AI was the star of 2023. ChatGPT, New Bing, GPT-4, Google Bard, MidJourney v5, Adobe Firefly and others showed us: AI is close and changing our world.
From February to March, hundreds of OpenAI's API applications emerged in research, reading, analysis, programming and more. The giants joined the game, integrating AI into their products. The internet saw a new situation not seen for over a decade because of AI.
So, an old question was asked again: will AI replace our jobs?
We thought low-level labor-intensive work was easy for AI; and creative, thinking, and emotional work was hard - but this was human fantasy.
The fact is: MidJourney and Stable Diffusion can generate realistic AI paintings and photos. Human's aesthetic ability is just noise points following some pattern for AI.
What about thinking? ChatGPT can talk to users naturally, answer questions, polish texts, extract information, and write drafts. New Bing and Bard added web search, which can summarize information based on data and models, and tell us the answer.
AI can do many things: comfort, listen, chat, design, analyze, edit. But can it feel? No. Psychology uses AI to make robots for CBT (cognitive behavioral therapy) and gets good results. Chatting and listening are easy for big language models.
But AI is not good at "action" and "movement". It can only use sensors to imitate the real world. It does not understand physical laws. Jobs that need these skills are hard for AI now. Even a toddler can do better. This is called "Moravec's paradox".
AI will change our lives and jobs in the future. But the AI we have now and soon is not AGI (Artificial General Intelligence). It is tool-type AI. It helps us with tasks, but it does not think like us.
Artificial general intelligence (AGI) is AI that can think like humans, understand their situation, and solve general human problems. It is common in sci-fi. If AGI exists, it will replace most jobs and change everything.
But don't worry: AGI is far from us. We can't make it now or soon.
Our current AI, like ChatGPT's language model or AI painting's diffusion algorithm, are models that train for specific situations, use huge data to fill in, and learn patterns and rules from data.
They process data better than humans, but they are limited by their design. They can only do specific tasks they are trained for and can't have "subjectivity".
For example, AGI is like a computer, but our current AI is like a vacuum cleaner or a washing machine. They do specific tasks better than humans, but they can't go beyond their design limits.
AI tools will change our work like machines replaced workers. Let's see: how will current and future AI tools affect our lives?
OpenAI published a paper with an interesting view. They listed 2087 "tasks" (DWA, Detailed Work Activities). Tasks are the smallest work units, and a job may have multiple tasks. For example, "decision making" is a job that may have tasks like "collecting information" and "evaluating priorities".
Then, OpenAI asked humans and GPT-4 to assess: if AI can do a task in half the time or less without losing quality, it is replaceable - OpenAI called it "exposed to AI". If most of a job's tasks are exposed to AI, the job is exposed to AI.
OpenAI's conclusion was: after human and GPT-4 evaluation, about 80% of known jobs will have at least one task exposed to AI, and 19% of jobs will have half of their tasks exposed to GPT.
Here are some specific examples:
This table shows how AI affects different jobs. The 100% part means: all tasks in these jobs can be done by AI. These jobs include: mathematicians, tax preparers, analysts, writers, web developers...
Some people think this means: AI will take over these jobs and people will lose work. Does it mean that in 19% of the jobs, half of the people will be jobless?
No. I use "affect" instead of "replace" because it means to improve efficiency. We should think of it as: people in these jobs can produce more than twice as much.
Some people may ask: isn't this still replacing? A task that needed 10 people for six months, now needs 5 people for six months, doesn't that mean 5 people may be fired?
No, the reason is simple: demand will increase.
An industry that can produce twice as much means lower costs and easier entry. Then more people will want to join this industry. As long as this industry is not full, its demand and supply will grow.
Many industries need more AI, which boosts their whole chain and creates more jobs. For example, e-commerce lowers the entry barrier and makes more people participate. It drives factories, brands, logistics, etc. The industry is not destroyed, but more prosperous.
Similarly: fewer artists in game studios, but more new studios; shorter novel cycle, lower cost and higher output for publishing and film, more novel writers...
Tool-type AI has no subjectivity, it needs people to use it. Everyone has limited time and energy. As long as society has division of labor, there is a lot of demand. I want to translate a book, design a website, edit a video, I can let AI do it-but I can't do everything by myself. I need to give some things to others. Many such demands create huge employment under AI acceleration.
AI connects everything and makes value exchange more efficient.
AI is not causing unemployment, but making society more detailed and effective in exchanging value. Many people can do things they couldn't before with AI services; many people can use their skills better with AI's help.
AI is a catalyst for a more prosperous and abundant world. It connects everything and speeds up information and value flow. The world has many complex systems with many links. Each link matches supply and demand, exchanging value. But many links are blocked, slow, or broken due to resources and reality. This causes resource mismatch and inefficiency.
For example, you may feel this: there are many movies and TV shows, but none of them are what I want to watch, and I can't find what I want to watch. Maybe the producers don't know or don't care about my needs. Or, I have an idea that I want to make it happen, and I think it can help many people, but the time and cost are too high, so I have to give up.
This is resource inefficiency. Situations like this happen in most industries. Supply and demand can't match well, or the supply chain is too long, making feedback hard and ineffective.
AI eliminates obstacles and accelerates the flow of information and value. The world becomes faster and more precise: products are updated more quickly, new products and demands emerge, ideas turn into reality.
AI connects everything as a new interface and a new "Internet". All products and services access AI. "Using AI" is normal.
Some scenarios:
- Writing code: tell AI requirements, it writes code, we improve it. Problems: ask AI solutions, it diagnoses online.
- Graphic design: use AI models and requirements to generate sketches, we select and improve them. Details: use AI to assist.
- Product development: AI analyzes user needs, extracts improvement suggestions, designs prototypes, tests products, collects data. We steer and decide.
AI will collect and upload our work data; manage projects; schedule resources; assign tasks. AI can also analyze data from different departments; warn us of potential problems.
AI can help us design new drugs; cure diseases; analyze literature from different fields; find potential topics and directions for research; provide experimental design and methods.
AI can analyze our life data and social media; match us with suitable friends and partners. If we want to be alone: we can chat with our own chatbot; no need for meaningless socializing.
Everyone will have their own AI assistant; trained based on their data; familiar with their preferences and habits. It will be our "second brain"; give us suggestions and opinions; big or small things.
Some of these scenarios are partly realized; some are still far away. But not for long. We will live in the physical world; the Internet; and AI.
AI will extend our tentacles and possibilities. We can delegate boring, repetitive work to AI, save time, learn new things, technologies, directions, ideas, and explore new possibilities and lives.
AI development has some risks to be aware of.
AI will be an "amplifier" for everyone - it will enhance your abilities. The more skills and knowledge you have, the more possibilities and value you can get with AI's help.
A simple logic is: AI can make "learning" easier. Obviously, the stronger your learning ability, the more AI can help you. This will widen the gap between you and others, and this gap will only grow bigger.
Similarly, AI can greatly increase "output". The faster and higher quality your output, the more obvious AI's effect on you. This will lead to a head effect, which will improve the overall quality of the industry while making some of the bottom participants more likely to be eliminated.
Therefore, although in the previous analysis, I proposed: the whole society can run more efficiently and better with AI's help, but at the same time, it may also exacerbate inequality, widen the gap between people, make the industry update faster, and make everyone have to "run around" to avoid being eliminated by the industry - this is entirely possible.
So, facing the wave of AI, my views have always been these points:
1)Specialize in your own strengths and professional fields, and constantly improve your professional level. 2) Think about how to apply AI to your profession and improve your output and efficiency. 3) Constantly receive new information, embrace new changes, expand your horizons, contact more people and industries, and think about what new sparks you can collide with.
AI is a tool. You and others can use this tool. Our enemies are not the tools, but the people who use them.
Society also has a hidden risk: it can exploit, alienate and control individuals.
When machines were invented, people thought they would replace workers and make work easier - did they? History shows otherwise. Machines led to more factories, more demand, and more work. This is not machines' fault, but capital's nature.
When computers appeared, people thought they could solve problems beyond human power, and we could relax. Did they? No. We depend on computers and work anytime, anywhere. Our time is divided into smaller pieces for new needs and tasks. Society sped up, but so did the workload.
When phones appeared, people thought our contact with the world would grow. But it turns out our thinking and life are limited on this small screen, drawn by endless social media, news and entertainment, ruled by "FOMO" (fear of missing out) and reward loops. Our vision narrowed instead.
The leap and breakthrough of technology not only brings about the improvement of social productivity but also often brings another result: exacerbating social Matthew effect.
People who have more resources always have more choices and are also more able to benefit from the social chain. But the operation of the social chain requires costs. It may bring more work higher social expectations and standards bind us more firmly in busyness.
The whole social chain may run faster and smoother but its effect may be more concentrated on the rich. And those who bear the burden of running the social chain may be you me every one of us.
This is a serious problem and it may also require the whole society to think together about how to deal with it.
Finally add a sentence:
In the near future with the development of AI there will definitely be a period of "pain". Old jobs are accelerated optimized and replaced but new industries have not developed yet. There will inevitably be a chaotic period which may affect many people.
But this period is also a period when you can find needs and opportunities from it and benefit yourself.
I wish you all well.
From February to March, hundreds of OpenAI's API applications emerged in research, reading, analysis, programming and more. The giants joined the game, integrating AI into their products. The internet saw a new situation not seen for over a decade because of AI.
So, an old question was asked again: will AI replace our jobs?
We thought low-level labor-intensive work was easy for AI; and creative, thinking, and emotional work was hard - but this was human fantasy.
The fact is: MidJourney and Stable Diffusion can generate realistic AI paintings and photos. Human's aesthetic ability is just noise points following some pattern for AI.
What about thinking? ChatGPT can talk to users naturally, answer questions, polish texts, extract information, and write drafts. New Bing and Bard added web search, which can summarize information based on data and models, and tell us the answer.
AI can do many things: comfort, listen, chat, design, analyze, edit. But can it feel? No. Psychology uses AI to make robots for CBT (cognitive behavioral therapy) and gets good results. Chatting and listening are easy for big language models.
But AI is not good at "action" and "movement". It can only use sensors to imitate the real world. It does not understand physical laws. Jobs that need these skills are hard for AI now. Even a toddler can do better. This is called "Moravec's paradox".
AI will change our lives and jobs in the future. But the AI we have now and soon is not AGI (Artificial General Intelligence). It is tool-type AI. It helps us with tasks, but it does not think like us.
Artificial general intelligence (AGI) is AI that can think like humans, understand their situation, and solve general human problems. It is common in sci-fi. If AGI exists, it will replace most jobs and change everything.
But don't worry: AGI is far from us. We can't make it now or soon.
Our current AI, like ChatGPT's language model or AI painting's diffusion algorithm, are models that train for specific situations, use huge data to fill in, and learn patterns and rules from data.
They process data better than humans, but they are limited by their design. They can only do specific tasks they are trained for and can't have "subjectivity".
For example, AGI is like a computer, but our current AI is like a vacuum cleaner or a washing machine. They do specific tasks better than humans, but they can't go beyond their design limits.
AI tools will change our work like machines replaced workers. Let's see: how will current and future AI tools affect our lives?
OpenAI published a paper with an interesting view. They listed 2087 "tasks" (DWA, Detailed Work Activities). Tasks are the smallest work units, and a job may have multiple tasks. For example, "decision making" is a job that may have tasks like "collecting information" and "evaluating priorities".
Then, OpenAI asked humans and GPT-4 to assess: if AI can do a task in half the time or less without losing quality, it is replaceable - OpenAI called it "exposed to AI". If most of a job's tasks are exposed to AI, the job is exposed to AI.
OpenAI's conclusion was: after human and GPT-4 evaluation, about 80% of known jobs will have at least one task exposed to AI, and 19% of jobs will have half of their tasks exposed to GPT.
Here are some specific examples:
This table shows how AI affects different jobs. The 100% part means: all tasks in these jobs can be done by AI. These jobs include: mathematicians, tax preparers, analysts, writers, web developers...
Some people think this means: AI will take over these jobs and people will lose work. Does it mean that in 19% of the jobs, half of the people will be jobless?
No. I use "affect" instead of "replace" because it means to improve efficiency. We should think of it as: people in these jobs can produce more than twice as much.
Some people may ask: isn't this still replacing? A task that needed 10 people for six months, now needs 5 people for six months, doesn't that mean 5 people may be fired?
No, the reason is simple: demand will increase.
An industry that can produce twice as much means lower costs and easier entry. Then more people will want to join this industry. As long as this industry is not full, its demand and supply will grow.
Many industries need more AI, which boosts their whole chain and creates more jobs. For example, e-commerce lowers the entry barrier and makes more people participate. It drives factories, brands, logistics, etc. The industry is not destroyed, but more prosperous.
Similarly: fewer artists in game studios, but more new studios; shorter novel cycle, lower cost and higher output for publishing and film, more novel writers...
Tool-type AI has no subjectivity, it needs people to use it. Everyone has limited time and energy. As long as society has division of labor, there is a lot of demand. I want to translate a book, design a website, edit a video, I can let AI do it-but I can't do everything by myself. I need to give some things to others. Many such demands create huge employment under AI acceleration.
AI connects everything and makes value exchange more efficient.
AI is not causing unemployment, but making society more detailed and effective in exchanging value. Many people can do things they couldn't before with AI services; many people can use their skills better with AI's help.
AI is a catalyst for a more prosperous and abundant world. It connects everything and speeds up information and value flow. The world has many complex systems with many links. Each link matches supply and demand, exchanging value. But many links are blocked, slow, or broken due to resources and reality. This causes resource mismatch and inefficiency.
For example, you may feel this: there are many movies and TV shows, but none of them are what I want to watch, and I can't find what I want to watch. Maybe the producers don't know or don't care about my needs. Or, I have an idea that I want to make it happen, and I think it can help many people, but the time and cost are too high, so I have to give up.
This is resource inefficiency. Situations like this happen in most industries. Supply and demand can't match well, or the supply chain is too long, making feedback hard and ineffective.
AI eliminates obstacles and accelerates the flow of information and value. The world becomes faster and more precise: products are updated more quickly, new products and demands emerge, ideas turn into reality.
AI connects everything as a new interface and a new "Internet". All products and services access AI. "Using AI" is normal.
Some scenarios:
- Writing code: tell AI requirements, it writes code, we improve it. Problems: ask AI solutions, it diagnoses online.
- Graphic design: use AI models and requirements to generate sketches, we select and improve them. Details: use AI to assist.
- Product development: AI analyzes user needs, extracts improvement suggestions, designs prototypes, tests products, collects data. We steer and decide.
AI will collect and upload our work data; manage projects; schedule resources; assign tasks. AI can also analyze data from different departments; warn us of potential problems.
AI can help us design new drugs; cure diseases; analyze literature from different fields; find potential topics and directions for research; provide experimental design and methods.
AI can analyze our life data and social media; match us with suitable friends and partners. If we want to be alone: we can chat with our own chatbot; no need for meaningless socializing.
Everyone will have their own AI assistant; trained based on their data; familiar with their preferences and habits. It will be our "second brain"; give us suggestions and opinions; big or small things.
Some of these scenarios are partly realized; some are still far away. But not for long. We will live in the physical world; the Internet; and AI.
AI will extend our tentacles and possibilities. We can delegate boring, repetitive work to AI, save time, learn new things, technologies, directions, ideas, and explore new possibilities and lives.
AI development has some risks to be aware of.
AI will be an "amplifier" for everyone - it will enhance your abilities. The more skills and knowledge you have, the more possibilities and value you can get with AI's help.
A simple logic is: AI can make "learning" easier. Obviously, the stronger your learning ability, the more AI can help you. This will widen the gap between you and others, and this gap will only grow bigger.
Similarly, AI can greatly increase "output". The faster and higher quality your output, the more obvious AI's effect on you. This will lead to a head effect, which will improve the overall quality of the industry while making some of the bottom participants more likely to be eliminated.
Therefore, although in the previous analysis, I proposed: the whole society can run more efficiently and better with AI's help, but at the same time, it may also exacerbate inequality, widen the gap between people, make the industry update faster, and make everyone have to "run around" to avoid being eliminated by the industry - this is entirely possible.
So, facing the wave of AI, my views have always been these points:
1)Specialize in your own strengths and professional fields, and constantly improve your professional level. 2) Think about how to apply AI to your profession and improve your output and efficiency. 3) Constantly receive new information, embrace new changes, expand your horizons, contact more people and industries, and think about what new sparks you can collide with.
AI is a tool. You and others can use this tool. Our enemies are not the tools, but the people who use them.
Society also has a hidden risk: it can exploit, alienate and control individuals.
When machines were invented, people thought they would replace workers and make work easier - did they? History shows otherwise. Machines led to more factories, more demand, and more work. This is not machines' fault, but capital's nature.
When computers appeared, people thought they could solve problems beyond human power, and we could relax. Did they? No. We depend on computers and work anytime, anywhere. Our time is divided into smaller pieces for new needs and tasks. Society sped up, but so did the workload.
When phones appeared, people thought our contact with the world would grow. But it turns out our thinking and life are limited on this small screen, drawn by endless social media, news and entertainment, ruled by "FOMO" (fear of missing out) and reward loops. Our vision narrowed instead.
The leap and breakthrough of technology not only brings about the improvement of social productivity but also often brings another result: exacerbating social Matthew effect.
People who have more resources always have more choices and are also more able to benefit from the social chain. But the operation of the social chain requires costs. It may bring more work higher social expectations and standards bind us more firmly in busyness.
The whole social chain may run faster and smoother but its effect may be more concentrated on the rich. And those who bear the burden of running the social chain may be you me every one of us.
This is a serious problem and it may also require the whole society to think together about how to deal with it.
Finally add a sentence:
In the near future with the development of AI there will definitely be a period of "pain". Old jobs are accelerated optimized and replaced but new industries have not developed yet. There will inevitably be a chaotic period which may affect many people.
But this period is also a period when you can find needs and opportunities from it and benefit yourself.
I wish you all well.
