With the appearance of ChatGPT, the world has entered the era of artificial intelligence.

Nowadays, the world is in the era of great acceleration of artificial intelligence, and various fields are constantly exploring and applying artificial intelligence technology, which also promotes the accelerated development of artificial intelligence. The following are some factors that accelerate the development of artificial intelligence:
  1. Explosive growth of data: With the development of the Internet, the cost of data generation and storage is getting lower and lower. At the same time, the popularity of various sensors, devices, Internet of Things and other technologies also makes a lot of data collected and stored, which provides more data bases for the training and application of artificial intelligence technology.

  2. Progress of hardware technology: With the development of computer hardware, especially the emergence of accelerators such as GPU and TPU, the training speed and effect of artificial intelligence have been greatly improved.

  3. Continuous optimization of algorithms: In the fields of deep learning and reinforcement learning, researchers constantly put forward new algorithms and optimization methods, making the application effect of artificial intelligence more outstanding.

  4. A large influx of investment: Artificial intelligence technology has broad application prospects, attracting a large number of investors and capital influx, which provides more financial support for the research and application of artificial intelligence.

With the continuous development and application of artificial intelligence technology, its influence will be more and more profound, changing our way of life and work.

ChatGPT, as a large-scale language model, represents the latest progress in natural language processing in the field of artificial intelligence. It is obtained through deep learning algorithm and a large number of corpus training, and can automatically answer users’ questions, generate articles, translate languages and other tasks.

The appearance of ChatGPT indicates that the field of artificial intelligence has entered a new era, that is, the era of large-scale pre-training model. Prior to this, the research in the field of natural language processing mainly relied on manually labeled data sets, and this method often required a lot of manpower, time and money. Now, using the large-scale pre-training model, we can automatically learn the laws of natural language through a large-scale corpus without manual annotation, which greatly improves the efficiency and accuracy.

In addition to the field of natural language processing, the large-scale pre-training model also shows great capabilities in computer vision, speech recognition, recommendation system and other fields. They can not only greatly improve the accuracy of the model, but also provide more possibilities for the application of artificial intelligence.

  1. Language understanding and generation: ChatGPT can be used to understand and generate languages, including natural language processing, text summarization, machine translation, question and answer system and so on.

  2. Personalized recommendation: ChatGPT can provide personalized product or service recommendations by analyzing users’ historical behaviors and preferences.

  3. Automated customer service: ChatGPT can be used to build an automated customer service system to provide customers with 24-hour uninterrupted support and service.

  4. Intelligent writing: ChatGPT can help people to quickly generate various texts, including press releases, advertising copy, product descriptions, articles and so on.

  5. Sentiment analysis: ChatGPT can identify and analyze people’s emotions, thus helping enterprises to understand users’ emotional tendencies and attitudes, so as to better improve the user experience.

  6. Self-driving: ChatGPT can help self-driving cars understand and respond to road conditions, traffic signs and other vehicle behaviors.

  7. Financial field: ChatGPT can be used for tasks in financial field such as risk assessment, anti-fraud and credit evaluation.

  8. Medical field: ChatGPT can be used for medical image analysis, medical record, drug research and development and other tasks in the medical field.

The application field of ChatGPT is constantly expanding and deepening, and many new application scenarios will emerge in the future.
Therefore, the appearance of ChatGPT marks the arrival of the era of great acceleration in the field of artificial intelligence. It is expected that in the next few years, the large-scale pre-training model will be widely used and popularized in various application scenarios.

What is the Industrial Internet of Things (IIoT)?

Industrial Internet of Things (IIoT) is defined as a set of devices and applications that allow large enterprises to create an end-to-end connection environment from the core to the edge. It also includes traditional physical infrastructure, such as containers and logistics trucks, to collect data, react to events and make more informed decisions with the help of smart devices.

The Industrial Internet of Things (IIoT) is an extension of the Internet of Things (IoT) and has many applications in the consumer field. IoT use cases include, for example, smart home devices such as Amazon Echo, which use Alexa voice recognition function to remotely turn off the lights.

In industrial operation, this technology has been widely used in business in the environment with complex infrastructure and large equipment. In contrast, the industrial Internet of Things can remotely manage the heating, ventilation and air conditioning (HVAC) system of the whole factory. This is just an industrial Internet of Things use case that simplifies and improves enterprise operation and management.

How does the Industrial Internet of Things work?

Industrial Internet of Things is a subclass of Internet of Things, and enterprises are redefining how to connect, monitor, analyze and take action on industrial data to reduce costs and promote growth.

The idea behind the industrial Internet of Things is to use the data generated by "dumb devices" in industrial facilities for many years. Intelligent machines on the assembly line can not only capture and analyze data faster, but also exchange important information faster, which helps to make business decisions faster and more accurately.

The integration of information technology (IT) and operation technology (OT) promotes the development of industrial Internet of Things. It is a network matrix connecting devices, which collects and analyzes data through sensor technology and integrates it directly into the platform as a service. Industrial Internet of Things will herald a new era of industrial use cases, and there are many opportunities for economic expansion.

Industrial Internet of Things collects a large number of field data from factory workshops, transmits them through connecting nodes, analyzes them on the server, and transforms the information into operational opinions on the cloud platform. This encourages enterprises to make better decisions for their specific markets and target audiences. In other words, the industrial Internet of Things is a system that connects edge devices such as actuators, sensors, controllers, connection switches, gateways and industrial personal computers (IPC) to the cloud.

What is the connection between Industry 4.0 and Industrial Internet of Things?

Industry 4.0 is the product of the fourth industrial revolution. The definition of the fourth industrial revolution is the integration of traditional automated manufacturing and industrial processes driven by intelligent technology and autonomous communication equipment.

The word "Industry 4.0", abbreviated as I4.0 or I4, appeared in 2011, which is an initiative of the German government to vigorously advocate the digitalization of industrial processes in the past 20 years.

As mentioned by the Boston Consulting Group, the industrial Internet of Things is the main pillar of Industry 4.0, in addition to additive manufacturing or 3D printing, augmented reality (AR), autonomous robots, big data analysis, cloud computing, network security, horizontal and vertical system integration and simulation. This is because the autonomous communication between machines and the decentralized digital environment can automatically solve the problems that needed manual intervention before.

Industry 4.0 covers the industrial Internet of Things, digitalization and sustainable development of enterprises in a wider scope. Industrial Internet of Things is the driving force behind Industry 4.0, without which there would be no Industry 4.0. In other words, the industrial Internet of Things is limited to data detection, data transmission, data calculation, data processing and intelligent applications in specific fields.

Architecture of Industrial Internet of Things

The typical industrial Internet of Things architecture describes the arrangement of digital systems, so that they can jointly provide the network and data connection between sensors, Internet of Things devices, data storage and other layers. Therefore, the industrial Internet of Things architecture must have the following points:

1. Internet of Things devices at the edge of the network

These are groupings of network objects located at the edge of the Internet of Things ecosystem. These locations are as close as possible to the location of the data source. These are usually wireless actuators and sensors in industrial environments. A processing unit or small computing device and a set of observation endpoints. Edge IOT devices may include traditional devices, cameras, speakers, sensors and other instruments and monitors in brown land environment.

What happens at the most remote edge of the network? Sensors get data from the surrounding environment and the items they monitor, and then convert the information into indicators and figures that can be analyzed by the Internet of Things platform, and turn them into operational insights. Actuators control the processes that occur in the observed environment. They change the physical environment in which data is generated.

2. Edge data management and initial processing

Without high-quality and massive data, complex analysis and artificial intelligence cannot give full play to their potential. Even at the sensor level, data processing can be performed.

In this respect, edge computing provides the fastest answer, because the data is preprocessed at the edge of the network and in the sensor itself. Here, numerical and aggregated data can be analyzed. Once the relevant insights are collected, you can move on to the next stage instead of sending all the collected information. This extra processing reduces the amount of data sent to the data center or cloud.

3. Cloud for advanced processing

The preprocessing ability of edge devices is limited. Although it has been as close to the edge as possible to limit the consumption of local computing power, users will need to use the cloud for deeper and thorough processing.

At this time, we must choose whether to give priority to the agility and immediacy of edge devices or to the advanced insights of cloud computing. Cloud-based solutions can perform a lot of processing. Here, you can aggregate data from different sources and provide insights that are not available at the edge.

In the context of industrial Internet of Things architecture, the cloud will have:

concentratorIn addition to telemetry and equipment control, it also provides a safe link with the field system. If necessary, hubs can provide remote connections to local systems across multiple locations. It maintains all communication elements, such as connection management, secure communication channels, and device authentication and authorization.

save: Used to store information before and after processing.

analyse: It is helpful for data processing and analysis.

User interface:It provides the visualization of transmitting the analysis results to the end users, usually through the Web browser interface, but also through e-mail, SMS and telephone reminders.

4. Internet Gateway

Here, sensor data are collected and converted into digital channels for further processing at the Internet gateway. After obtaining the aggregated and digitized data, the gateway transmits it through the Internet so that it can be further processed before uploading it to the cloud. The gateway is still a part of the edge data collection system. It is adjacent to actuators and sensors, and performs preliminary data processing at the edge.

Gateways can be deployed as hardware or software:

hardware: The hardware gateway is an autonomous device. It provides wired (analog and digital) and wireless interfaces for downstream sensor connection. Internet connection is also provided, whether locally or through a standard link to a router.

software: On a PC, you can install a software gateway instead of connecting a hardware gateway. The software runs in the background or foreground, and provides upstream and downstream communication links as hardware entry points, and the PC provides physical interfaces. The software-based gateway can access the visual sensor settings and sensor data presentation through the user interface.

5. Connection agreement

Transmission of data across industrial Internet of Things systems requires protocols. These protocols should be in line with industry standards, well-defined and secure. The protocol specification can include the physical characteristics of the connection and wiring, the procedure for establishing the communication channel and the data format sent through the channel.

Some common protocols used in industrial Internet of Things architecture include:

Advanced Message Queuing Protocol (AMQP):This is a connection-guided, two-way, multiplexed and compact data coding message transmission protocol. Unlike HTTP, AMQP is built for cloud connection for IIoT.

MQ Telemetry Transmission (MQTT):This is a compact client-server message transmission protocol. MQTT is beneficial to IIoT devices because of its short message frame size and minimum code space.

Restricted application protocol (CoAP):This is a data reporting protocol that can be deployed through the transport layer, including User Datagram Protocol (UDP). CoAP is a compressed version of HTTP developed for IIoT requirements.

6. Industrial Internet of Things platform

The industrial Internet of Things system can now coordinate, monitor and control the operation of the whole value chain. These platforms control the device data, and manage the analysis, data visualization and artificial intelligence (AI) tasks of edge devices. In some cases, they can also transmit sensors directly to the cloud and back.

Industrial Internet Reference Architecture (IIRA) can be used as a reference for developing complex systems in the field of industrial Internet of Things. Generally speaking, IIRA’s framework advocates enterprises to use systematic methods to design frameworks, including feedback and iteration. In addition, the report suggests customizing the industrial Internet of Things design for specific business sectors, such as energy, health care, transportation and government use.

Advantages of Industrial Internet of Things

1. Improve efficiency

The biggest advantage of industrial Internet of Things is that it can help enterprises realize automation, thus maximizing operational efficiency. In addition, physical devices can be connected to software solutions through sensors to continuously monitor performance. This enables enterprises to better understand the operational efficiency of specific equipment and the entire fleet. In addition, the industrial Internet of Things realizes data-driven decision-making and remote monitoring of all production processes.

2. Increase production

By increasing the equipment utilization rate, organizations with IOT manufacturing processes may improve their productivity. As mentioned earlier, network devices provide continuous data flow, which can provide in-depth understanding of the operation of the devices. This can improve the overall efficiency of the equipment and maximize the performance of the machine during the running time. In addition, the use of industrial Internet of Things equipment also improves the utilization rate of human capital. Smart devices can be used to perform trivial, repetitive and dangerous activities, thus freeing employees to engage in other more strategic production-related work.

3. Reduce mistakes

The use of industrial internet of things forces enterprises to automate production operations. Eliminating human factors from industrial operation can eliminate the inefficiency that leads to defective products exiting the assembly line. With the reduction of quality defects, the profitability of enterprises will be improved due to the improvement of customer satisfaction and brand awareness.

4. Forecast maintenance requirements

Predictive maintenance is a strategy to avoid asset failure by analyzing production data to find patterns and predict upcoming problems.

The sensor of industrial Internet of Things can be integrated into industrial equipment, which can send out state-based management notice. These sensors record the temperature, humidity and other environmental variables in the work area, as well as the influence of material composition and transportation factors on transportation. All these data are useful for predictive maintenance. Therefore, asset failures can be avoided, expenses can be reduced, and machine downtime can be minimized.

5, to ensure the safety of workers

Intelligent manufacturing can achieve higher security, and all industrial IOT sensors cooperate to monitor the safety of employees and workplaces. Comprehensive safety system can protect workplace, production line and personnel. Once an accident happens, the whole facility can be notified, activities can be stopped, and senior management can mediate to solve the problem. This incident may also produce useful information, which can be used to avoid similar incidents in the future.

6, save energy costs

Industrial operation is the main source of global power supply, which is not conducive to sustainable development and the overall bottom line. Using sensors and small devices to continuously monitor the system may find inefficiencies that lead to waste. This includes not only monitoring equipment, but also comprehensive services, such as adjusting the temperature, water use, humidity and lighting of equipment. In addition, with the development of Internet of Things technology, sensors consume less energy, which is undoubtedly a boon.

7. Improve on-site service and customer experience.

Industrial Internet of Things can help improve the provision of on-site services. It is determined by time, context and technical personnel’s participation in specific service operations. The industrial Internet of Things also allows real-time data visibility. This means that original equipment manufacturers (OEMs), end consumers and any other interested parties will understand the risks and difficulties that arise, so as to gain a positive experience.

Nowadays, the industrial Internet of Things is the main product of large enterprises, and it is also one of the key products provided by major cloud providers such as Microsoft and Amazon Web Services (AWS). Industrial Internet of Things extends the capabilities of advanced data analysis and cloud to industrial applications, such as equipment maintenance, factory operation, supply chain management and personnel safety. Data from the industrial Internet of Things platform can even help to simulate and test products in a digital environment, perfectly integrate digital systems with physical systems, and improve industrial achievements exponentially.

The bank exploded, and the United States is going to have another financial tsunami?

Image source @ vision china

A brief introduction to Chinese businessmen

On March 11, 2023, the Silicon Valley Bank of the United States, which managed $175.4 billion in deposits, collapsed.

Founded in 1983, it is the 16th largest bank in the United States and one of the largest banks in Silicon Valley. By the end of 2022, the total assets under management of Silicon Valley banks reached 212 billion US dollars.

This is the biggest bankruptcy case of American financial industry since the subprime mortgage crisis in 2008.

01 trillion banks, can’t hold on.

On March 9, Silicon Valley Bank released a bad news:

The bank lost $1.8 billion because it sold $21 billion of securities in its portfolio. In order to avoid the liquidity crisis, the bank decided to sell common shares and preferred shares to raise $2.25 billion.

This bad news caused the share price of Silicon Valley Bank to plummet by more than 60% in a single day, and its market value evaporated by 9.4 billion US dollars.

On March 11th, even worse news came. The California Department of Financial Protection and Innovation announced the closure of Silicon Valley Bank, and the Federal Deposit Insurance Corporation was appointed as bankruptcy administrator to take over Silicon Valley Bank.

Although the Federal Deposit Insurance Corporation provides standard insurance of up to $250,000 for each depositor, it is not clear how much this will help the big customers of Silicon Valley Bank.

Silicon Valley Bank is very special. Most of its customers are technology startups and investment institutions, and it is a "to B" bank.

Since its establishment, the reputation of Silicon Valley Bank has been very good. So far, Silicon Valley Bank has supported more than 30,000 startups and more than 700 investment institutions, including facebook and Twitter. In the credit market of start-ups, it accounts for more than 50%.

The business of Silicon Valley Bank is not limited to the United States. It has branches in China, Indian, Israeli, British, German … almost the most important entrepreneurial market in the world.

In 2012, Shanghai Pudong Silicon Valley Bank was established. With the title of the first Sino-US joint venture bank in China, it has served more than 2,000 local science and technology enterprises in China. From artificial intelligence and big data to enterprise services, medical and health care, industrial internet and new consumption, all the hot spots are not absent.

Why did such a seemingly powerful and reputable bank collapse suddenly?

The answer is related to a key word:Federal reserve.

During the COVID-19 epidemic, in order to stimulate the economy, the US Treasury and the Federal Reserve passed a crazy rescue plan in 2020, injecting about 5 trillion US dollars into the financial market. In March of that year, the Federal Reserve cut interest rates by 150 points and kept the federal basic interest rate near zero. In 2021, Biden’s government continued to increase the code and approved the US rescue plan of 1.9 trillion US dollars again.

A large amount of water injection and the loose policy of zero interest rate have caused a surge in financial transactions and "stimulated" Wall Street into a prosperous period.

The Bank of Silicon Valley is also the beneficiary of this round of prosperity. In 2021, the amount of deposits in the technology stock bull market created by the liquidity of the Federal Reserve surged. At the end of 2019, the deposits of Silicon Valley banks were only about 60 billion US dollars, and in 2021, it reached the order of 190 billion US dollars.

After eating a lot of deposits, banks want to make money. The general way is to borrow money from commercial institutions and put the money out to eat interest.

The deposits of Silicon Valley banks originally belonged to these startups. Moreover, in recent years, the growth of technology and consumer industries has slowed down, and the financing needs of startups are declining, and loans cannot be released at all.

So, there is only one way for Silicon Valley Bank:Use customer deposits to invest directly.

In terms of investment direction, the choice of Silicon Valley banks is still relatively cautious. It allocates more than half of its assets to mortgage-backed securities.

The return on this investment direction is not high, and it is between 1.5% and 2% in yield to maturity. But the advantage is stability, not to mention that the cost of deposits used by Silicon Valley banks for investment is generally around 0.25%, which is still a steady profit.

However, the policy hand of the Federal Reserve changed this situation. The Federal Reserve, which had injected water crazily before, then began to raise interest rates crazily.

Since March 2022, the Federal Reserve has raised interest rates seven times in 2022, with a cumulative increase of 425 basis points. Eventually, the Fed will raise the target range of the federal funds rate to 4.25%-4.50%, reaching the highest level since the subprime mortgage crisis in 2008.

After entering the interest rate hike cycle, the investment assets held by Silicon Valley banks suffered a blow.

A very simple logic is:Even the federal funds have an interest rate of 4.5%. Who will hold mortgage-backed securities with a yield of only 1.5%?

As a result, the price of mortgage-backed securities began to plummet. At the same time, due to the change of financing environment, startups began to consume deposits continuously, and Silicon Valley banks were forced to sell assets to cope with liquidity, which led to a large number of floating losses in Silicon Valley banks and had to sell investment securities at a loss.

Until March 9, the Silicon Valley Bank, which really couldn’t survive, finally stood up and announced the fact that it had cut its own meat and suffered a sudden loss.

As soon as this news came out, it triggered a panic in the market, and everyone inevitably thought that there would be a big problem in the liquidity of Silicon Valley banks.

The collapse of the whole system of Silicon Valley banks inevitably happened.

02 Wall Street, chilly.

Under the Federal Reserve’s interest rate policy of "first drop and then increase", Silicon Valley Bank is not the only financial institution that has experienced ups and downs.

During the boom of Wall Street, the Wall Street financial giant Goldman Sachs Group also benefited greatly.

Its net income reached $59.3 billion and net profit reached $21.6 billion in fiscal year 2021, an increase of 137.25% year-on-year, setting a record high.The annual transaction income of the five largest investment banks in the United States, such as JPMorgan Chase and Citigroup, reached $100 billion for the first time in more than a decade.

In order to cope with the increasing business volume, Goldman Sachs set off a recruitment boom. By the third quarter of 2022, Goldman Sachs had 49,100 employees, a surge of 34% compared with 2018.

The number of employees in all walks of life has also greatly expanded. From the first quarter of 2020 to the third quarter of 2022, the number of employees at Morgan Stanley surged by 34%. The number of employees in JPMorgan Chase and Citigroup increased by 13% and 17% respectively.

But the good times won’t last forever. The interest rate policy of the Federal Reserve, after pushing the financial market to the top of the mountain, made it fall down severely.

In 2020-2021, under the bottomless water injection of fiscal and monetary policies, the American stock market was super prosperous and hit a new high. At the end of last year, the US Dow Jones index rose nearly 19%, the Nasdaq index rose 21%, and the S&P 500 index rose 27%.

By September 2021, their P/E ratios had risen to 25.9 times, 39.9 times and 26.6 times before the epidemic respectively, while their P/B ratios had increased by 6.73 times, 5.9 times and 4.51 times.

However, the interest rate hike in March, 2022 made the valuation of US stocks extremely cold. Since 2022, the benchmark S&P 500 index has fallen by 19.8%, or it will be the biggest annual decline since 2008. It’s like riding a "crazy roller coaster", described by Federal Reserve Chairman Powell.

The decline in the stock market has also slowed down IPO financing sharply, which has greatly reduced the business volume of Goldman Sachs. The number of transactions decreased by 73%, while the transaction amount decreased by 95%. Goldman Sachs’ financial report in the second quarter of 2022 showed that its net profit fell by nearly 48% due to the 41% decrease in investment banking income.

All kinds of unfavorable factors swept through, which made the net profit of Goldman Sachs increase between -40% and -50% in the first three quarters of 2022. At the end of October last year, judging from the results announced by the giants, JPMorgan Chase’s net profit fell by 17% year-on-year, and Bank of America’s net profit fell by 8% year-on-year.

This is really an ominous sign:The prosperity achieved by the Fed’s massive water injection a few years ago is coming to an end, and everyone should prepare for a hard life.

"Our future life will be very bumpy, and we must be more cautious in financial resources, especially in the short term",In 2022, David Solomon, CEO of Goldman Sachs, publicly stated many times that he hoped to control expenditure in the future.

The initial warning came from July. When Solomon saw the bleak second-quarter earnings report and judged that the US financial market would remain depressed for a long time to come, he immediately took precautions and told everyone in advance that Goldman Sachs would probably slow down recruitment and cut expenses in the future.

In September, Goldman Sachs first resumed its annual performance evaluation, which was suspended for two years during the epidemic. At that time, 500 people were laid off, and those who kept their jobs later found that their year-end bonuses were much lower than expected.

Regarding the future trend of Goldman Sachs, CEO David Solomon’s statement in an interview is:This will depend on the next Fed trend and inflation trend.

With the overall income of Wall Street investment banking business shrinking by about 40%, the largest banks have said that they will continue to lay off employees, and the whole industry is at the forefront of layoffs and optimization.

"Some people will be fired. We are laying off employees moderately around the world. For most enterprises, this is the situation after many years of development. "Following Goldman Sachs, james gorman, CEO of Morgan Stanley, also announced the layoff plan to the media.

Morgan Stanley has nearly 82,000 employees, and this time it announced about 1,600 layoffs, accounting for 2% of its total employees. In November 2022, the financial report data released by Morgan Stanley showed that its profit in the third quarter of fiscal year fell by 29% and its revenue fell by 12%.

Under such circumstances, JPMorgan Chase, Bank of America, Citigroup and other financial institutions are also considering cutting bonuses by one third or more. The incentive pay of wall street investment banks may drop by more than 45% as a whole.

Closely accompanied by the coldness of investment banks, there are deep worries about the prosperity and decline of the US stock market and anxiety about the Fed’s continued interest rate hike.What’s more, Wall Street generally has no hope for next year’s US economic forecast.

"The US stock market will usher in the worst year since the global financial crisis, and corporate profits will also suffer the same fate",Mike Wilson, chief US stock strategist at Morgan Stanley, warned the outside world.

The uncertain prospects of global economic growth make investors hesitate to invest for a long time. "We must assume that we will have some rough times in the future," David Solomon told the media. "We must be more cautious about financial resources and organizational scale.

On November 30, 2022, Powell once again publicly stated to the public that the fight against inflation in the United States is "far from over" and the Fed’s interest rate hike means "will continue for some time". He believes that,"In order to control inflation, we can even sacrifice certain employment and economy."

Continued aggressive interest rate hikes have made the market’s expectations of the US economic recession increase day by day. However, Biden said that the current economic slowdown is "not surprising" and he insisted that "the United States is still on the right path". And insist that even if there is a recession in the American economy in the future, it will be a very "slight" recession.

Wilson was very worried about the recent signs of weakness in the American economy under the warning of a recession. Perhaps the only good news is that Wilson assured the public that "fortunately, there are no signs of systemic financial risks or difficulties in the real estate market".

However, some business people represented by Musk are not so optimistic about the future.

Elon musk, CEO of Tesla, has warned people on many occasions that the US economy may face a serious recession lasting for one to two years. On November 30th, he wrote on Twitter again,"The Federal Reserve is continuously amplifying the possibility of a serious economic recession, and the US economic trend is worrying".

Jeff Bezos, the founder of Amazon, bluntly said in front of the media that the economic recession in the United States is close at hand. He even suggested that the American people should cut down their expenses as much as possible in the next few months and not buy big items such as cars and televisions.

The pessimistic attitude of entrepreneurs comes from their sober observation of themselves and their peers.

Just in January 2023,IBM announced that it will lay off 3,900 employees and reduce the number of its 280,000 employees by 1.4% in order to divest its medical and health business.

In 2022, the major technology giants in the United States have laid off more than 60,000 people. Many people think that this is the end, but they don’t know that this is just the beginning.

As soon as it entered 2023, Amazon confirmed the start of a new round of layoffs, and the number of layoffs is expected to reach 18,000, making it the largest layoffs in Amazon history;

Microsoft announced that it will lay off 10,000 employees at the end of March, and Google will lay off 12,000 people. Twitter’s goal is to reduce the total number of employees to the pre-IPO level.

Dr. Rouriel Roubini, who accurately predicted the financial tsunami in 2008, is even more pessimistic. He believes that the US economy will not only face inflation and fall into recession, but also usher in a "stagflation debt crisis".

"High inflation will bring higher interest rates, slow growth and weak labor market conditions, and it will also bring pain to families and enterprises," Powell explained to the public.

"We must control inflation. I hope there is a painless way to do this. But it didn’t. "

03 real estate, hidden worries reappear

As a core fixed asset, American real estate is actually in a delicate or even bad situation.

During the epidemic, the Federal Reserve continued to inject water for two years, which led to the financial prosperity and even the bubble on Wall Street, and also made American real estate once again reach a dangerous height and face risks.

In 2020-2021, the American property market experienced a boom that had never happened in the past two decades.

The term "How much will American housing prices rise in 2021" has increased by 350% in Google’s search frequency within one week.

"The price is almost heaven." A property buyer said excitedly, "Basically, it depends on grabbing. It is common for dozens of people to grab a suite. It is impossible to buy a house without increasing the price."

In April 2021, Mr. Zhang, a Chinese resident living in San Diego, California, told the media that a good house in the neighborhood of his home could be sold in about 10 to 15 days after it was hung up. In order to impress the original owner, the buyer must increase the price by at least 10%-20%.

"Every day, there are about thirty or forty groups of people waiting in line to see a suite. Many people come directly with cash, and sometimes they are directly robbed by the people in front before they get to themselves."

In 2021, 6.12 million properties were sold in the United States, reaching the highest performance since 2006. During the post-2020 epidemic, house prices in the United States rose by 43% in two years. At the same time, the data shows that house prices in 20 cities in the United States soared by 14.9%, the highest level since November 2005.

In the first half of 2022, the hot trend of the US property market continued. In February, house prices rose by 19.8% year-on-year, the highest increase in 25 years and four times the average increase in the past decade. In June, the median selling price of American houses reached the top, reaching $413,800.

Then, the violent interest rate hike by the Federal Reserve pushed the US property market to a cliff.

At the end of October, the mortgage interest rate in the United States exceeded 7%, the highest level in 20 years.

With the surge in interest rates, the confidence index of American home buyers quickly fell to 39, the lowest level since the 1980s. In December, the confidence index of NAHB home builders in the United States was only 31, lower than that in February 2007. An index below 50 indicates that the real estate market is in a contraction period.

Both buyers and sellers lack confidence, so that the originally fiery property market quickly entered the freezing period. At the same time, mclaughlin, chief economist of Kukun Company, boldly predicted that the good days when working families could easily buy high-quality housing anywhere in the United States were over. The soaring housing prices in the past two years may force them to extend the time for buying houses by five to ten years.

The rising interest rate has greatly increased the repayment pressure of buyers; The continued high housing prices have hit the purchasing power of buyers. In the end, the demand for house purchase dropped sharply, and house prices changed from rising to falling.

In 2022, the total sales of finished houses decreased for 11 consecutive months, and house prices could not keep increasing in the second half of the year and began to fall back. Diane Swonk, chief economist of KPMG, predicts that house prices in the United States may fall by 15% in 2023.

The sudden slowdown of the once-soaring real estate market has made the value of stocks firmly bound to real estate extremely uncertain. Nouriel roubini believes that the stock value will shrink by as much as 40% in 2023.

After experiencing the soaring housing prices in the United States during the two years of the epidemic, many people began to wonder whether the real estate bubble in the United States has piled up again. After all, one of the keys to the financial tsunami in 2008 was the real estate bubble. Now the United States is facing the double pressure of real estate and stock market, which makes people worry that a bigger storm is brewing.

Many people have forgotten that in 2002, former Federal Reserve Chairman Alan Greenspan said,"The prosperity of our real estate market is actually composed of a large increase in mortgage debt, which is unsustainable".

On the eve of the financial tsunami, American house prices rose by 10% every year, by 17% in 2005, and by more than 50% in the four years from 2003 to 2006.

In the five years from 2002 to 2006, the Federal Reserve raised interest rates 17 times in a row. As a result, the interest rate of housing loans has soared, and house prices have collapsed and fallen wildly.

The collapse of the property market brought down the stock market and plunged by 50%. Then the unemployment rate increased greatly, and household consumption and business investment decreased rapidly. Finally, the American economy fell into recession. Then it triggered a financial tsunami that affected the whole world.

Although Federal Reserve Chairman Powell is optimistic, he believes that the current American real estate is very stable, and there will be no recurrence of the 2008 crash and no economic recession. The mainstream view of the market also believes that the rapid rise of the US housing market during the epidemic period is not the same as that in 2008.

However, nowadays, large banks have closed down, technology and finance enterprises have laid off a large number of employees, the stock market has flourished and declined, and the hidden troubles in the housing market have reappeared. More and more people feel that this story is deja vu.

While they are worried, they keep speculating in their hearts: Under the operation of the Federal Reserve, will the collapse of the Silicon Valley Bank be just the beginning? Will the United States set off another financial and economic catastrophe?

Google released the "super brain" PaLM-E in history, and robots have become versatile since then.

What happens when ChatGPT has vision?

Editor’s note: ChatGPT has grabbed most of the limelight in the AI field during this time. But recently, an AI model PaLM-E launched by Google, which has visual ability and can guide robots without special training, has also shown impressive capabilities. The emergence ability of this largest visual language model so far makes people think about general artificial intelligence. The article comes from compilation.

On Monday, a team of artificial intelligence researchers from Google and Technical University of Berlin launched a multimodal visual language model (VLM), which is called Palm-E. The model has 562 billion parameters and integrates vision and language for controlling robots. Researchers claim that this is the largest VLM ever, and it can perform various tasks without retraining.

According to Google, PalM-E can generate an action plan for a mobile robot platform (developed by googlebot) with a mechanical arm, and then execute it by itself, just by giving it a high-level command, such as "Give me the rice cake in the drawer".

PaLM-E realizes this by analyzing the data from the robot camera, and the whole process does not need to preprocess the scene representation. In this way, there is no need for human beings to preprocess and annotate the data, and the control of the robot can be more autonomous.

In the demo video provided by Google, PaLM -E carries out the instruction of "get me a bag of rice chips from the drawer", which includes several planning steps and visual feedback from the robot camera.

This model is also flexible and can respond to the environment. For example, the PaLM-E model can guide the robot to the kitchen to take out the rice cake bag. Because PaLM-E is integrated into the control system, it can be tolerant of possible interruptions during the task. In a video example, the researchers put the rice cake bag picked up by the robot back several times, but the robot will find the rice cake bag again and pick it up again.

In another example, the same PaLM-E model is shown to control the robot autonomously through tasks with complex sequences. Previously, such tasks often required manual guidance. Google’s research paper explains how PaLM-E transforms instructions into actions:

We demonstrated the performance of PaLM-E in challenging and diverse mobile control tasks. In the setting, we mainly follow the setting of Ahn and others. (2022), that is, robots need to plan a series of navigation and manipulation actions according to human instructions. For example, give the instruction "I spilled my drink, can you bring me something to clean up?" After that, the robot needs to plan an action sequence including "1. Find the sponge, 2. Pick it up, 3. Give it to the user, 4. Put it down". Inspired by these tasks, we developed three use cases to test PaLM-E’s embodied reasoning ability: fitness prediction, fault detection and long-horizon planning. Low-level policies come from RT-1 (Brohan et al., 2022), which is a transformer model. It can use RGB images and natural language commands, and then output end-effector control commands.

PaLM-E belongs to the "next-token predictor", so it is called "PaLM-E" because it is based on Google’s so-called "PaLM" large language model (similar to the technology behind ChatGPT). By adding sensory information and robot control, Google "visualized" PaLM.

Because it is based on the language model, PaLM-E can continuously observe, for example, images or sensor data, and encode them into a series of vectors with the same scale as language tags. In this way, the model can "understand" sensory information in the same way as language.

Google also provided a demonstration video, which showed that a robot "gave me a green star" under the guidance of Palm-E. The researchers said that this green star "is an object that this robot has not directly touched before."

In addition to the RT-1 robot transformer, PaLM -E also draws lessons from Google’s previous work on ViT-22B. ViT-22B is a visual transformer model released in February this year. ViT-22B has been trained in various visual tasks, such as image classification, object detection, semantic segmentation and adding subtitles to images.

Google Robotics is not the only research group dedicated to robot control using neural networks. This research reminds people of the paper recently published by Microsoft (ChatGPT for Robotics), which also discusses the control of robots by combining visual data with large language models in a similar way.

Robots aside, Google researchers have observed some interesting effects, which are obviously because PaLM-E uses a large language model as its core. First of all, it has the performance of "positive migration", which means that it can transfer the knowledge and skills learned from one task to another. Compared with the robot model with single task, the performance of the former is significantly higher than that of the latter.

In addition, they also observed a trend of model scale: "The larger the language model is, the more it can maintain its language ability when training with visual language and robot tasks-in terms of quantity, the PaLM-E model with 5620 parameters almost maintains all its language ability. “

PaLM-E is the largest VLM reported so far. Although we have only received the training of single image prompt, we have observed the emergence of emerging abilities such as multimodal thinking chain reasoning and multi-image reasoning. Although this is not the focus of our work, PaLM-E has set a new SOTA (Best Performance) on the OK-VQA benchmark.

——Danny Driess

Researchers claim that PaLM-E has demonstrated its emergent abilities, such as multi-mode thinking chain reasoning (which allows models to analyze a series of inputs including language and visual information) and multi-image reasoning (which uses multiple images as inputs to make reasoning or prediction), even though it has only been trained with single image cues. In this sense, as the deep learning model becomes more and more complex, PaLM-E seems to continue to surprise people.

Google researchers also plan to explore more applications of PaLM-E in real-world scenes, such as home automation or industrial robots. They hope that PaLM-E can stimulate more research on multimodal reasoning and embodied AI.

The word "multimodal" is very hot now, and we will hear more and more in the future, because major companies want to make general artificial intelligence that looks like human beings to perform general tasks.

Translator: boxi.

Engineers built AI chat bots several years ago, and Google has been obstructing artificial intelligence efforts.

It is reported that a few years ago, a former Google engineer developed a conversational AI chat robot.

But for security reasons, Google executives blocked their efforts to release it to the public.

Google is now catching up with Microsoft’s artificial intelligence and plans to release its artificial intelligence chat bot this year.

Google is expected to release Bard, a highly anticipated AI chat robot, in the near future. But according to a recent report, a few years ago, two former Google engineers urged their former employers to release a similar chat robot to the public, and they met with resistance.

According to former colleagues, around 2018, Daniel de Freitas, a Google research engineer, began to work on artificial intelligence projects, with the goal of creating a dialogue and chat robot that imitates human speech. Noam Shazeer is a software engineer in Google’s artificial intelligence research department and later joined the project.

De Freitas and Chazel can build a chat robot, which they call Meena. They can discuss philosophy, talk about TV programs at will and make puns about horses and cows. They believe Meena can fundamentally change the way people search online.

However, Google executives said that the chat bot did not comply with its artificial intelligence safety and fairness standards, and they renamed it LaMDA, which will become the language model behind Budd. Executives have repeatedly tried to send robots to external researchers, add chat functions to Google Assistant, and release demonstrations to the public.

Although CEO Sandahl Picha personally asked them to stay and continue to develop chatbot, Defree Tass and Chazel were disappointed with the response of the executives and left Google at the end of 2021 to start their own company. Their company is now called Character.Ai, and since then it has released a chat robot, which can play characters such as elon musk or Mario of Nintendo.

Since 2012, Google has been obstructing its artificial intelligence efforts, and it is nothing new for Google to hesitate to release its artificial intelligence tools.

It is reported that in 2012, Google hired computer scientist ray kurzweil to study its language processing model. According to TechCrunch, about a year later, Google acquired DeepMind, a British artificial intelligence company, which aims to create artificial general intelligence.

However, due to moral concerns about large-scale monitoring, academics and technical experts refused to use the technology, and Google promised to limit the way it uses artificial intelligence. In 2018, Google terminated its project of using artificial intelligence technology for military weapons in response to employee opposition.

But Google’s artificial intelligence plan may finally have a dawn now, although the discussion about whether its chat bot can be launched responsibly continues. Bard, the company’s chat bot, will be launched after Microsoft releases its own chat bot through Bing, and Microsoft’s stock is rising.

Last month, Google’s Bard chat bot made a factual error in its first public demonstration, and Google employees quickly called the announcement "hasty" and "clumsy". John Hennessy, chairman of Alphabet, agrees that Google’s chat bots are "not really ready to launch products".

1:0! Harland outbursts: 28 goals in 26 games broke the team history record again, and Manchester City was 2 points behind Arsenal.

In the early morning of March 12th, Beijing time, the 27th round of Premier League continued, and Manchester City played away against Crystal Palace. In the first half, Rodley’s shot was saved by Guaita, and in the second half, Foden’s free kick was saved by Guaita. Then Alvarez turned to lead the ball and volleyed it high, Gundogan made a point and Harland hit it. In the end, Manchester City beat Crystal Palace 1-0 away.

Manchester City beat Newcastle 2-0 at home in the last round to win the local derby. However, Nalsen’s winner made the Gunners still lead the standings by five points, making it difficult for Manchester City to defend the Premier League title. Crystal Palace, headed by former Arsenal captain Vieira, has never won a game in the last 10 games, only four points higher than the relegation zone before the game, but they won 1 win and 1 draw against Manchester City in two rounds last season. In the first leg, Manchester City won 4-2 at home, and Harland scored a hat trick.

Although the key matchup of the second leg of the Champions League knockout round is about to come, Guardiola did not have a clear rotation in this game. The two main players, De Braune and Walker, all sat on the bench, and the four central defenders, Ake, Diaz, Akanji and stones, lined up in the defense line; Rodley, Gundogan and B are in the midfield; Grali, Foden and Harland teamed up to attack the Trident.

There is a big gap between the two sides in terms of strength. Crystal Palace directly stationed troops to defend, while Manchester City formed a siege. In the second minute of the opening, Grali made a cross from the left, Diaz headed the ferry, and Rodri shot was saved by Guaita bravely. In the 4th minute, Grali made a breakthrough through a dragon, and then a shot in front of the restricted area missed slightly. In the 26th minute, Akanji made a long-range shot outside the restricted area, and the ball grazed the crossbar. In the 28th minute, Rodley went straight to the left, Grali made a cross on the left, and Aker inserted it and passed it again. Harland outflanked in front of the door and pushed it high. In the first half, Manchester City had 72% possession and 8 shots, but the score was still 0-0.

In the second half, Manchester City continued to exert pressure. In the 55th minute, Grali won an excellent free kick for Manchester City, and Foden shot the free kick directly. Guaita once again turned the ball away. Seeing that the situation continued to be deadlocked, Guashuai took the lead in taking the initiative to replace people, and Alvarez went on stage to strengthen the attack. In the 60th minute, Grali sent a straight plug from the right, and Alvarez turned and led the ball, then volleyed high.

Until the 78th minute,Gundogan was thrown to the ground by Ollie in the penalty area, and the referee decisively awarded a penalty. Harland stood in front of the penalty and hit it, 1-0!Manchester City finally took the lead, which is Harland’s 28th Premier League goal this season. Even if Kane scored twice this round and increased the number of goals to 20, there is basically no suspense in the Premier League Golden Boot. It is worth mentioning that this goal is also Harland’s 10th away goal in the Premier League this season, and now his home and away league goals have reached double digits. Previously, only aguero in Manchester City achieved double-digit goals at home and away in the Premier League in the 2014/15 season, and Harland became the second player in the history of Manchester City to do this.

Thanks to the only goal scored by Harland, Manchester City took away three points with a 1-0 away win and won four consecutive victories in various competitions. In the case of one more round, the difference with Arsenal was temporarily narrowed to two points, and Manchester City will face Red Bull Leipzig at home next week, which will determine whether Manchester City can enter the key battle of the Champions League quarter-finals.

Guardiola: The Premier League is bigger than the Champions League! Manchester City fights for the League, and I don’t rely on Harland to win.

De Blau got Guardiola’s support in his recovery, but the Manchester City manager hinted that if he didn’t play his best, there would be no guarantee that he would get the starting position. De Braune recently missed two games due to injury, and resumed starting in last week’s game. Guardiola called his performance in the first half "passive", but after the intermission, the Belgian’s performance improved and scored a wonderful goal.

However, in the game against Newcastle last week, De Braune was somewhat out of step, which prompted the manager to tell him to "get back to basics" in this week’s training. With Manchester City’s away game against Crystal Palace on Saturday, Guardiola supported him: recovering from a slight decline.

Guardiola explained: De Braune is ready to play, just like other Manchester City players. On the first day of our arrival, everyone stayed late for training. At the same time, Gua Shuai also said: He reserved the possibility of rotating the lineup at Selhurst Park. I know him well and know how important he is. At the same time, there are ups and downs, just like Phil Forden and other players, they compete with other players. At a certain stage of the season, it is normal that one player may be better than another player and the other player may be worse. I hope the team can win this special game. Football is always so unpredictable. In these seven years, De Braune has always been himself. When he missed the game, he was either rested, tired or injured.

Guardiola said that the next week will be "decisive" for Manchester City, which will be their last Premier League game before the international match day, followed by the second leg of the Champions League 16 against Leipzig and the FA Cup quarter-final against Burnley. After announcing that De Braune was ready for these games, Guardiola insisted: The only focus of Manchester City is to go to London to challenge Crystal Palace.

"I will never give priority to the Champions League. We are fighting for the Premier League. The most important thing is the league. We have had a long week, so we need to have a good rest, rest for two days and try to play well. This has nothing to do with Leipzig. The strength of Crystal Palace deserves attention. They have many ace players in the frontcourt, and their games are unbelievable. We have to pay attention to a lot. The attack of Crystal Palace is incredible. They scored two goals through positioning. We must be careful. The players know this! "

Guardiola also talked about how Manchester City adapted to Harland’s tactics: "Find this balance in the right way, because when Harland starts running with strength, he tends to give him the ball because he is going to score!"! But there were several accidents in the game against Newcastle. I won’t rely on Harland alone. Our tactics are very rich! "

So for this news, fans, do you have anything to say? If you like this article, welcome to pay attention to Beta and chat with the stars and the ball game.

How will the AI big model craze affect developers and enterprises?

With the continuous development of artificial intelligence technology, the upsurge of AI big model is on the rise. Large model refers to a deep learning model with billions or tens of billions of parameters, which can be used in various fields such as natural language processing, image processing and speech recognition. The rise of the AI ? ? big model will have a far-reaching impact on developers and enterprises. Below we will analyze it from the following aspects:

Ten trillion parameters, can you pile up a general artificial intelligence?

# Headline Creation Challenge ## One knowledge point every day ## 10 trillion parameters, can you pile up a general artificial intelligence? #

Although the current large-scale language models, such as GPT-3 and GPT-4, have more than 10 trillion parameters, these models still cannot be regarded as the realization of General Artificial Intelligence (AGI).

General artificial intelligence refers to an artificial intelligence system that can carry out various intelligent activities like human beings. It has the ability of self-learning and adaptation, and can cope with various situations and tasks. Although the current large-scale language models can perform well in some tasks, they are still only in the category of "Narrow AI".

In addition, general artificial intelligence needs to have more abilities, such as perception, exploration, understanding and emotion, which are still not realized by simple parameter stacking. Therefore, to achieve real universal artificial intelligence, we need to innovate and make progress in algorithms, hardware and software.

However, to achieve real universal artificial intelligence, we need to innovate and make progress in the following aspects:

  1. Algorithm: Although the current large-scale language models can perform well in some tasks, they are still only in the category of "narrow artificial intelligence". Therefore, we need to research and develop more general artificial intelligence algorithms, such as reinforcement learning, meta-learning and transfer learning.
  2. Hardware: In order to realize general artificial intelligence, a lot of computing resources are needed to support the operation of artificial intelligence system. Therefore, it is necessary to research and develop more efficient and intelligent computing hardware, such as quantum computers and neuron chips.
  3. Software: To realize general artificial intelligence, we need to develop a more perfect software system, including autonomous learning and adaptability, emotion and consciousness. This requires us to deeply study the essence and mechanism of human intelligence and apply it to artificial intelligence systems.
  4. Human-computer interaction: To realize general artificial intelligence, we need to research and develop more natural and efficient human-computer interaction methods, including voice, images, gestures and other forms. This requires us to conduct in-depth research on human interaction patterns and habits and apply them to artificial intelligence systems.

To sum up, the realization of real universal artificial intelligence needs innovation and progress in algorithms, hardware, software and human-computer interaction. Although the current large-scale language model parameters have reached 10 trillion, they still cannot be regarded as the realization of general artificial intelligence. We need to make continuous efforts and exploration to realize the real general artificial intelligence.

ChatGPT

Tevez: I am sad to leave Manchester United, but he made me miserable.

[Tevez: I was very sad to leave Manchester United, but he made me very sad] Tevez said in an interview that he wanted to stay at Old Trafford, but Ferguson’s false promise meant that he had no other choice. When Manchester City provided him with a core position that Manchester United did not have, he seized the opportunity.

Tevez said: "I was on loan to Manchester United. Ferguson told me that they would buy me, and then he brought Berbatov. I didn’t need to think too much because I was angry with Ferguson. As a coach, he is phenomenal. He has been in a club like Manchester United for a long time, but I have a little contradiction with him. "

"They told me at the time,’ We will buy you, but we will also bring Berbatov. Don’t worry, I’ll bring him to the competition with you. But we will talk to your agent and reach an agreement on the contract and transfer. But they didn’t call my agent, nothing. Time is running out and they want to lower my price. Every time I go on stage, I behave, and people start calling my name. This is a year-long process. "

"Later, I almost reached an agreement with Manchester City. My family and I went to Abu Dhabi to meet him and told him that I was going to Manchester City before the Champions League final."

"It’s like a dagger for Ferguson. Me too, because I love Manchester United. But for me, he didn’t let me go on stage for a whole year. He made me miserable. This makes me sad because I love Manchester United. I like playing at Old Trafford. It gives me that feeling. Then Manchester City came and they told me that they wanted me to be the core of Manchester City. He showed me the club’s plan today, that’s all. "