After all, where is the "scary" of the heliocentric? In fact, it only reveals a fact: human beings are not the center of the universe, and have not received special care from the Creator. Therefore, some people say that ChatGPT is the heliocentric theory of the AI era.
On July 13, 2023, Musk announced his ambitions in the field of AI: an AI company called xAI was born with the goal of "understanding the nature of the universe."
The original intention of human beings to develop artificial intelligence (AI) is to create a technology that can help us become faster, better and stronger. But obviously, the great success of ChatGPT in the near future has ignited the concerns from the heart of human beings - its (we dare not call him or her) rapid growth, its intelligence and eagerness to learn, especially its "autonomous" learning and discrimination ability, as if a kitten kept at home suddenly showed its lion's claws.
Musk's slogan for AI is to understand the universe-this is to abandon the original intention of "serving mankind" and to snatch the intelligence halo that originally belonged only to humans.
As early as the 16th century, human beings were deprived of their "C position" by science, technology and facts, causing a huge panic, and it was Copernicus who brought this panic. After all, where is the "scary" of the heliocentric? In fact, it only reveals a fact: human beings are not the center of the universe, and have not received special care from the Creator. Therefore, some people say that ChatGPT is the heliocentric theory of the AI era. From then on, the subject of the sentence "understanding the nature of the universe" may no longer be only human beings, and human beings are no longer the center of intelligence...
1
understand the world?
It is meaningless to talk about "understanding the world" without talking about intelligence.
The intelligence enlightenment of "silicon man" can be said to come from the Bayesian method. The Bayesian method is based on the probability of different data observed under a given assumption, the prior probability and the data itself (acquisition of information), and then combines the unknown prior information with the existing sample information (generates a sample) to obtain the posterior information, and then calculates according to the Bayesian formula (training optimization), and draws a method or conclusion for judging the unknown. When you cannot accurately know the essence of a thing, you can only rely on experience to judge the probability of its essential attributes, that is, to judge what it should roughly look like. So, is our "gut feeling" worth anything? Musk said to "understand the nature of the universe", but what is the truth of the essence? What does it mean to understand? Bayesian methods may be able to answer these questions.
Does the truth lie? "
The learning experience of ChatGPT should be very familiar to "Carbon Man": Chat GPT is based on a large model. You can say that it is an infinite neural network composed of countless brains. Massive text samples from the Internet, books and other human creations all over the world can be fed to this huge network; then, this neural network is trained to generate text that is "similar" to the massive information in the world. Moreover, it can start from a "hint" and continue to generate "text similar to its training data". For example, when you give it a letter C, it will infer that you may want Cat or Chat.
In fact, ChatGPT will continue to improve its language based on massive amounts of information. It is by far the AI with the strongest chatting ability, but you will feel that it is "ghosting" from time to time. For example, a child who is just learning to speak, when he/she sees the current news for the first time and imitates saying the word "murderer", will be corrected by parents "that is called 'murderer'". But after that, the kid can't be wrong. ChatGPT can not only produce coherent human language, but also follow the "hint" to say "that person is the murderer" according to the content that has been "read". But ChatGPT didn't really judge who the murderer was, but just said something that "sounds pretty reliable".
"Carbon Man" does not have countless brains, and does not have unlimited time to swallow massive amounts of information. If the "silicon man" can make up for these weaknesses and become an increasingly excellent student, many of the intelligent behaviors originally exclusive to the "carbon man", such as writing, design, data processing, etc., will appear weak in an instant...
To be honest, have you ever asked ChatGPT: who is the most beautiful person in the world?
2
Where is the way out for "Carbon Man"?
Human beings are not limited by the evolution of technology, but only by the goals they set themselves.
——S. Wolfram
Understanding the world is an innate magical ability of human beings. Children aged three or four can answer who has more stars in the sky or children in the kindergarten class. But the real world is very complicated, and the problem will also become complicated: what is the approximate order of the number of stars? The child may not be able to answer it, but he/she may be confused about "millions", "hundreds of millions"... any large number, but he/she will not say "tens" and "hundreds". The child is very smart! Most of us are "smart" at this stage: collecting samples, calculating probabilities, optimizing experience, and forming judgments. So far, the intelligence of AI may also stay at this stage.
However, why are some people "smarter"? For example, how did people like Newton, Galileo, and Einstein figure out the concepts of "acceleration", "gravity", and "space-time"? Why did they invent something as powerful as calculus? What is the difference between the heads of these human wizards and the heads of "Muggles" like you and me? Is the gap between them and us the ultimate gap between "carbon man" and "silicon man"?
Mathematician Mikael Launay tells a story. He and a colleague set a date for their next meeting, and he heard her murmur, "Today is April 20, and 14 days later is the 34th, which is 34 minus 30—May 4th." It dawned on him that the colleague had invented a date that didn't exist: April 34th. There is more than one way to calculate dates. For example, many people think that April 20th will be April 30th after 10 days, so May 1st will be 11 days later, and May 4th will be 14 days later. Because April 30th is followed by May 1st, the transition of the month interrupts the natural growth of numbers-instead, the natural arithmetic thinking must be deliberately interrupted to figure things out? Instead, fellow mathematicians figured out how, using something that didn't exist.
Lonay was inspired: the universe is out there, how do we understand it? As a mathematician, Luo Nai believes that mathematics is the "first tool": mathematics is like an umbrella, opening up a "virtual" world, allowing us to walk in the real "heavy rain", and when we get the method in this virtual time and space, we can bring the results back to reality for verification, and then solve problems. So, does this approach work when it comes to bigger, more complex problems?
This is not just a way of scientific research, but a way of understanding everything in the universe. This is how Newton got the law of universal gravitation. Hooke also saw gravitation, but he failed to use mathematical methods to verify and explain it. This is why Newton believed that he defeated Hooke—he stood on the shoulders of giants.
3
Human scientific thinking is good at making two compasses: one is called "practical" and the other is called "elegant".
Just like the mathematician mentioned above, she invented "April 34" out of thin air to derive the date. Not only is it simple and convenient, but more importantly, this method is more in line with basic thinking habits and can quickly help her sort out the situation and solve problems. This way of thinking is natural and typical for a person with mathematical training.
Elegant and practical, mathematics is the scientific soul of AI
Thinking, understanding, exploring, and creating "non-existence" are the great qualities of human beings, and they are also "imported goods" that AI (at least temporarily) is difficult to obtain. What does not exist is abstract: the "concrete" in reality can be turned into a kind of "idea" in a thinking model (such as a mathematical model), so that abstraction becomes something that can be imagined and seems to exist, and it just appears in the middle of thinking out of thin air. In the final analysis, the way out for "Carbon Man" is to "move forward"—those who change the world are those who dare to be "different" and innovative.
Looking back at the history of science, the progress humans have made over the centuries is dizzying. However, once we think of all that we don’t know, how many unknown and “non-existent” things in this world—no matter how far away they are from us today, or they may have happened, right under the eyes of human beings, but we still can’t understand them through our own perception and imagination—“Carbon Man” will feel uneasy.
But "Silicon Man" will not, it will not be anxious about "non-existence". But one day, it may really understand like we do today, and find it very interesting to understand the world. Right, AI?