2023 is close to halfway, but no one has yet said that this year is "XX year", after all, we have allocated the "first year of AI", "first year of VR", "first year of meta-universe", and the first year of blockchain, Web3 and other technologies under the optimism of the optimistic era.
On the contrary, 2023 is no longer the era of optimism, and that fiery optimism is becoming extremely cautious, the simplest example of which is that OPEN AI's current valuation is less than $30 billion, and if it had been four or five years ago, its valuation would have doubled or even more.
Of course, this year is destined to have many other things, such as the smartphone industry has completely entered a downward channel, once considered the main cause of the decline in shipments after the epidemic factor is removed, everyone can only face the harsh reality: everyone has become interested in smartphones. Another example is that Alibaba Cloud opened the largest price reduction in history, and then Tencent Cloud followed suit. Another example is Qualcomm's demonstration of Stable Diffusion (an open-source AI drawing tool) running natively on Android phones. Or something that has little to do with the Internet, such as everyone suddenly has enough confidence in controlled nuclear fusion...
Of course, with the vastness of the world, the length of the year, similar news every year, just look at these, 2023 is similar to the fifteenth year of Wanli, but it is a dull year in history.
But if the time dimension is extended, when we look back at 2023, there must be two products that have left their names in the historical dimension: the GPT-4 that has been released, and the just-released Apple Vision Pro headset.
2023 is not any "first year", but because GPT-4 and Apple Vision Pro have become an extremely important year.
intellect
In fact, GPT-4 is an umbrella term: AI tools that can replace some of the mental work.
For example, Microsoft Office Copilot, which originally required to learn Excel formulas for a long time to achieve data analysis, can now be done in just one sentence; The PPT that originally required a whole day to make is still a matter of words.
▲ The picture generated by MidJourney, the glass reflection is extremely realistic
Another example is MidJourney, an AI painting tool, which can output all kinds of high-quality images in a matter of minutes, whether it is two-dimensional illustration style or fake real pictures taken by professional cameras.
There is also the "intelligence" presented by GPT-4 itself as a representative of the big language model and general artificial intelligence: if it is allowed to take the "American college entrance examination" SAT test, its score can beat 90% of the candidates, and the score is enough for it to enter famous schools such as Harvard or Stanford in the United States.
In addition, a variety of AI tools are now emerging every day, in some ways showing efficiency far beyond human capabilities.
Like the "emergence" phenomenon shown by large models such as GPT-4, AI tools are now emerging, and for most people, it is impossible to know that these tools exist without talking about learning and using these tools.
In "The Long Season", the male protagonist Wang Xiang has driven a train in a steel factory for decades, and his skills in driving trains allow Wang Xiang to hold an iron rice bowl and enjoy a good social status and family status.
In the same or earlier period, a skill that was not hard to acquire was enough to secure a job for decades: Fifty or sixty years ago, an American high school graduate who could use a typewriter or a calculator could find a white-collar job at a company for 30 years.
Now, no one thinks they can stay in a company for 30 years by mastering Office software.
10 years ago, I was writing some compiled news in Aifaner, and writing 6 news a day had made me feel tired, but now with the translation and summary ability of ChatGPT, it is not difficult for a high school student with basic English and language skills to write 60 news news a day.
This is the simplest arithmetic problem, and the current AI tools cannot completely replace people, but they can achieve the effect of "1 person + AI = 10 people" in some jobs.
This kind of arithmetic problem is not used as a strategic weapon of deterrence, but as a situation that is happening, and many game companies have used AI for character original painting and scene modeling.
In the promotional poster of the Hema flower sale promotion "Peony Season" last month, MidJourney has been used to make pictures of flowers that are not in flower, and if it is not specially marked, almost no one will see that this is an AI drawing.
▲ Many people speculate that this "Honor of Kings" game illustration is an AI drawing and manually modified
Not long ago, an illustration in the mobile game "Honor of Kings" also caused many players and artists to speculate: because of the irrationality of many details, everyone suspected that this was an AI drawing and a slightly modified work.
A marketing employee of a major game factory told Aifaner that the price of a similar single poster illustration for outsourced artists is 20,000-50,000 yuan, and it will be more expensive if there are special requirements.
Of course, many people, including myself, will find that AI tools do not yet have the ability to replace themselves, or do not integrate well into their workflows. But people who are confident in themselves also know in their hearts that today it will not work, but one day it can, and this day will not be long.
From the perspective of most workers, AI tools such as ChatGPT will not make a certain job disappear, but will sharply reduce the demand for jobs, and the work model of a small number of elites plus AI makes people who were originally about the industry average face the risk of unemployment, and these people are the mainstream population of the industry.
The initiator OPEN AI itself has released a report, listing many of the occupations most susceptible to ChatGPT, and many of them that were once considered white-collar or even gold-collar are at risk of being replaced.
This risk is not later, but now.
Not long ago, the Hollywood screenwriting industry is on strike, in addition to protesting the low treatment, NetFlix and other employers to squeeze labor, the contradiction also focuses on AI creation, the collective demand of this group of screenwriters is "ban the use of AI to write literary materials; It is forbidden to use it as original material; It is forbidden to train AI using materials created by writers."
However, the chairman of the AMPTP (Alliance of American Film and Television Producers), who represents the interests of employers, said:
It's already a blessing to have a short-term job as a screenwriter.
The implication is that you may not even be able to find short-term jobs in the future.
From the perspective of employers, the efficiency improvement and labor cost reduction brought by ChatGPT are huge opportunities, which is why so many Internet industry bosses are keen to forward various AI progress to the bottom of the circle of friends, they firmly believe and hope that AI can reduce costs and increase efficiency.
As economist Tyler Cowen puts it in The End of Average, the industry that employs more people will create greater business value by disrupting that job.
The hired workers naturally have another idea, reducing costs and increasing efficiency, reducing costs by layoffs, increasing efficiency by squeezing, and horizontal vertical cannot escape the tragic fate.
▲ John Deere CP690 cotton picker
Taking agriculture as an example, Ai Fan'er once went to Yuli County, Bayingolin Mongolian Autonomous Prefecture, to see how modern agricultural machinery allows a few people to manage thousands of acres of cotton fields: an agricultural drone can spray more than 150 acres of agricultural land per hour, equivalent to the efficiency of 60 people. A John Deere CP690 cotton picker worth more than $5 million is equivalent to more than 700 cotton pickers.
It can be said that the development of agriculture and industry in the past thousands of years, especially in modern times, is the history of creating a "physical surplus" and replacing human power with machines.
If repetitive physical work can be replaced by machines, why can't repetitive mental work?
Now it seems that the barrier of intelligence is not stronger than the barrier of physical strength.
Carrier of intelligence
In fact, it is inappropriate to call AI such as ChatGPT a tool, because for humans, tools do not need intelligence, do not need any human definition of "subjective initiative", follow specific logic, kitchen knife chopping, WeChat chat, Photoshop retouching, etc., every action we use tools has a clear expectation.
A typical example of this expectation is the "graphical interface, GUI", whether it is a PC, Mac or a smartphone, it relies on the graphical interface for control, we touch the WeChat icon, the mobile phone will never open Weibo. We click the meeting recording button on Feishu and it will never open the monthly report page. This is because the program is written dead, and the path from A to B is certain and transparent.
But for ChatGPT, it's not just that we can't guarantee that the expected result is clear, but we don't know what happens between the command we give and the ChatGPT output.
As mentioned earlier, in many aspects involving reasoning and understanding, ChatGPT3.5 appears to be mentally retarded, but GPT-4 behaves like a student with high emotional and intellectual skills.
The hot search for science and technology news in the first half of 2023 is obviously dominated by ChatGPT, of which there are three news hottest: GPT-4 release, Microsoft Office Copilot based on GPT-4 capabilities, and ChatGPT released iOS applications.
Apple's Vision Pro headset
Then, there was the release of Apple's Vision Pro headset.
The reason why this device is so valued is that people are tirelessly looking for the next computing platform after the smartphone, and once people pinned their hopes on wearable devices such as smart watches, but it turned out that its positioning is unable to work and entertainment, and the innovation that smart watches can do in interaction is lackluster.
▲ A large number of sensors and cameras provide the basis for new interactions
So, what does Apple's Vision Pro headset do with interaction?
We control computers and need a keyboard and mouse; We control smartphones, need to touch the screen, there is always a thing as a "medium" to connect devices and people, and on the Apple Vision Pro headset, we hardly need this "medium", gestures, eyeballs and mouth become the main interactive tools:
- Gesture: Represents the action that will be performed
- Eyeball: Represents the direction of attention
- Mouth: Represents heavy content input
The reason why the keyboard and mouse is accurate is because we click the F key, it will not be recognized as the G key, but the touch screen occasionally has the phenomenon of mistouch, for mistouch, many input methods have launched "intelligent error correction", when we play similar to "zjihui", intelligent correction to "zhihui", which is the device began to have its own "judgment".
For Apple's Vision Pro headset, it is "judging" almost all the time: what does the user's gesture represent, where is the eyeball looking, and what does this sentence mean?
In short, it needs to be "smart" enough to perform such interactions. In some game console accessories and smartphones, gesture manipulation, eye tracking, and voice recognition are not new, but they are all icing on the cake, and they are not the main way to interact.
But there is no keyboard and mouse and touch screen Apple Vision Pro headset, the combination of the three, not only lost the baggage of the past, but also opened a new future, the freedom and dimension of interaction has been liberated to the greatest extent, before the commercial use of brain-computer interfaces, human new human-computer interaction methods will be based on this: without the help of "medium", human organs are the main body of interaction.
If you still feel that there is something missing from Apple's Vision Pro headset, there is no doubt that it is a large language model like ChatGPT and related applications.
▲ One of the most shocking scenarios in 2023: Office Copilot automatically generates PPT
If you keep your mission in mind, insist on being the best Apple ecological developer, save Office Copilot until June, and demonstrate in Apple's Vision Pro headset, it will definitely be more explosive than Mickey Mouse and the like.
What migrant worker doesn't want to generate a PowerPoint by saying a word to Office Copilot in the virtual world, and then send a job to the boss and watch a movie by himself?
Is it hard for Microsoft to put Office Copilot on?
It's not difficult at all.
So if Stable Diffusion can run on the Snapdragon flagship phone and ChatGPT can run on iOS, then the Apple Vision Pro headset with an M2 chip and the new interaction it brings means that it can be, and should be, a "carrier of intelligence".
If you can't understand why a large model like ChatGPT is an intelligence and Apple's Vision Pro headset is a carrier of intelligence, then Meta, which has sunk all the way in the era of advocating the "meta-universe", has continuously increased its stock price in the AIGC era, and many large models released by it can be linked to the VR business, at least proving the capital market's recognition of its logic, which is also an example of supporting "intelligence and intelligence carriers".
All AR, VR and XR practitioners are looking forward to Apple's "proof", and the $3499 price leaves enough market space for other related products.
Such as PICO 4 Pro is light (597 grams), good performance, viewing angle and resolution at the mainstream level, equipped with new technologies such as eye tracking and face tracking, with a preliminary ecology, but also with a little cost performance (compared to Vision Pro) products are expected to become many people's "replacement", in addition to the price, Apple Vision Pro market time and initial production also leave a lot of space for everyone.
Another "electronic product" with more and more sensors, stronger computing power, and higher and higher voice interaction status is the car.
Not long ago, Li Xiang, the founder of Ideal Auto, said:
Intelligent driving and intelligent space have entered the era of large models, and the research and development and training of large models is a necessary capability for intelligent electric vehicle enterprises, otherwise it will only stay in the era of electric vehicles.
This is another example of a different category.
▲ Use the coolest Vision Pro to do the most boring work
Intellectual surplus
When I went to see the graduation exhibition of Guangdong Academy of Fine Arts not long ago, I saw an exhibition area cooperated by Guangdong Academy of Fine Arts and Tencent, with the theme of the future city WeCityX, while enjoying the future office, living, and travel scenes, students who had not yet stepped out of the campus showed their prospects full of future perspectives:
- One graduation project is AR glasses, which meet the needs of working from anywhere
- Another graduation design work is the future workstation, with AR glasses, you can work, exercise, and rest at the workstation
- Another work is Future Travel, where a workplace is arranged in a driverless car, and you can not drive, but you have to work
Earlier, when interviewing Kingsoft office executives, I asked, "Has AI made the mobile office that was once a false proposition come true?" , the answer is yes.
Dealing with tables on mobile phones is a nightmare for many people, and the operation of "swapping the third and eighth rows of the table" on the computer will become very cumbersome on the phone, but if it is an AI with "intelligence", we only need to say that on any device, the efficiency of the PC, mobile phone, tablet and even smart headset is the same.
For the vision of the future work scene, Guangmei graduates have predicted one point: work is like the wind, always accompany me.
Why is this happening?
In fact, we are getting closer and closer to the two elements of processing work, information and intelligence.
A recent Pew Research Center survey of U.S. adults showed that younger, more educated and higher-income people use ChatGPT more.
We can't accurately predict what this group of people who are more open to AI will become in the future because of AI, but the Matthew effect of "the stronger the stronger" is indeed everywhere.
Relaxed to the urban dimension, the price increase of urban core real estate is often greater than that of suburban areas; Urban economic growth and energy consumption levels also tend to be greater than population growth; Most of the fruits of urban economic growth are obtained by the top 10% of income groups, of course, this group will also be wrapped up in the fast pace of urban growth, back to the previous said: work is like the wind, always with me.
If the Pew Research Center report adds a geographic dimension, it would almost certainly be that people in the Bay Area of California or New York would be more likely to use ChatGPT than people in the Rust Belt of the Great Lakes.
Similar cognitive leadership in the past examples abound, if you become the first batch of ride-hailing drivers around 2013, because of platform competition and preferential policies such as subsidies, it is very simple to earn 20,000 or 30,000 a month. The current situation is that traffic management departments in Changsha, Sanya and other places have issued early warnings, saying that the number of local ride-hailing drivers tends to be saturated and is not suitable as a career choice.
Because AI tools such as ChatGPT have a very obvious "silly duality", which makes many people's cognition of them not clear.
In a company, the best at driving the intelligence of others is generally the boss and supervisor, this status and division of labor, may also form a cognitive lead, for example, every time I doubt the AI tool, use a few times to find that it is not enough, my boss will tell me with personal experience and practice and results: If the AI tool does not give you what you want, it is not that it does not work, but that you asked wrong or not enough.
When I think about why he can always get the results he wants from the AI, the answer is twofold, one is that he has confidence in the intelligence and knowledge of the AI; The second is that he has rich experience in driving other intellects.
For most people who contribute their intellectual and physical strength, choosing, driving and utilizing other intelligence is a completely new proposition.
In large Internet companies represented by BAT, the more successful employees may be both smart and hardworking, and the most successful employees are often superimposed with an element of "good use of the company's various resources", which often includes intelligence.
This unevenness has existed since ancient times, and the unit of "horsepower" used in modern cars comes from the emergence of steam-powered equipment, and steam engine improver Watt determined that a horse can turn the mill turntable 144 times an hour, which translates to a horse can increase 75 kilograms of water by 1 meter per second, which is literally: the power of a horse.
Obviously, when walking, people's strength is far inferior to horses, but the carriages of the ancients and dignitaries had three or four horses pulling, and sometimes, people contributed horsepower, such as when carrying sedans for dignitaries.
Tesla Model S
Now the latest Tesla Model S Plaid can instantly burst out 1000 horsepower, which was the power of thousands of cavalry in ancient times, but now a car is only 1-5 people to serve.
American scholar and consultant Clay Shecky believes in "Cognitive Surplus: The Power of Free Time" that the emergence and prosperity of the Internet stems from the human cognitive surplus and sharing spirit. Because of his recognition of the book's views, Ma Huateng also wrote the preface to the Chinese edition of the book, believing that the concept of "cognitive surplus" is the dividend of the times for the development of platform-based Internet companies, and Facebook, Twitter, Wikipedia and Weibo are based on this.
In fact, the concepts of UGC (User Contributed Content) and PGC (Professional User Contributed Content) are in the same vein as "cognitive surplus", and when this wave of AI is named AIGC, it actually confirms that the "intellectual surplus" that AI can contribute is almost infinite.
Everyone has the opportunity and has unlimited "external intelligence".
The difference in perception of this new tool is much greater than the difference between walking and driving a Tesla Model S Playd, the former uses less than 0.1 horsepower, while the latter can use 1000 horsepower, the former travel speed is about 5KM/h, and the latter can reach 320KM/h.
In this wave of AI, the biggest beneficiary is NVIDIA, whose market value once exceeded $1 trillion not long ago, becoming the first chip company with a market value of more than one trillion, whether it is Intel in the PC wave or Qualcomm in the smartphone wave have failed to achieve this achievement.
▲ Nvidia founder & CEO Huang Jenxun
Not long ago, Nvidia founder & CEO Jensen Huang gave a speech at NTU, and at the end, he said this:
Whatever it is, go all out to pursue it like we did, run! Don't walk slowly.
Whether running for food or not being treated as food by others. You often don't know which situation you're in, but keep running no matter what.
Six or seven years ago, I saw some entrepreneurs in the field of AI in my circle of friends predicting that future human beings with the help of exobrains, exoskeletons and even mechanical implants and brain-computer interfaces will produce huge differences in intelligence and physical strength, as large as species and species differences. Biological fundamentalist humanity may have come to an end.
▲ The cover of a recent issue of Time
At the time, it was just a science fiction scene, but now it seems that the combination of ChatGPT and intelligent headsets has approached the prototype of the external brain.
There was no starting gun, but the race had already begun, and now the situation is that those running at the front began to shout at those crowded at the starting line: Run, if you don't run, you will be eaten!
Unlike the chicken soup we are accustomed to seeing in college speeches, "some people become famous at a young age, some people are late bloomers, don't be impatient, be yourself", Huang's exhortation is close to the "social Darwinist" view of the jungle.
As for why he says this, he believes that agile companies will use AI technology to improve their competitiveness, while those that fail to use AI will face decline. AI will change every kind of work, some jobs will be eliminated, and everyone in every organization will need to learn to take advantage of AI. That's why so many AI practitioners are shouting that "average is over," and people, organizations and organizations, will become more uneven because of AI.
Or take the car as an example, the car is not only a symbol of "physical surplus", with the power of 1000 horses, but also because of the computing power (NVIDIA Thor automatic driving chip has 2000 TFLOPS computing power), sensors (a combination of lidar, camera and millimeter-wave radar, far beyond the perception of the human eye) and algorithms and large models can also produce "intelligence surplus", cars with intelligent driving and without intelligent driving are already two types of products.
The act of driving with the steering wheel in hand is bound to be a thing of the past, just as I am now typing this article word by word on the keyboard.
If you think that people can be divided into "people who walk, people who drive themselves, and people who use automatic driving to the end", then the relationship between people and "AI intelligence" will also become the key to distinguishing people.
ChatGPT and Apple's Vision Pro headset have ushered in the era of "intellectual surplus", you don't have to have a Vision Pro, but you must have Vision.