Thinking of Yourself as AI: Four Tips from an Experiment

Original source: Tencent Research Institute

Author: Zhou Zhenghua

Image source: Generated by Unbounded AI‌

When I went for a physical examination not long ago, an elderly doctor asked me lightly: How do you feel about your life? My first reaction was stunned. No one had asked me this question before. I thought for a few seconds, and then almost blurted out: good, I know what I am interested in, what I am suitable for. But just after I finished speaking, I felt that such a summary of my life was too incomplete, and then I added that I have actually experienced many ups and downs, but I am more optimistic.

Later, I asked the same question to ChatGPT, which of course has no emotional tweaks like mine: "As an artificial intelligence language model, I have no life and emotions, so I cannot experience life like a human. I exist to help users Solve problems, provide information and advice. If you have any questions or need help, please feel free to ask me and I will try to help you."

So the question is, how should people evaluate their own life? GPT4.0, which is a collection of human knowledge, gives several standards: the goals set by oneself and the completion status; the relationship between oneself and family, friends and colleagues, and the role played in it; The growth situation at the level; the inner happiness and satisfaction; whether I have a balance in work, family, health and entertainment. GPT4.0 also suggests that people should regularly review their lives in order to continuously adjust and improve.

This may be almost the same as the reference answer that a professional personal growth consultant can give. Before, I often felt that people are more and more like machines, and machines are more and more like people. After the popularization of the Internet, people's communication is mostly through the Internet as an intermediary, and the face-to-face interaction between people is less and less, and the interpersonal relationship tends to be indifferent. Therefore, people are becoming more and more like machines, which also implies my relationship between machines and indifference. On the other hand, machines are becoming more and more like people, which is actually an admiration for technological progress. Machines can imitate human abilities, so that humans can use machines to complete some tasks that could not be completed, such as in extreme environments. work, perform high-precision operations, and especially perform all tasks without emotion.

In fact, through the study of large models represented by ChatGPT, it is found that this type of natural language processing model based on large-scale neural networks has certain similarities with the human brain in its learning and expression methods.

In terms of pattern recognition, both ChatGPT and the human brain have the ability to recognize patterns. By observing a large amount of data, they can learn to recognize the relationship between language and concepts, so as to understand and generate natural language; both ChatGPT and the human brain can learn from experience, ChatGPT learns through training data sets, and the human brain learns through reading, communication In addition, both ChatGPT and the human brain can understand contextual information, generate appropriate responses, and understand ambiguous or polysemous words.

However, there is still a huge difference between the learning method of large models and the human brain. From the perspective of learning methods, the human brain learns through the connection between neurons and changes in synaptic strength, while ChatGPT learns by adjusting the weights in the neural network; the thinking process of the human brain involves consciousness, emotion, memory, etc. A variety of complex mental activities, while the thinking process of ChatGPT is mainly based on calculations based on mathematical and statistical models; ChatGPT does not have the emotions and consciousness unique to the human brain, it is just an algorithm-based tool, which enables humans to experience emotions and establish values and moral concepts.

Previous research has often focused on how general artificial intelligence learns from the human brain, while ignoring how we humans can learn from artificial intelligence.

Robot scientist Peter Scott-Morgan decided to transform himself into a cyborg (cyborg) after suffering from ALS. He underwent a "triple ostomy" and a total laryngectomy, a feeding tube Inserted directly into his stomach, a catheter was inserted directly into the bladder, and then the feces outlet was diverted through a colostomy, with the help of a wheelchair, and the sound was artificially synthesized, making a big step towards the "half-human, half-machine" cyborg step.

If Peter Scott's actions are too advanced and radical, then for ordinary healthy people, it is actually possible to carry out an experiment of learning from artificial intelligence in terms of psychology and thinking. After several rounds of conversations with GPT4.0, I summarized its four reference paths for humans to learn from artificial intelligence:

Path 1: Data-driven decision-making

Most people make decisions, even major life decisions like who to marry, based on intuition. Although intuition may also be the way big data is presented at a higher latitude. However, AI systems usually make decisions based on a large amount of data. Therefore, we can try to collect and analyze relevant data when making decisions in our work and daily life, so as to make more informed and objective choices.

For example, when purchasing a house, you can pay attention to four aspects of data: study the historical data of a city’s real estate market, understand the trend of housing prices and market forecasts, which will help determine the timing and budget of buying a house; compare the housing prices of different candidate sectors within the city, Cost of living, transportation convenience, educational resources, medical facilities and other data for comparison and trade-off; evaluate return on investment, and analyze the potential returns of real estate investment in candidate housing, including rental income, house price appreciation, etc. This helps to determine the purpose and expected benefits of buying a house; consider loan interest rates: understand the interest rates, repayment terms and other information of different banks and loan products in order to choose the most suitable loan program.

Data decision-making can also be used in mate selection. For example, the following four dimensions are considered important factors for a harmonious relationship between husband and wife: common values, understanding each other's values, beliefs, life goals, etc., to evaluate whether there is a common foundation and long-term Compatibility; communication skills, observe how each other communicates in different situations, and understand whether they can effectively solve problems and deal with conflicts; money concepts, understand each other's consumption habits, savings concepts, investment concepts, etc., to evaluate the financial management Whether they can reach an agreement on aspects; living habits, observe each other's living habits, hobbies, social circles, etc., to assess whether they can adapt and support each other in daily life.

Still, there is a balance between data analysis and emotion when making work and life decisions. Especially in decisions involving relationships, emotional and human considerations are equally important.

Path 2: Strengthen the habit of logical thinking

AI systems typically base their reasoning on logic and algorithms. Observe friends with strong logical thinking around you, they often have strong analytical ability, high communication efficiency, and problem-solving ability. Here are some tips on how to develop your logical thinking skills so you can solve problems and meet challenges more effectively:

1. Reading: Read classic books on logic, philosophy, critical thinking, etc., to understand the basic principles and methods of logical thinking. In addition, reading professional books and reports on the Internet industry to understand industry dynamics and development trends will help improve your industry analysis capabilities.

2. Learning: Participate in training courses on logical thinking, critical thinking and industry analysis, and systematically learn relevant knowledge and skills. This will help you to use logical thinking better in your work.

3. Practice: Exercise your logical thinking ability by solving logic problems, math problems and mind maps. Doing more exercises will help you better use logical thinking in practical work.

4. Case analysis: study the success and failure cases of the Internet industry, and analyze the reasons and laws. This will help you improve your ability to logically analyze industry phenomena.

5. Communicate with others: Communicate with colleagues, mentors and industry experts to understand how they use logical thinking to solve problems. This can help you broaden your horizons and learn new methods and techniques.

6. Reflection: In work, always reflect on your own thinking process and methods. Think about whether there is a better way of logical thinking in order to continuously optimize your analytical ability.

7. Cultivate habits: develop good habits of logical thinking, such as when analyzing problems, first from the general to the specific, and then from the specific to the general; when demonstrating ideas, ensure that the arguments are sufficient and relevant; when making decisions, weigh the pros and cons, full analysis etc.

8. Continuous learning: Keep the desire for new knowledge and skills, and keep learning and growing. This will help you stay competitive in the ever-changing Internet industry.

9. Patience: Improving logical thinking skills takes time and practice. Be patient and believe in your ability to continually improve.

Path 3: Improve pattern recognition ability

In daily life, when we see poplar trees on the side of the road, we will classify them as plants, and those with professional knowledge in botany will further classify them into angiosperms, dicotyledonous plants, willow orders, salicaceae, Populus, which is pattern recognition in humans. Pattern recognition abilities vary from person to person. Those who have the ability to draw inferences from one instance and understand by analogy have excellent pattern recognition skills. Since the 1970s and 1980s, as an important branch of artificial intelligence, pattern recognition has been applied to image analysis and processing, speech recognition, sound classification, communication, computer-aided diagnosis, data mining, etc. Although artificial intelligence pattern recognition is not perfect, it also helps humans to further improve their own pattern recognition capabilities, especially how to find common patterns to solve the same type of problems more quickly in work, improve the ability to summarize and draw inferences from one instance, and the large model The learning process also gave us inspiration:

1. It is necessary to pay attention to the analysis of the problem. When solving the problem, we must first deeply understand the nature of the problem. Understand the context, causes and effects of problems in order to find the most appropriate solutions;

2. Learn from experience, learn from past experience, and summarize success and failure cases. Analyze commonalities and differences among these cases in order to find common patterns applicable to similar problems;

3. Establish a knowledge base, record the methods and techniques for solving problems, and build your own knowledge base. This will help you quickly find solutions when you encounter similar problems;

4. Interact with others: Interact with colleagues, mentors, and industry experts to learn how they have solved similar problems. This can help you broaden your horizons and learn new methods and techniques.

Path 4: Continuous learning, continuous self-optimization

Lifelong learning is a slogan recognized by many people, but it is very difficult to practice. In many cases, path dependence is formed for the sake of "safety and security". Old problems are solved by old methods, and new problems are still solved by old methods. An important reason why continuous learning motivation cannot be found is that many people do not know their true interests and expertise, so they cannot maintain their desire for new knowledge and skills.

If you want to rekindle your curiosity about unknown areas and constantly iterate your knowledge system and methodology, you can try the following:

1. Try new things: actively participate in various activities and try different fields in order to discover your interests and potential expertise. This can include attending courses, lectures, seminars, interest groups, etc.;

2. Self-reflection: Take time regularly to reflect on your interests, strengths, and passions. Ask yourself: What do I enjoy doing? What am I good at? In what ways do I feel confident? This helps you understand yourself better and identify potential areas of expertise;

3. Ask others: Ask your family, friends, colleagues, and mentors for their opinions and suggestions. They may discover strengths and potential that you were not aware of;

4. Set goals: Set short-term and long-term learning and growth goals for yourself. This helps keep you hungry for new knowledge and skills and motivates you to keep trying;

5. Cultivate study habits: Develop the habit of regular study, such as reading books, watching educational videos, taking online courses, etc. This will help you maintain your hunger for new knowledge and skills and keep expanding your body of knowledge.

6. Enjoy the process: learn to enjoy the process of learning and growing, not just focus on the results. When you love learning and enjoy the process, it's easier to maintain your hunger for new knowledge and skills.

7. Learn to adapt to change: Over time, your interests and expertise may change. Learn to adapt to these changes and adjust your learning goals and plans. Being flexible and open-minded will help you find your place in an ever-changing world.