity.

”Working from home was once considered to improve productivity, but now some bosses believe it will benefit many people供偷懒的借口生成式AI自身也可能拖累生产力例如，如果AI能够完美配合你的各方面需求定制娱乐将会怎样？此外，对于拥有能够即刻生成大量文本的系统意味着什么，多数人可能都没有想清楚。

GPT-4对于应对规划应用的邻避人士来讲简直是上帝的礼物，只需要五分钟，他就可以生成一封精致的1000页纸的反对意见书接下来，就需要专门有人来应对这封意见书垃圾邮件也将变得更难监测欺诈案件数量将飙升银行将需要在防范攻击方面投入更多，并补偿那些发生损失的人。

在充斥着AI的世界里，律师数量将成倍增长Brown Rudnick律所的Preston Byrne说：“1970年代，你用15页纸就能够完成数百万美元的交易，因为重新打字实在是太痛苦了”“应用AI将让我们在第一稿中就覆盖1000个最有可能发生的极端案例，然后各方将就此争论几个星期。

”在美国的经验法则是，如果你预期的补偿金额不到25万美元，那么起诉就没有意义，因为上庭就需要花这么多钱现在，诉讼的成本可能将降到约等于零此外，教师和编辑将需要确认读到的东西不是由AI创作的OpenAI发布了一个程序可以帮他们做这件事，为其自身科技带来的问题给世界提供了一个解决方案。

AI可能会以人们今天难以想象的方式改变世界，但这并不意味着会使经济天翻地覆Fogel写道，他的论证“并不是在驳斥铁路在19世纪美国发展中扮演决定性角色的观点，而是在证明该观点所基于的现实基础完全不像人们通常认为的那样牢固”。

在21世纪中叶的某个时间，可能也会有一名研究生成式AI的诺贝尔奖得主得出同样的结论