Since the launch of ChatGPT on December 2, 2022, two years have passed, and AI is reshaping various industries at an unprecedented speed, completely changing the way people acquire knowledge, accumulate experience, and gain insights into the world, and refreshing the mode and habits of human-computer interaction.
In this wave, a huge figure that cannot be ignored has emerged: Microsoft. In fact, this veteran tech giant of the PC era is regaining control of AI era issues. In the past few years, Microsoft has invested $13.75 billion in OpenAI, the leader of the AI revolution, to provide computing engines for various AI companies using Azure. More importantly, Microsoft has provided the most valuable confidence for the faster arrival of the AI era.
Zhang Qi, a "veteran" of Microsoft, has joined Microsoft's US headquarters since 2002. He is now Microsoft's senior global vice president and president of Microsoft's AI Asia Pacific region. He is responsible for the team of Internet business and artificial intelligence platform in Asia. Zhang Qi has over 20 years of experience in big data, artificial intelligence algorithms, and other fields, making outstanding contributions to the implementation of Microsoft's AI technology.
After the rise of generative AI represented by OpenAI's ChatGPT, AI entered a new narrative chapter. Zhang Qi found that when AI shows the ability to surpass humans in content, code generation, and other aspects, the productivity of many jobs has been greatly improved, and the original organizational structure and production relations will face tremendous changes. So in 2023, he proposed the concept of "solo entrepreneur", using the power of AI, one person can also have the ability to match the work of a team.
But with technological innovation comes another concern: Will AI trigger a large-scale wave of unemployment? Zhang Qi's view on this is that some existing jobs will definitely be impacted, but at the same time, the abilities of each individual will be infinitely amplified. Those without programming backgrounds may be able to develop their own applications through large models, and those who cannot draw can use AI tools to generate high-quality artwork.
In an interview with Blue Whale News, Zhang Qi even proposed his vision for how future work will evolve. The company system that has lasted for hundreds of years may face the impact of AI. Companies may no longer need to organize and hire employees, but instead break down their work into a series of tasks and distribute them to crowdsourcing platforms.
Zhang Qi describes himself as a "technical optimist" who does not want to be limited to linear thinking inertia and underestimates the possibilities of the future. AI will change many things that are currently unimaginable. And predicting the invisible changes brought about by exponential growth is what excites and shocks many people in our industry He said.
The following is a recorded conversation between Zhang Qi and Blue Whale News, with some deletions and adjustments made to ensure the original intention.
Don't overestimate the present, don't underestimate the future
Blue Whale News: Why is the concept of "solo entrepreneur" proposed?
Zhang Qi: In this AI revolution led by AIGC and big models, it will be found that an important thing is that previously it was thought that "knowledge" was a unique ability of humans, but now it can actually be generated in large quantities at a very low cost through AI, and can generate a full range of content that can be mixed, including text, images, videos, and sound.
Not only content production, but also AI can be used to summarize while reading articles. When writing something, give AI a topic that can be completed in one or more very standard languages. Previously, it took many years of training to develop aesthetic sense in order to draw a good picture. Now, with a prompt input, it can be done. AI has greatly improved the production efficiency of content and knowledge generation.
In the concept of 'solo entrepreneur', the most crucial transformation actually lies in the ability to generate code. It's not that code is no longer important, but rather that humans may not need to write code directly and express it in natural language to tell machines what they want to accomplish. Large language models can help you generate code.
Why is code important? We say 'software changes the world', and the efficiency improvement of the last wave of information revolution is actually due to the arrival of computers, which are behind the productivity improvement of their programming code in completing all previous stages.
Although there are millions of software programmers, it was still a scarce resource before the AIGC wave, and you will find that the ability of this scarce resource is now largely replaced by large language models. So everyone will become a programmer, which is the most important dimension for improving efficiency.
With the support of this generation of AI, content summarization, code generation, and error correction may have been unique abilities of humans that require a lot of training to obtain, but now they have become very low marginal cost and easy to acquire abilities. Things that used to require a lot of manpower can now be completed with fewer people, resulting in better efficiency.
So, if the means of production and productivity change, will the production relations also change? The more important message covered by the concept of "solo entrepreneur" is, what kind of social changes may be brought about in such an environment? What can a person do? There is a lot of room for imagination here.
When there is such an efficiency improvement, many issues need to be reconsidered. Why maintain the original organizational structure when building prototypes? Is it really necessary for six or seven people to coordinate in order to complete the initial idea, prototype construction, and even several iterations of a product? I don't want it any more.
Organizations are effective, but once they reach a certain level, efficiency will decrease because coordination is crucial. When a person and AI achieve full line communication, it will greatly improve organizational efficiency. It doesn't require such a large organization, and there's no need for changes in hierarchical aspects. Even considering whether this traditional approach is the best.
Of course, this state is more suitable for the MVP (Minimum Feasible Product) stage. If we really want to mass produce products on a large scale, building a product that serves hundreds of millions of people still requires a lot of teamwork.
Friends in the investment field are very excited to hear this. Originally, when it came to investing in a project, it was usually a PPT presentation. Some people may build a very simple prototype and use it to make judgments and conduct research. Sometimes you're betting on your cognition, and other times you're betting on some luck.
But now it's completely different. With very little investment and the support of AI, prototype products can be built, at least in a more formed state and an MVP, which will greatly improve their investment judgment and decision-making ability, including the probability of success.
The investment amount will also be different, the same 1 million can do more things. The width and speed of trial and error may vary. Previously, it used to take several months to receive feedback, but now it may only take a week.
Blue Whale News: With the launch of various AI tools, the threshold for entrepreneurship will decrease. What will be the competitive factors among entrepreneurs in the future?
Zhang Qi: I think a product manager with unique domain knowledge in a particular field may become a very important or scarce talent in the future.
What is the specific knowledge in a field? It's about understanding the field, not just summarizing knowledge. For example, a doctor knows the entire process of a hospital, its strengths and weaknesses, and what patients really need to pay attention to. These are domain knowledge that can help people find opportunities to improve efficiency or services within existing systems.
After possessing this cognition, with the support of AI tools, the process of discovering these opportunities can be greatly compressed and accelerated. Previously, there may have been more resistance, firstly due to capital issues, and secondly due to team coordination issues. But now that these things have all become smaller, the changes a person can make are very significant.
In the environment of a solo entrepreneur, once a person has a deep enough understanding and a great willingness to change something, everything becomes more possible today.
Blue Whale News: Will computing power or data become the most important means of production in the future?
Zhang Qi: The importance of data is undoubtedly increasing. In addition, GPU technology is iterating at a very fast speed now, and the capabilities of data centers are constantly increasing. Models are also becoming more optimized and larger. This continuous iteration indicates that computing power will become an important part of new production resources.
But is it in a restricted state or an infinite state? It depends on how to think about the problem. I am actually a technology optimist, always forcing myself to think about the future with an exponential attitude because there are too many possibilities that go beyond our understanding today.
People tend to overestimate the changes of one or two years and underestimate the changes of ten years. Human thinking tends to be linear, and it is difficult to appreciate the possibility of exponential growth, which is a problem of habitual thinking.
The changes brought about by AI are already happening in reality, but I believe that what we can see is still linear. And anticipating the invisible changes brought about by exponential growth is what excites and shocks many people in our industry.
Will AI make more people unemployed?
Blue Whale News: We have discussed many changes related to productivity or production relations, how will demand change? When many problems are solved and a lot of money or manpower is saved, will it create new demands or leave many people with nothing to do?
Zhang Qi: If we only limit ourselves to the current work state, the efficiency of a job will definitely be reduced, and the required manpower will be compressed, which will definitely affect some existing jobs.
But if we let go of our thoughts, everyone has a desire for change, and at the same time, what can be done can be magnified, and some things that were previously unimaginable can be accomplished. If billions of people around the world could release their own tools for more people to use, what would the world be like?
I think it's difficult to simply think about future manpower compression based on current thinking. When infinite space is revealed, all supply and demand relationships will undergo significant changes.
Blue Whale News: Since the Industrial Revolution, organizations like companies have continued their history for hundreds of years. What impact will AI have on corporate governance and social division of labor? Will there be any changes in our way of working in the future?
Zhang Qi: Work is actually about acquiring employees' specific skills and time through employment relationships to complete the company's tasks.
In the era of AI, many existing workflows can be thought about through agents. Breaking down something into tasks is important. When work is transformed into tasks, certain tasks require people with a deep understanding of specific fields to complete, such as architecture, or some related to product positioning and user cognition, which is crucial for the overall product form and strategy. Some tasks require manual and AI assistance, while others may be fully automated.
In the future, it may not be necessary to hire employees and buy out their time to complete the work. In fact, work will transform into a series of tasks. This system can help you decide which tasks to distribute to where, greatly improving efficiency.
I think crowdsourcing could be an interesting intermediate where tasks can be placed on a crowdsourcing platform. These tasks may be in different stages or states. When people can subdivide work and mobilize needed resources in different resource pools, I believe that the definition of work itself will change.
'Solo entrepreneur' does not refer to a person who simultaneously serves as an engineer and product manager. In fact, a person can have an AI system or assistant to complete multiple roles through these 'clones'. When a person has a strong desire to complete a specific task, the ability of AI can "produce" the required developers or testers. Adding AI to one person is equivalent to having a team.
Blue Whale News: What are the thresholds for people to use AI? Do you need a high educational background?
Zhang Qi: First of all, you need to understand what you want to accomplish, which is your dream. Secondly, one must have the ability to ask questions and learn how to effectively interact with large models.
At present, it is indeed necessary to have some training to effectively use prompt engineering, or to handle code issues and logical problems. Under this combination, if you also have a strong educational background, you will definitely be more efficient in your areas of expertise.
Blue Whale News: If efficiency and structural optimization reach their extreme, will it exacerbate social division? A small number of people who possess knowledge and skills, have received training, and are more likely to interact with AI can meet the vast majority of society's needs, while the majority of the remaining population may only have needs and no longer participate in production?
Zhang Qi: I don't think this will exacerbate the division of human society. In fact, it may enhance the abilities of many people and quickly reach a very high level.
Taking painting as an example, I did not have the ability to create artworks before, so the growth and business opportunities related to it were all closed to me. However, now with computing resources such as Midjourney or DALL-E AI tools, I can directly ask questions or tell it what I want to draw through typing or even voice interaction, and the door to opportunity suddenly opens.
This is no longer something that only professionally trained painters can do, and today it is entirely possible for hundreds of millions of people to achieve it. What possibilities will arise when this door is opened? Will it widen the wealth gap or provide opportunities for more people to reach previously impossible states? If so many people can achieve it, what state will this exponential growth reach?
The original product integrating AI and the new AI super application will coexist for a long time
Blue Whale News: We just discussed a lot about the future, but we need to get back to reality. In the entire AI industry, people are making various attempts, whether in model development or application. What is your opinion on this? What directions are worth paying attention to next?
Zhang Qi: In this wave of AIGC, the collaboration between Microsoft and OpenAI can be written into business school textbooks and is considered a highly forward-looking and creative way of cooperation.
Microsoft's positioning of product form is "Copilot" (intelligent co pilot), which is very important. People still dominate, and with Copilot's support, they can enhance their abilities and help them complete tasks more effectively, rather than replacing people or their judgments. It aims to assist rather than harm users.
There are several products worth mentioning. For example, GitHub Copilot is actually the product that truly leads Copilot innovation, significantly improving developers' work efficiency by providing features such as code generation and testing. Copilot for Microsoft 365 is part of our productivity tools, including Word, PowerPoint, Excel, and Outlook. I often use Outlook to write emails, and with just one sentence, it can provide elegant expressions from grammar to sentence structure, often better than my own writing. Teams Copilot is also a great product that automatically generates meetings.
In fact, using this AI capability is quite simple. But in the right context, it played a huge role. These examples, although simple, demonstrate how technology can make productivity improvements within reach.
We are still the first company to truly introduce large models into search engines, and released a new Bing search on February 7th. Search engine is important because it is the most important dimension to obtain information and knowledge in the Internet era. The traditional method of providing 10 links to obtain information has now been transformed into a knowledge aggregation search method. Search engines have become more efficient and knowledge driven, which has a significant impact on everyone's productivity improvement.
Another important dimension is AI for Science, and the application of AI in the field of science may not have received sufficient attention. For example, Google's AlphaFold can predict the structure and interactions of living molecules with unprecedented accuracy. AI can be applied in the fields of materials science and medicine, and can achieve results that previously took a long time to discover in a short period of time.
AI for Science has a profound impact on humanity's fundamental understanding of the world and the advancement of the most basic disciplines. This transformation will have a compound effect, and new materials and discoveries may trigger a series of new innovations and developments. The investment in AI for Science should be long-term and patient, which is very important.
Blue Whale News: These changes are actually adding AI capabilities to existing products to improve the efficiency of existing scenarios, rather than creating a completely new closed application environment. Will this become a trend in the future AI era?
Zhang Qi: This coexistence state may last for a long time. Copilot can be a standalone product form, but it can also be integrated into many existing apps. Large model companies both domestically and internationally have their own applications, and the functions of these applications are also integrated into other products.
Although it hasn't happened yet, there is a consensus that in the AIGC era, there may be a super application like WeChat that becomes an independent entry point. Everyone is striving in this direction because it will indeed change many things.