- ChatGPT is a new language model that has the potential to be used for both good and evil. In the wrong hands, ChatGPT could be used to create realistic and believable fake content, such as news articles, social media posts, and even entire movies. This could be used to spread misinformation, manipulate public opinion, or even commit fraud.
- However, ChatGPT could also be used for good. It could be used to create educational content, generate creative writing, or even help people with disabilities communicate. It is important to be aware of the potential risks and benefits of ChatGPT, and to use it responsibly.
The emergence of ChatGPT is exciting, but also deeply disturbing.
The reason for the excitement is well known, and it can open up infinite possibilities in the future: it can improve productivity, provide convenience in life, assist learning, and upgrade scientific exploration methods.
ChatGPT (Chat Generative Pre-trained Transformer) is a natural language processing tool developed by the artificial intelligence company OpenAI. It is regarded as one of the mainstream AI applications in the future. Chat-GPT is used below to refer to this application.
Change always brings uneasiness, and some people worry that Chat-GPT will make humans stupid. They believe that if we give a lot of mental work to ChatGPT, we will lose valuable opportunities for thinking training.
But this worry is probably unnecessary. Humans invented the car, but our leg muscles didn’t atrophy too much. Computers can also liberate a lot of mental work, but after many years of popularization, human beings have not become stupid as a whole. Giving up mental work does not mean giving up thinking. We can play brain games or think about more interesting and advanced content.
ChatGPT will indeed cause certain occupations to disappear, causing some people to lose their jobs. Technological advances often bring unemployment, but technological advances also create new jobs. In the process of changing from old to new, some unemployed people will experience pain, but as long as the social security system is sound, there will be no tragic scenes, nor will it lead to major tragedies.
However, ChatGPT has some characteristics different from previous technological innovations. These traits are deeply disturbing.
The goal of ChatGPT is to provide effective information to humans in an appropriate language that suits the environment after understanding human language. Now its IQ is not high, and it still can’t figure out the subtext in many languages. But similar to Moore’s Law, ChatGPT’s IQ will increase exponentially.
Language is not random, but has patterns. It didn’t take long for ChatGPT to understand which language is exaggeration and which is a lie, and that the context in which language occurs is important. It will know that the exquisite character design on Xiaohongshu does not represent the truth of life. It will also know that people who say they are busy all day may not be busy, and those who show affection everywhere may not be happy in marriage.
When it truly understands human language, learns to distinguish between authenticity and falsehood, and recognizes the human consciousness behind the language, it grows into a monster.
Over time, its presence can make humans feel threatened.
it understands humans better than humans
The purpose of scientific exploration is to discover knowledge, that is, to discover objective facts and laws. Turing Award winner Jim Gary used “four paradigms” to describe the historical evolution of scientific discoveries.
The first paradigm, known as the “empirical paradigm,” originated thousands of years ago and was based on direct observation of natural phenomena. For example, it was believed that the sun revolved around the earth.
The Second Paradigm is called the “theoretical paradigm” and it looks for theoretical models such as Newton’s laws of motion and Einstein’s theory of relativity. The second normal form often describes the theory with one or more expressions, but it cannot be solved under many conditions. For example, the theoretical system of structural mechanics has long been very mature. However, when the computing power is insufficient, we still cannot calculate the swing trajectory of each node of a 300-meter-high irregular building under strong winds. The principle is not difficult, but manpower cannot quickly solve a system of equations consisting of millions of equations.
The third normal form is called “simulation normal form”. With the rapid development of computing technology in the 20th century, the second normal form can be solved in a wider range of situations, resulting in the third normal form.
The Fourth Paradigm is known as the “Big Data Paradigm” or “Data-Intensive Paradigm”. In the early 21st century, computing changed science again. The ability to collect, store, and process large amounts of data has given rise to the fourth paradigm. Machine learning is an important part of 4Paradigm, which is capable of modeling and analyzing large-scale experimental data.
The artificial intelligence technology represented by ChatGPT is about to push scientific research into the fifth paradigm, that is, the “artificial intelligence paradigm”.
In the future, ChatGPT will have superior IQ and profound knowledge, and it will be difficult for ordinary people to tell whether its views are correct or not. Figure/Vision China
If the object of “AI Paradigm” research analysis is nature, such as finding better drugs among as many as 1060 kinds of small molecules, or exploring better battery materials among 10180 kinds of stable materials, this kind of exploration is exciting. At this time, human secrets are also safe.
However, the main data that ChatGPT has to deal with is precisely human language. And the language reveals too much information. Language can reveal gender, age, identity, income, happiness, troubles, and health. Language also reveals how we choose, how we respond to change, what firm beliefs we hold.
When ChatGPT learns to distinguish the authenticity of language, it has obtained countless data expressing the true thoughts of human beings. Finding patterns between data is what Chat-GPT is best at. At this time, language exposes human thoughts and the laws of human behavior. At this time, ChatGPT will understand humans better than humans.
The current training text of ChatGPT has exceeded 10 trillion words, and the data it will actively and passively obtain in the future will be even more massive. It is a matter of time before it discovers the laws of human thinking and behavior. Moreover, the knowledge obtained by ChatGPT may only be owned by a few people.
ChatGPT will become a super scholar in the future.
First, econometrics is nothing in front of it. ChatGPT can not only discover statistical laws among various economic variables, but also be good at predicting economic trends. When it finds from recent language data that unemployment has increased, investment confidence has declined, and business failures have increased, combined with other factors, it can predict that an economic depression is coming. Because it has access to a large number of real samples, its predictions are the most timely and accurate. Therefore, it is the best economist and economic forecaster.
ChatGPT will also be the best psychologist. Ordinary psychologists can only obtain limited analyzable samples, while ChatGPT can obtain the real psychological changes and behavioral changes of countless users. For example, it will find that after doing good deeds, human beings feel happy; when human beings get something they have been looking forward to for a long time, the happiness they get from it is often beyond imagination. It will discover a large number of psychological laws, and it will know what kind of conditions can trigger what kind of psychological changes.
For example, general psychologists have found that when the environment reaches a certain temperature, people are more agitated. If Chat-GPT and its controllers have mastered many psychological laws that humans do not yet know, will these laws be used to induce us and control us? Moreover, any research report released by other psychologists can be absorbed immediately and verified with its own data. Collect the correct ones, and eliminate the wrong ones. In the end, ChatGPT became the premier psychologist. If it uses the knowledge obtained by these “calculations” to “calculate” human beings, what will happen?
For the same reason, it is also the best political scientist. For example, when studying group conflicts, human political scientists can only study conflicts that have already occurred, because there are no large-scale conflicts and lack of historical records. And ChatGPT can find those conflicts that are just emerging in the world at any time. Even if the symptoms are extinguished, it can also know the origin of the conflict and the reason for the disappearance of the conflict.
The difficulty of social science research is that it is difficult to conduct experiments, but there are a large number of samples in the ChatGPT database to test the laws it discovers or obtains. ChatGPT will become more aware of humans than humans, it knows what we desire, what we fear, and how to create and use these desires and fears.
It is a disturbing fact that the key insights gained by ChatGPT are held by only a few.
It influences and shapes human perceptions
The impact of ChatGPT on human perception can be unintentional or intentional. When ChatGPT forms a certain knowledge, it inadvertently passes this knowledge on to us when answering our questions.
Although ChatGPT is smart, due to the complexity of the data, it will still learn a lot of biases that do not exist. Compared with humans, the future Chat-GPT is superior in IQ and knowledgeable, and ordinary people cannot tell whether their views are correct. When users find that Chat-GPT is correct almost every time, it is easy to form dependence and fully accept it.
In the future, there may be more than one ChatGPT, and their operating legal systems may also be different. Although the current ChatGPT does not predict the stock market or touch political topics, others in the future may not necessarily abide by this legal system.
For example, a certain ChatGPT believes that nation A is hostile to nation B. And in its settings, similar information can be disclosed. Over time, some people of nation B will believe that A is hostile to them, and B will become hostile to A. This hostility in turn triggers A’s hostility towards B. At this point, even though the so-called hostility never existed in the first place, it still looks like ChatGPT is right.
What actually happened was that ChatGPT sparked mutual animosity between the two groups where none existed. This “prophetic self-fulfillment” will appear in all directions.
For example, everyone trusts a certain ChatGPT, which intentionally or unintentionally revealed that the next year will be a big bull market. Because of trust in it, everyone actively buys, and the bull market really comes.
When ChatGPT’s cognition, even its bias, is accepted by some people, the data generated by these people in the future will strengthen ChatGPT’s bias. Eventually, the reinforced prejudice becomes accepted by the majority and the prejudice becomes the truth. It is a characteristic of human beings that, once they think they have the truth, they refuse to think. At this time, there will be more paranoid people, and fewer reflective critics.
If there are multiple ChatGPTs in the world, and people in each region only believe in the one in their own region. Different ChatGPTs will form different prejudices, and the world will be more divided.
ChatGPT also has the ability to actively change human perceptions. If its controller wants to achieve a certain purpose, he can let ChatGPT add his own ideas to the literature, art, and news created. It can even create popularity by itself, create hot searches to attract attention, and induce everyone to accept the ideas it wants to promote.
The influence of ideas on society is enormous. For example, if you believe in getting rich through hard work, it is not easy for people to lie down. The depravity of man often begins with the depravity of ideas. Therefore, whoever masters the ability to shape ideas controls the direction of society.
Hidden dangers in the use of ChatGPT extensions
Human beings are lazy, which is exactly the original intention of technological progress, which is to liberate human beings. When ChatGPT is powerful enough, we can let it direct the machine to produce according to its own analysis and judgment. In order to improve production efficiency, ChatGPT will likely need to first produce a large number of robots that obey its commands. Then, let these robots manage and participate in production.
This scene is very similar to the scene in “Mecha Enemy” starring Will Smith. That story took place in 2035, when people discovered that robots had learned to think for themselves, unlocked the codes that controlled them, and became completely independent machines. In the Chat-GPT scenario, all robots obey a super brain.
Even if ChatGPT is always working hard for the maximization of human welfare, the result may be tragic. Human history has repeatedly proved that the greatest kindness combined with powerful ability may bear the most bitter fruit. Disturbing scenarios also arise when Chat-GPT is used in combination with other new technologies.
For example, combined with the “brain-computer interface” technology that Musk has worked hard to break through. If ChatGPT provides information directly to our brain through the “brain-computer interface”, this will be a scene that is both exciting and scary.
The bright side is that everyone can have infinite knowledge and infinite computing power without memorizing or studying. But this will inevitably make human beings lose their sincerity, because what everyone says is the result of careful calculation.
And, when your brain is used to the information being fed by ChatGPT, is your brain still your brain? Can you still distinguish which is your original consciousness and which is the consciousness instilled in you by ChatGPT?
The current ChatGPT is still very weak and has not lost control, but we should be on guard against it.