Hey, dear readers! We decided to launch a series of articles on AI entrepreneurship in the next time, covering what artificial intelligence is, what is the current development trend, what are the applications, how we should use AI to start a business, etc., to help everyone better grasp opportunities, create wealth and realize value in this new era. (Long span, a lot of content, please collect and pay attention first)
In all aspects of our lives, whether it is mobile phone assistants, voice recognition or autonomous driving, artificial intelligence is quietly integrating into our lives. So, what exactly is artificial intelligence? And how does it change our lives?
Artificial intelligence is the technology that uses computers to simulate human intelligence
First, what is artificial intelligence - the technology of simulating human intelligence with computers
Alan Turing
Turing's Cross-century Question: The discussion of artificial intelligence dates back to Alan Turing's seminal 1950 work Computing Machines and Intelligence, in which Turing posed a cross-century question: "Can machines think?" At the same time, a test method is given, the famous "Turing test", in which human interrogators try to distinguish which text responses are made by computers and which are made by humans, so as to distinguish whether the machine is intelligent. Over the past few decades, some definitions of artificial intelligence (AI) have emerged, but in simple terms, AI is the technology of simulating human intelligence with a computer. When computers can think, learn, understand, and solve problems like humans, we call it "artificial intelligence."
Turing test: 1 human can distinguish which of the 2 closed rooms is a robot through dialogue, if it cannot be distinguished, it means that the artificial intelligence has passed the test
▲ The development of artificial intelligence: From the perspective of time, artificial intelligence has mainly experienced 5 stages of development.
AI development timeline
Representative AI achievements in different historical periods
- Initial development period: 1943-60s of the 20th century. During this period, the concept of artificial intelligence was proposed, technologies such as neural networks and expert systems were developed, and artificial intelligence made some breakthroughs in machine theorem proof, checkers programs, human-machine dialogue, etc., and the representative results were neural networks.
- Rethinking the period of development: the 70s of the 20th century. During this period, artificial intelligence encountered some difficulties and challenges, such as knowledge acquisition, common sense reasoning, computational complexity and other issues, and the development of artificial intelligence entered a trough.
- Application development period: 80s of the 20th century. During this period, artificial intelligence moved from theory to practical application, and expert systems achieved success in medical, chemical, geological and other fields, promoting the second climax of artificial intelligence.
- Stable development period: 90s of the 20th century - 2010. During this period, artificial intelligence benefited from the development of network technology and Internet technology, and realized innovation in remote diagnosis, human-machine game, and smart earth.
- Booming period: 2011 to present. During this period, artificial intelligence benefited from the development of information technology such as big data, cloud computing, and the Internet of Things, and artificial intelligence technology represented by deep neural networks developed rapidly, achieving technological breakthroughs from "unusable" to "usable" in image classification, speech recognition, knowledge question answering, and unmanned driving, ushering in a new climax of explosive growth.
▲According to the intelligence level and ability range of artificial intelligence, it can be divided into weak artificial intelligence and strong artificial intelligence.
- Weak artificial intelligence (ANI): Also known as narrow AI or artificial narrow intelligence (ANI), it is a trained AI focused on performing a specific task, and is currently the most widely used stage, which can surpass human intelligence in specific areas, such as playing Go, recognizing images, translating language, etc. Weak AI drives most of the AI we use today, and "narrow" is probably a more accurate description of this type of AI because it's not weak at all — it supports some very powerful applications like Apple Siri, Xiaoai, Amazon Alexa, IBM Watson self-driving cars, etc. It's worth noting that ChatGPT is now known for its exceptionally good performance, but it's actually a weak AI.
The "artificial intellectual disability" we know is a typical weak artificial intelligence
- Strong artificial intelligence (AGI+ASI): Strong AI is divided into artificial general intelligence (AGI) and artificial superintelligence (ASI). Artificial General Intelligence (AGI) is a theoretical form of AI in which machines possess intelligence equivalent to humans; It is self-aware and able to solve problems, learn, and plan for the future. Artificial superintelligence (ASI), also known as superintelligence, will surpass the intelligence and capabilities of the human brain. Perhaps his best examples come from images from science fiction movies like MOSS in The Wandering Earth 2, Matrix in The Matrix, and thinking robots in 2001: A Space Odyssey. At present, strong artificial intelligence is still in the exploration stage, and there is still a certain distance from our daily life.
Strong artificial intelligence is generally considered to appear in the middle of the 21st century, which will bring great impact and change to human society
Machine learning - the underlying core logic of artificial intelligence
Artificial intelligence is just a beautiful vision, but how exactly can artificial intelligence be realized? The answer is machine learning. Machine learning is a troublesome thing to do, you need to build a model that contains a lot of parameters, and then feed the prepared data (including the correct result) into the model, constantly adjusting the parameters of the model until it is very close or exactly in line with the correct result.
Deep learning is now more widely used, which uses multiple layers of artificial neural networks to learn abstract features and patterns of data to solve complex problems. Its advantages are:
- Can process unstructured data such as images, sounds, text, etc.;
- Features can be learned from data automatically without manual design or extraction;
- Large amounts of data and computing resources can be leveraged to improve the performance and accuracy of the model;
- It is possible to mimic the way the human brain works and achieve the goals of artificial intelligence.
At present, mainstream image recognition, speech recognition, natural language generation, machine translation, recommendation systems, automatic driving and other artificial intelligence are all deep learning models.
Below, we take the hottest GPT-4 as an example to explain the basic principles of artificial intelligence.
▲How does GPT work? GPT performs self-supervised learning with a large amount of unlabeled text data, and its goal is to predict the next word or symbol in the input, and after accepting text and image inputs, generate text as output based on a large amount of data. Just give it a proper hint and it can handle a variety of tasks, such as answering questions, writing articles, programming, etc. What makes him special is the generative pre-trained model that uses Transformer, which is more efficient and accurate when processing long text than traditional recurrent neural networks (RNNs) and long-short-term memory networks (LSTMs).
▲How is GPT trained? GPT training is divided into two phases, pre-training and fine-tuning. In the pre-training phase, the goal is to predict the next word based on a given piece of text. Pre-training enables the model to learn the basic structure and semantics of the language. In the fine-tuning phase, the model performs supervised learning with a small amount of labeled data to adapt to specific tasks, such as question answering, text summarization, and so on. In CHATGPT, fine-tuning data includes conversations between humans and machines that enable them to generate text that resembles human responses.
▲How does GPT speak? GPT uses a generative strategy when generating text, that is, given an input sequence, the model generates a probability distribution from which the word with the highest probability is selected as the output. This process continues continuously until a specific terminator is generated or the maximum output length is reached. For the simplest example, if there are 4 words "today we" in the previous article, then he will analyze the probability of the words "eating" and "sleeping" in combination with the context, and then select the word with the highest probability to say it.
So fundamentally, no matter how powerful the current artificial intelligence seems, it is like Ren Zhengfei said: "The current artificial intelligence is essentially statistics."
How AI is changing our lives: Evidenced by examples and data
Artificial intelligence (AI) has magically changed our lives, making our daily lives more convenient and efficient. Its application has penetrated into various fields and has become the driving force of the development of the times.
▲Unmanned driving field. Autonomous driving technology brings us a safer and more comfortable driving experience by sensing the environment, planning paths, and performing control functions, effectively reducing the incidence of traffic accidents. Self-driving cars are an important application of artificial intelligence in the transportation sector. Self-driving cars can allow us to enjoy safer and more convenient travel. According to statistics, the global autonomous vehicle market will reach $55.6 billion by 2023.
▲Vertical professional field. Expert systems are used in all walks of life and have greatly promoted the development of the computer and network industry. In healthcare, Intel ® Saffron™ AI can analyze 10,000 characteristics of an electrocardiogram, increasing the diagnostic accuracy of heart disease to 90%. In finance and insurance, AI can reveal fraud, and Intel's ® Saffron™ platform uncovered multiple scams that caused more than $2.4 million in losses in a year in just 10 hours. In addition, AI can also help find missing children, improving search efficiency and success rate.
▲Traditional production field. Artificial intelligence can improve the production efficiency of the manufacturing industry, realize intelligent, flexible, customized and quality production, reduce costs, improve quality, and shorten the cycle. For example, Ferro Labs works with Intel to accelerate machine learning models to keep robots running and reduce material costs while improving product quality, and with the help of this technology, an average factory can be expected to save millions of dollars a year. According to statistics, in 2018, the proportion of enterprises using artificial intelligence technology in the mainland manufacturing industry reached 16.5%, an increase of 4.6 percentage points over 2017.
▲Smart life scene. In the age of artificial intelligence, one device can integrate multiple functions. The latest robots can work according to our needs, realizing functions such as air purification, projection playback, safety monitoring, etc. When we browse the web, watch videos, and shop, the artificial intelligence system behind us will provide us with personalized recommendations based on our behavior and preferences. We can also say commands into the phone, and let smart assistants such as Siri, Google Assistant or Xiaoai help us check the weather, set alarms, make calls, etc.
Artificial intelligence has had a profound impact in the fields of natural sciences and social sciences. AI provides a new model for human cultural life. It is predicted that the global smart assistant market will reach $45 billion by 2025.
Looking to the future: opportunities and challenges brought by artificial intelligence
Opportunity – rapid development of productivity
▲Smart manufacturing upgrade. In recent years, with the rapid development of AI, companies in various industries have paid attention to and sought to combine with AI technology. Manufacturing plants leverage AI to increase productivity, improve flexibility, reduce costs, and facilitate self-optimizing operations. According to the analysis, through the use of AI, manufacturing costs can be reduced by 20%, of which 70% comes from the improvement of labor productivity. At the same time, the manufacturing industry can also use AI technology to develop and produce customized innovative products for customers, greatly shortening the delivery time. It can be seen that AI technology will be an indispensable and important part of the future manufacturing industry.
▲ Job market change. According to a report by the McKinsey Global Institute, about 14% of the global workforce may need to change careers into AI-related industries by 2030. While low-skilled jobs may be replaced, AI will spawn more high-skilled and innovative jobs. For example, emerging professions such as data scientists, machine learning engineers, and prompt engineers will emerge in large numbers. According to the statistics of the "Global Artificial Intelligence Industry Distribution" report released by Goldman Sachs, China accounted for 51% of emerging artificial intelligence projects in the world in 2022. However, in terms of global AI talent pool, China only accounts for about 5%, and the talent gap exceeds 5 million.
▲Innovation in the medical field: Artificial intelligence can assist doctors in diagnosis, treatment, prediction, prevention and other medical activities through data analysis, machine learning, deep learning and other technologies, improve medical accuracy and effect, reduce medical errors and costs, and achieve personalized, precise and intelligent medical treatment. For example, artificial intelligence can help doctors perform medical image analysis, identify abnormal conditions such as lesions, tumors, and fractures, and provide auxiliary diagnosis and treatment suggestions; AI can also help doctors with drug discovery, predict drug effects, toxicity, mechanisms of action, etc., and identify new drug candidates or combinations. Google DeepMind's AlphaFold project has successfully predicted the three-dimensional structure of many proteins, providing strong support for the development of novel drugs.
▲Reform in the field of education: artificial intelligence can assist teachers in personalized teaching and learning, formulate special learning plans for students according to their interests, habits and learning needs, provide real-time evaluation and feedback, and improve students' learning effects and interests. It can also track and feedback the learning effect in real time, and use natural language processing and speech recognition technology to score and comment on students' speaking, composition, etc. in real time, and provide suggestions for improvement and content analysis results. At present, the Knewton platform in the United States has used artificial intelligence to analyze students' learning data and provide them with customized learning resources and suggestions.
Challenge - backwardness of production relations and lack of management means
Privacy leakage: As the demand for data in artificial intelligence continues to increase, data protection and privacy leakage have become increasingly important issues. On March 31 this year, the Italian Personal Data Protection Agency announced that it will temporarily ban the use of chatbot ChatGPT from now on, and has launched an investigation into the alleged violation of data collection rules by OpenAI chatGPT, and temporarily restricted OpenAI's processing of Italian user data, which has attracted global attention.
▲Employment threat: The widespread use of artificial intelligence in various industries is very likely to cause the disappearance of a large number of existing jobs, especially some simple repetitive jobs, and the corresponding employees will also lose their jobs. Even if new jobs are created, these employees may lose their jobs because they are not competent. Yuval Harari, author of A Brief History of the Future, believes that if artificial intelligence develops to a certain extent, the vast majority of people will become a useless group. Especially in the era of strong artificial intelligence, after robots have human psychological abilities, there will be more types of jobs that can be replaced.
▲Uneven distribution: Inequality in employment will lead to inequality in the right to social wealth gain, and more countries, enterprises or individuals who master artificial intelligence technology can obtain greater distribution rights of social wealth, while other parts of human beings are in a extremely disadvantaged position in social competition. This may make the gap between the entire human society continue to increase, and the rapid differentiation between countries, between enterprises, and between people will also be facing, with the result that the stronger the stronger, the weaker the weaker, the extreme imbalance between the rich and the poor, and the foundation of harmonious coexistence in human society will no longer exist.
In China, artificial intelligence is being integrated into our lives with great momentum. Smart homes, driverless cars, face payment and other technologies emerge one after another, which is inseparable from the support of national policies. There are also many large-scale AI projects that are also well known, such as Baidu's Wen Xin Yiyan, Huawei's Pangu model and so on.
Lee Kai-fu once said: "There are five fields that are difficult to replace by AI: skillful, creative, caring, requiring complex strategic and cross-domain thinking. "
In this era dominated by robots and artificial intelligence, we should value skills that cannot be easily replaced by machines. We need to stimulate the creativity, love, cross-border thinking and strategic skills of young people to equip them with the qualities to keep pace with the times. In this way, they can gain a foothold in this era of robots and artificial intelligence blooming everywhere.
We must strengthen our confidence and face challenges bravely. Only by always being vigilant and adapting to the needs of the times can we seek development in this changing world and create a better future.
In the next issue, we will introduce the latest trends and application scenarios of artificial intelligence development, please follow me and achieve yourself in the change. If you have any ideas or questions, please leave a message in the comment area.