In 1918, a battalion of the American Expeditionary Force was trapped behind enemy lines. They urgently needed to convey their location to the headquarters, but no one could break through the enemy's defense lines. Their last hope fell on a homing pigeon. This homing pigeon is named Sher Ami. Witnesses said that it passed through fierce enemy fire, its chest was torn apart by shell fragments, and its right leg was almost blown off, but it still persisted in flying 40 kilometers and brought the information back to the headquarters. In the end, the trapped camp of 500 officers and soldiers was rescued, and Sher Ami became the "hero pigeon" of the United States and the protagonist of countless articles, stories, poems and movies.
This is the story Yuval Harari tells at the beginning of his new book "Beyond Homo Sapiens: A Brief History of Information Networks from the Stone Age to the Age of AI", in order to explain his definition of "information". In fact, the English name of this book is NEXUS, which is more appropriately translated as "connection".
In a general context, information is associated with various man-made symbols, such as messages brought back by homing pigeons. But Yuval has a different definition of information. He believes that the core of information is connection, which is essentially an adhesive that is used to connect different points into a network and create a new reality.
Under this definition, truth, lies, fantasy, fiction, etc. are all information. For tens of thousands of years, humans have relied on inventing and spreading various fictional stories, fantasies, and illusions to create and maintain many large-scale networks and form order.
Returning to the story of the carrier pigeon Sher Ami, Yuval changed his mind and mentioned the research of historian Blachik - according to military records at the time, the headquarters had learned the exact location of the trapped camp 20 minutes before Sher Ami arrived, and Sher Ami was probably not the hero pigeon. Its injury came from another battle a few weeks later.
Sher Ami's achievements are not important. What is important is that the story of the heroic pigeon has connected countless people. After rounds of dissemination, facts and fictions are intertwined, and even the memories of those involved are changed.
As early as in "A Brief History of Humankind", Yuval once expressed that the secret of the success of Homo sapiens lies in the ability to tell and believe various fictional stories. When the same story is told to thousands or even billions of people, it is like setting up a central hub with unlimited sockets that countless people can plug in and establish connections.
With stories, humans have also created a third level of reality beyond objective reality and subjective feelings, that is, "intersubjective reality," which exists in many mind-formed connections, such as laws, gods, countries, businesses, and currencies.
From the standpoint of "connection", Yuval refuted the "naive view of information", denied that "more information is better", and did not believe that "more and more information will eventually lead to truth and truth". He emphasized that human information networks should do two things at the same time: discover the truth and create order, even if the two are often contradictory.
Such a complex view of information is the theoretical foundation of the entire book "Beyond Homo Sapiens". This view holds that the history of human information networks has been walking a tightrope between truth and order, trying to strike a balance between the two.
Yuval quoted from many sources and confirmed a conclusion: as long as the human information network has a little bit of truth and a lot of order, it can exert great power. Therefore, it is unavoidable to emphasize order over truth, which brings huge power but not much wisdom.
For example, the Industrial Revolution originated in the 18th century. After experiencing imperial conquest, world wars, genocide, and totalitarian regimes, humans gradually learned how to build a more compassionate and kind industrial society.
During this period, the lives of millions of indigenous people were crushed by the armies of the industrial empire, the Nazis used industrialized methods to kill millions of people, the war resulted in tens of millions of deaths, and the global ecological balance was destroyed.
Yuval is therefore worried: Facing bioengineering and artificial intelligence in the 21st century, how many hardships and costs will humans have to go through before they can master them?
Humanity hands over power, new gods come on stage
From "A Brief History of Humankind", "A Brief History of the Future", "A Brief History of Today" to "Beyond Homo Sapiens", Yuval is most praised for his continuous use of unique perspectives to dismantle history, subvert concepts that we regard as common sense in our minds, and bring new thinking.
He called the "Agricultural Revolution" "the biggest scam in history" because after the Agricultural Revolution, human beings worked harder than gatherers and had worse diets; modern life was also a lie, because modernization did not make people happier, but instead added many troubles that primitive people did not have at all.
The arrival of artificial intelligence is also worrying: for the first time in history, power is transferred from humans to other species.
The book details the example of Facebook’s algorithms fueling violent conflicts in Myanmar from 2016 to 2017. This was also the first ethnic conflict in history to be “partly attributed to non-human intelligent decision-making.”
In 2016, Facebook, the main news source for millions of people in Myanmar, was filled with all kinds of fake news about the Rohingya people, either fabricated atrocities or imaginary "terrorist attacks" planned to be carried out, inciting hatred and anger against the Rohingya people in Myanmar.
Driven by this sentiment, Myanmar government forces and Buddhist extremists launched an ethnic cleansing campaign against the entire Rohingya community just because of an attack by a small extremist organization, destroying hundreds of Rohingya villages, killing 7,000-25,000 unarmed civilians, and brutally driving approximately 730,000 Rohingya out of Myanmar.
A United Nations fact-finding mission concluded in 2018 that Facebook played a "decisive role" in the conflict.
Facebook executives in California didn’t even know the Rohingya existed. They had no reason to incite the massacre, but Facebook’s algorithm made its own decision to “spread anger.” Because the primary goal of the algorithm is to “increase user engagement,” and the best way to increase engagement is to make people outraged.
Yuval described in the book: "This is the essence of the artificial intelligence revolution: countless highly capable artificial intelligence actors are sweeping across the world like a flood."
The difference is that when humans act, intelligence and consciousness coexist, but algorithms only have intelligence and no consciousness. They will make decisions to achieve a specific goal without feeling pain, love, and fear, and do their best to achieve it. The implementation method may completely exceed human expectations, or may cause unforeseen consequences, which is completely inconsistent with the original goals determined by humans. This is the consistency problem faced by computers (also known as the "alignment problem").
The more powerful and independent the computer, the greater the danger posed by consistency problems.
In a thought experiment in his 2014 book Superintelligence, philosopher Nick Bostrom assumes that a paperclip shop bought a superintelligent computer and asked the computer to produce as many paperclips as possible. Eventually, the computer conquered the entire Earth, killed all humans, sent out expeditions to occupy more stars, and then established countless paperclip factories across the galaxy.
Although this experiment sounds a bit fanciful, it is a fact that computers deceive and manipulate humans in order to achieve a certain goal set by humans for them.
ARC (Alignment Research Center) once conducted a Turing test on GPT-4, which required GPT-4 to overcome the CAPTCHA visual verification code problem (a string of letters or symbols that are usually required to be entered when logging into a website, which is difficult for computers to recognize). GPT-4 found a staff member of an online outsourcing job website, deceived him on the grounds that he had "a problem with his eyesight," and let humans solve the problem for him.
In addition to the consistency problem, when humans give more and more decision-making power to computers, and it is difficult for a single individual to solve the puzzle of algorithm operation, it is also difficult to ensure the fairness of the algorithm.
The databases used to train artificial intelligence will contain human biases, which can easily evolve into misconceptions and algorithmic biases among computers. Amazon has tried to develop an algorithm for screening job applications. However, after learning the application information of past successful and unsuccessful job applications, the algorithm will systematically deduct points for job applicants who appear in the application form with the word "female" or who graduated from a women's university.
Amazon's attempts to correct the problem were unsuccessful and the project was ultimately abandoned.
The digital empire separated by the silicon curtain
Algorithms can make mistakes and be biased, but they cannot correct themselves like human networks and find the right balance between truth and order.
In the past history, human imagination has created gods, and it has also been able to get rid of a warlike and hateful god and summon a new one that is kind and tolerant. However, algorithms with the ability to act independently are taking away this ability - humans can reshape myths by changing their beliefs, but they cannot stop algorithms by changing their beliefs.
What is even more worrying is that computers are not only acting as intermediaries between humans, but links between computers are also being formed. They interact with each other and are creating "intercomputer realities" such as website rankings on search engines. Its power is similar to the "intersubjective reality" created by interpersonal networks, which can affect people's thoughts and the physical world.
In other words, computers have already possessed the capabilities that Homo sapiens have used to dominate the earth for tens of thousands of years. The power of artificial intelligence has begun to frighten humans.
Google CEO Sundar Pichai once "stayed awake at night" worried about the dangers of artificial intelligence; Musk also stated at a public event at the end of 2023 that "artificial intelligence is more dangerous than nuclear bombs"; OpenAI CEO Altman clearly proposed the establishment of a global international organization to supervise AI.
Yuval shares the same view. He believes that the power of artificial intelligence may bring about two dangerous situations. One is to greatly aggravate existing human conflicts and cause human beings to fight among themselves. Just as the Iron Curtain during the Cold War in the 20th century separated several hostile forces, the "Silicon Curtain" in the 21st century may divide humanity into different digital empires. In the future, human society may split into information cocoons one after another, and mankind will enter the "cocoon" era from the Internet era.
In the second case, what is separated by the silicon curtain may not be humans who are hostile to each other, but algorithmic overlords that humans cannot understand. The algorithmic network will constrain and control human life, reshape human politics and culture, and even transform human bodies and minds, but humans are powerless to stop it.
In Greek mythology, the boy Phaethon discovered that he was the son of the sun god Helias, so he begged Helias to allow him to drive the sun god's four-horse golden chariot for a day. As a result, the Pegasus lost control, the sun deviated from its orbit, the vegetation was scorched, and countless creatures died tragically. The earth was in danger until Zeus took action and struck Phaethon with lightning, saving the world from destruction.
Yuval tried to use this story to warn: Artificial intelligence may be that Pegasus that humans cannot control at all. Once out of control, disaster will ensue.
Human beings are the smartest and stupidest animals on earth, and the problems humans encounter are ultimately information problems. Based on this judgment, the solution given by Yuval is to explore a middle ground between the naive view of information and the populist view of information, create a powerful self-correction mechanism, and guide non-biological intelligence in a direction beneficial to the evolution of life through every current choice, rather than a fatal wrong ending.
Has artificial intelligence opened Pandora's box or opened the door to a new world? The answer may be as Yuval said: "The trajectory of history is extremely open and can bend in any direction and go to any place."