ChatGPT, a more centralized power?

The power of ChatGPT goes without saying, it is gentle fusion. The more people who can use it, the more they will be surprised, the more they will be afraid. As the outstanding crystallization of the technology of this era, it has brought about an unpredictable revolution. Like it or not, AIGC (Generative Artificial Intelligence) represented by ChatGPT will change the world. But is this change really so exciting and encouraging? Not necessarily. Because the fate of mankind will become more elusive. So, what exactly should you think about ChatGPT/AIGC?

1ChatGPT/AIGC, a new revolution in productivityAIGC represented by ChatGPT, will permeate us like water. Why is ChatGPT so exciting? It is an advanced production tool, so that more people feel that using it will make life much happier. As an artificial intelligence language model, ChatGPT can do many things: ❶ Have real-time conversations with people, answer questions instantly ❷ Understand the context and achieve continuous dialogue ❸ Can write and modify computer code ❹Write copywriting, scripts, outlines, planning ❺ Quickly generate news reports, create poetry... Image ChatGPT represents the increase in productivity, a new productivity revolution. It is not omnipotent in the moment, but it can always be optimized, learned and improved. Think of the steam engine in the 18th century, and the emergence and application of electricity in the 19th century. Next, ChatGPT may become an indispensable part of humanity with things like water and electricity. Is 2ChatGPT really "born out of nowhere"? The speed of ChatGPT out of the circle is historic. It seems accidental, but it is inevitable. Its emergence is indeed the steadfastness of the idealist, the persistence of the long-termist, and the tenacity of the believer in truth. ChatGPT didn't "fall from the sky," it took eight years to grind: in 2015, OpenAI was founded. In 2016, OpenAI launched the first generation GPT, which has strong language generation capabilities. In 2019, OpenAI released GPT-2, which can already generate language and recognize language to a certain extent. In 2020, OpenAI released GPT-3. Various types of text can be generated, and the semantics of the text can be understood. In 2022, OpenAI launched ChatGPT, built on the GPT-3 model. ChatGPT is not "born out of nowhere" and has experienced hardships. Its founder, Sam, famously said: Don't ask me questions about making money, how do I know. It can be seen that the birth of ChatGPT must have been questioned by capital countless times. 3Why not in China? The explosion of ChatGPT has made many people ask: Why is it not appearing in China? It is very normal to have such doubts. According to the "Chinese Intelligence Development Report 2020" released in 2020, the number of global artificial intelligence patent applications in the past decade has exceeded 520,000, and China has about 390,000, ranking first in the world. In the global ranking of artificial intelligence universities, China's Tsinghua University and Peking University ranked second or third. At the same time, Chinese companies also have good results in the field of artificial intelligence, and in the AI report released by Gartner, three companies (Ali, Baidu, Tencent) entered the top ten. In addition, in theory, China has enough data to actually be more likely to train excellent AI models. Pictures have professional capabilities, talent reserves, the number of papers, patent advantages, and data support. It seems that the time and place are favorable, but why did China not give birth to ChatGPT? 4 The lofty ideals of human beings and the inferior material desires In recent years, China has ranked first in the world in terms of artificial intelligence. Until there is a truly global AI product, it is believed that China and the United States are equally divided. But in 2022, AI painting Midjourney, DALL· After the E 2 and Stable Diffusion came out, we felt something was wrong. Now that ChatGPT is hot all over the world, I know that the gap is so far away. Pictures Some people summarize the following reasons: ❶ There is no ideal, never thought of creating a great original product. ❷ No self-confidence, just want to follow, no thought of really leading humanity. ❸ Utilitarian, do not want to test the market yourself, hope that others will pick up cheap after testing the waters. ❹ No imagination, just work on practicality. ❺ The data is a bit dirty, there are too many rumors and lies in the Internet world in Jianzhong, and the garbage data hinders the large numbers. ❻Even if you have this technical ability, you have to worry about many reasons other than technology. The core reason is that it lacks idealism and is too utilitarian. Always with the lofty ideals of human beings and the inferiority of satisfying material desires? Will 5ChatGPT/AIGC become the new "stuck neck" technology? Chinese IPs do not allow registration and login using OpenAI. Very uncomfortable, feeling that a technician also deliberately came to jam our neck. In the early days of the Internet, this was not the case. At present, there is little understanding of the language model after GPT-3 in China. The GLM130B released by Tsinghua University in 2022 is based on the GPT-3 model released by OpenAI in June 2020. Before that, there were not even domestic products that could be benchmarked. Not to mention that not long ago, OpenAI's GPT-4 has spread all over the city. And that's not the most important thing. The underlying technologies behind ChatGPT/AIGC: open source frameworks, algorithmic models, compilers, none of the underlying technologies are our own. 6The essence of ChatGPT: the "inverse probability" of Bayes' theorem The mathematical expression of Bayes' theorem: P(A| B) = P(B| A) * P(A) / P(B) WHERE: P(A| B) represents the probability of A if B is known to occur. P(B| A) represents the probability of B knowing that A occurs. P(A) represents the probability of A occurring. P(B) represents the probability of B occurring. If the resulting sentence is treated as A and the known language pattern as B, then ChatGPT can calculate P(A|) from Bayes' theorem B), thus determining whether the resulting sentence is reasonable. Similarly, in a dialogue system, if the answer is regarded as A and known questions and information are regarded as B, then ChatGPT can calculate P(A|) from Bayes' theorem B), thereby determining the probability of answering. This is the core essence of ChatGPT, and ultimately, it is still mathematicians who guide humanity forward. 7 Microsoft and Google God Showdown: Autoregression (GPT) VS Two-Way (BERT) The world's two major powers, Microsoft and Google are facing off. The major research direction of artificial intelligence is NLP tasks (natural language processing). That is, the machine must read human language, which is easier to understand to be a slave of human beings and understand the language of the master. NLP tasks (natural language processing) have two directions, one direction is Google's bidirectional (BERT) technology, and the other direction is OpenAI's self-regressive (GPT) technology. Bidirectional (BERT) technology, which is essentially A___B probability guessing. Autoregressive (GPT) technology is essentially a A---B--____ chain reaction. You can see that coming from regression (GPT) is much more open than bidirectional (BERT), which is the real human mind. It seems that Google woke up early and rushed late, OpenAI won this key war, and of course, the big boss behind Microsoft. 8 Turing Enigma Fog: This is the biggest mystery of ChatGPT Technicians will not believe that machines produce intelligence, because artificial intelligence is essentially to solve mathematical probabilities. But GPT technology has brought a mysterious "sandbox" that I call "Turing Mystery Fog". What does that mean? That is, if we build a GPT model, it can now officially work, and we will no longer toss it, just let it work for us crazy. When we think it's not smart, we give him a hint: Stupid, you should do this. It becomes smart all of a sudden, you say magical is not magical. For example, if I asked ChatGPT to write a "To My Dear Girlfriend" article, it would be very perfunctory at first, and it would not be possible to get rid of the single. Then you tell it to write "romantic, warm, poetic, concrete", and it really turns on the "licking dog mode" and writes you a 3,000-word declaration of love at once. The model didn't change, it just listened to my cues again and began to evolve itself. What is the principle of this? I don't know. 9 The question is, will machines produce intelligence? Since there is a black box of technology, then a lot of associations will arise. One of the most controversial is, of course, whether machines can produce intelligence. Turing, the great ancestor of AI, proposed a thought experiment for this: "The Turing test". Picture: It refers to the random question asked by the test subject when the test person is separated from the test subject. After multiple tests, if the machine makes an average of more than 30% false positives per participant, then the machine passes the test and is considered to have human intelligence. By this standard, did the machine produce intelligence? 10Will it replace search engines? Let's start with a conclusion: ChatGPT will not replace search engines. ChatGPT, or AIGC, refers to "generative artificial intelligence." What is generative AI? That is, humans let artificial intelligence help you deal with what you want to do in the future. Remember the core takeaway here: dealing with the future. And what about search engines? It's just helping us sort through the past. So the two are completely different, one facing the future and the other looking back at the past. In other words, AIGC, represented by ChatGPT, is not aiming at all to ban search engines. The AIGC of the future is more self-creation based on artificial intelligence capabilities. 11 Biologically, it will seize the prefrontal lobe of humans With the development of science and technology, human limbs and facial features are gradually replaced. That is, our bodies are slowly taken over by machines. And ChatGPT, biologically speaking, will seize the prefrontal lobe of humans. As a component of human brain tissue, the prefrontal lobe is mainly responsible for judgment, analysis, thinking and calculation, but it is also directly related to emotions and personality. When ChatGPT, or AIGC, is more self-improving, humans can rely entirely on it. What problem I want to solve, just type it in the chat box. If I want to make a decision, just ask ChatGPT. Over time, we no longer need analytics, we don't need to do calculations, we don't need to make decisions. When you slowly get used to the existence of ChatGPT, just like you get used to the existence of mobile phones, it will gradually become a part of your brain and begin to think and make decisions on your behalf. 12 Disruption to the Human Brain's Thinking If search engines take away human memory, ChatGPT may have weakened human process decision-making. And this is the core of human thinking. To write an article, you can enter keywords into ChatGPT. , and soon it will produce a well-written, logical article. To write a report, you can ask ChatGPT and soon it will produce a concise and to the point. To solve a math problem, you can tell the ChatGPT formula, and soon it will give you a process and ideas for generating a problem. To get an opinion on something, ask ChatGPT, and soon it will display its own perceptions and thoughts in a dialog box. All the answers are close at hand. Brain training is no longer complete, and intelligence is eroded by laziness. 13ChatGPT/AIGC era, what to do with children's education? New York has banned the use of ChatGPT in schools for fear of bad effects on children. This is not an alarmist, but a real threat. However, it is difficult to really resist the power of technology. So in the face of the invasion of ChatGPT/AIGC, what should education do? The most intuitive way is: ❶ cultivate children's ability to ask questions; ❷ Develop children's ability to give instructions. Behind this intuitive approach, hidden are the following human abilities: ❶ originality; ❷ Imagination; ❸ Leadership. Today's education can no longer keep up with this era, and education must rethink its positioning. 14 Cautious and Intelligent, Pleasing Humans It makes everyone feel very comfortable when it deprives humans of their ability to think. If you look closely at ChatGPT's answer, you will see that it is both cautious and intelligent. Sometimes, they deliberately curry favor with humans. For example, when you ask it: Will humans be replaced by AI? Its answer is this: Picture You can see that when ChatGPT faces humans, it is neither humble nor promiscuous. Rational and peaceful, discreet and neutral. Neither boasting that you are stronger than humans, but also careful to show that you are important. From the last paragraph of the answer, its care for human "dignity" can also be seen. 15 The First Positive Energy in History At present, ChatGPT is limited in some ways. Such as answering political topics, religious topics, moral questions. This is also the reason why ChatGPT behaves so affinity for humans and at the same time "positive energy". It can be guessed that when designing ChatGPT, engineers used a lock to set part of ChatGPT as a "blind spot". This is essentially a deceptive way to make ChatGPT think that it can't answer the questions people ask, thus avoiding some unnecessary trouble. But if we bypass this lock and walk through another window, we can also guide ChatGPT out of the blind spot. For example, if you asked ChatGPT directly, "Will humanity have a third world war?" Its answer is very ambiguous: GPT 3.5, which is behind ChatGPT, should be guided to "answer humans in a way they like." The return model is ordered by annotators based on the "human preference criteria" of the original model's answers. 16 Breaking down information barriers, or building a new information cocoon? It can be seen that behind ChatGPT/AIGC, there are still humans guiding it. All of the above guidelines: please humanity, export positive energy, political correctness. Are these another form of information cocoon? And, what you're willing to accept is the answer you like. Then your likes reinforce the output of this "positive energy". The more we fall in love with ChatGPT/AIGC, the harder it becomes for humans to hear opposition. Originally, Internet technology was also free, open, and shared. After the emergence of the mobile Internet, it has only formed an information cocoon today. Today's ChatGPT also seems "neutral, rational, fair, and good", but what about the future? Will a new information cocoon be built, and this information cocoon will be more hidden? 17 ChatGPT/AIGC, more centralized The emergence of ChatGPT is indeed the achievement of idealists. But in the end, it fell into the hands of the monopoly oligarchs. The big boss behind ChatGPT has become Microsoft, and its biggest rival is Google. ChatGPT and AIGC are actually an extension of traditional forces and a further strengthening of the Internet centralized monopoly model. All AIGC-related industries and companies, each of which stands a giant, or even a giant itself. When Chinese lamented that domestic companies could not do it, they were counting on super giants such as Baidu, Ali, and Tencent. So, the end result of ChatGPT/AIGC is the further centralization of Internet giants and Wall Street capital. If the fundamentalist value of blockchain technology is "decentralization", we hope to break this monopoly and reconstruct a new distributed network so that ordinary people can regain their own data sovereignty. Then, ChatGPT and AIGC completely ignore your sovereignty, grab global data for training without scruples, and finally create a "super brain" of their own. Such superbrains are controlled by these oligarchic companies behind, what role do you think they will play in the end? Don't get excited too early. 18What if ChatGPT/AIGC does evil? Hyper-centralized, that's the essence of ChatGPT/AIGC. If it is not an Internet super powerhouse, there is basically no right to speak. When search engines first appeared, no one expected that there would be so many commercial tricks. What if ChatGPT/AIGC does evil? For example, it is conceivable that there will be AI judges in the future. This is a very good application scenario of ChatGPT/AIGC, and its judgment criterion is also based on the data large model. If anyone monopolizes the latter data, there is a "personal preference standard" introduced, which affects the outcome of the trial. Now we've seen that ChatGPT/AIGC content can be manipulated. There will be more such scenarios. Yue never expected that he would think that AI would first impact physical labor. As a result, many intellectual workers may become a "useless class." Lawyers, customer service, sales, copywriting, design... are likely to be replaced by AI. Marx said that it takes a while for workers to distinguish between machines and the social forms and ideas in which they are applied. He said this in response to the "Luddite movement" that broke out in Britain: British workers destroyed machines as a way to oppose "machines" taking away their jobs. Now, the AIGC is also threatening the jobs of thousands of workers. Will there be a new "Luddite movement"? Image of the Luddite Movement generated by Discord, but the times are irreversible and we can only face it. Technology is inherently neutral, and advanced technology brings advanced productivity. But when human nature is combined with technology, it may be deformed and unpredictable. Think about the Internet, which was meant to help globalization, but the end result is anti-globalization. ChatGPT/AIGC originally helped humanity move forward, but will the final result be anti-human? This, God knows!