AI Product Manager Transformation Trilogy – Cognition (2) “Deconstructing the Essence of AI”

With the rapid development of artificial intelligence technology, the role of AI product managers has become increasingly important. However, many practitioners often feel anxious and at a loss in the face of rapid changes in AI. As the second part of the “AI Product Manager Transformation Trilogy”, this article delves into the essence of AI, hoping to help everyone.

1. Preface:

By cognition in the first chapterCognition (1) “Look directly at AI, understand AI”With the rapid development of AI technology in the past two years, the more people around you use AI, the more anxious and urgent you may feel, or even at a loss.

People around me often ask me: I think those AI technologies are so cool and powerful, but I don’t know how to do it.

Over time, the gap between those who successfully use AI and those who fail to use it effectively will become more apparent.

This will give the latter four states: wrecked, out of focus, missed and out of control.

1. Crash:

There is a very appropriate saying, that is, “trial and error costs are low, and miss costs are extremely high”. In the face of this general trend, it may be eliminated due to a wreck. So, how not to miss this opportunity?

This requires us to take the initiative to recognize and try; What is the most important thing to start learning AI? It is the “beginning”. If you never take this step, stay where you are forever.

2. Out of focus:

In fact, many companies will coexist with AI and their main business, and even encourage a few employees to take the initiative to try AI technology, but they will find that it does not have a huge effect, worrying about input and output, and thus losing confidence in AI capabilities.

Essentially, I think the AI entry point must be business-oriented. We need to find those businesses with high business value and high AI maturity, that is, the greatest common denominator of business and model. Using these businesses as an entry point can quickly achieve results in a short period of time and build confidence.

3. Mistakes and loss of control:

When I talk about these two together, many people make mistakes at the moment, manifesting as not knowing what they are doing so that they will not be replaced by AI;

However, if you put a lot of energy into AI, there will be conflicts between your traditional workflow and AI workflow time; How to balance? When I first started doing my first complete AI work flow, I forgot to eat and sleep to devote myself to thinking, so much so that it would affect my main work content.

I think the essence behind the mistake and loss of control is actually the problem of business goals, if we can combine the AI power point we find with our own business modules, it will not only not get out of control, but also help your work efficiency.

Therefore, if you want to get rid of your “AI anxiety”, the best way is to solve this series of problems, out of focus, crashes, out of control, and mistakes, which is rooted in our limited understanding of AI.

Many people feel anxious because they think that the development of AI seems unrelated to them. So why is it not about me? Because you are not really involved, dare not engage, do not act, and always stand by.

My purpose today is to raise everyone’s awareness and understand the essence of AI, so that you can really take action.

I divide the essence of AI into three categories,

• The essence of technology: You need to understand what AI is and what it is not;
• Business Essence: Clarify what AI can and cannot do in order to effectively apply it to business practices
• The essence of man and machine: It discusses the relationship between AI and people. Does AI need people? Who does it reinforce? Who does it replace?

2. The essence of AI technology

Is the key to “artificial” and “intelligent” “intelligent”?

I believe that the essence of AI technology is to simulate human intelligent behavior.

The starting point of all modern artificial intelligence is neural networks; Or to be precise, a simulated neural network;

There are many neurons in the human brain, which receive signals through the dendrites in the neurons, and then transmit information through synaptic output, thinking, and decision-making, and these neurons can transmit information through electrochemical signals. Therefore, the neural network is composed of “neurons” as the basic units, and the neuronal structure is composed of inputs, computing units, and outputs.

So if you say this, do you suddenly understand it more vividly?

Simulate human intelligent behavior through computer technology, including learning, reasoning, judgment, decision-making, etc. We are trying to equip machines with human ability to perceive, think, and act, so that they can handle complex tasks autonomously.

So:The essence of AI technology is the in-depth exploration and imitation of human intelligence.

Algorithms are like the “brains” of AI, determining how machines process and analyze data and make decisions. By continuously optimizing algorithms, AI can learn patterns from massive amounts of data, thereby achieving prediction and judgment of unknown data.

For example, in the field of image recognition, AI can recognize different objects, scenes, and people by learning large amounts of image data; In the field of natural language processing, AI can understand and generate human language, enabling intelligent question answering, machine translation, and other functions.

AI is also silicon-based intelligent life

Many people think that AI is essentially a tool, and they believe that in the AI era, there is no essential difference between AI and computers, IT tools, and Internet applications.

However, in fact, the original definition of the concept of artificial intelligence in 1956 was to give machines human-like intelligence to replace human work.

In the first 50 years, AI developers tried to achieve this through instrumental logic, computer methods, i.e. logical or symbolic methods, and programming logic. But these attempts did not really implement artificial intelligence.

Despite this, the above explorations still led to later IT, computers, calculators and other electronic products. But they are tools, not artificial intelligence.

In the 21st world, AI represented by AlphaGo began to be known to everyone.

So I think the development history of AI is actually the evolutionary history of silicon-based species, which is constantly growing, evolving, and integrating into human society.

With the advent of deep learning and neural network technologies, and now with the advancement of science and technology, with the increase in parameters and attention mechanisms, artificial intelligence is able to recognize not only objects, but also relationships. (For example, now AI can understand the apple you are talking about through semantics, whether it is an Apple phone or a fruit.) ）

Deep learning/neural networks: Helps AI identify objects.

Attention Mechanism/Transformer: Helps AI recognize relationships.

How does AI fit into humans?

We humans use language, and AI has learned natural language, which allows humans to communicate with it, which is called NLP.

What are the characteristics of human language? You will find that human language can describe, reproduce, predict, and create the meaning of everything in this world, which is a set of world codes.

When computer AI has this set of world code, it can carry everything, show intelligence similar to humans, and this set of code can also communicate with us naturally, so humans have a path to AGI. In short, humans themselves are capable of general intelligence.

Secondly, the evolution of AI is actually the same as the evolution of the human brain. You can see the schematic diagram below.

The evolution of the human brain is actually from the brainstem, cerebellum, occipital lobe, parietal lobe, temporal lobe, and finally to the prefrontal lobe and posterior frontal lobe.

The evolutionary logic of artificial intelligence is very similar to that of the human brain.

• [Brainstem and cerebellum]: In the early days of the computer age, intelligence had not yet developed, and computers could only complete the most basic tasks according to people’s instructions
• [Occipital and parietal lobes]: With the development of deep learning and neurons, perceptual intelligence begins to appear, and machines can recognize objects and respond based on this.
• [Temporal lobe]: With the development of deep learning and neural networks, computers are gradually able to recognize natural language and communicate with humans through natural language.
• [Prefrontal lobe, posterior frontal lobe]: Based on natural language and large models, it began to have intelligence similar to human reason, capable of cognizing the world, recalling, connecting, predicting, communicating, reasoning, imagining, and creating.

Currently, although AI technology has made great progress, most AI systems are still specialized AI (Narrow AI) and can only perform specific tasks in specific fields.

In the future, I believe that artificial intelligence (AGI) with universal capabilities will definitely come, and I hope it can flexibly cope with a variety of complex tasks like humans and achieve the comprehensiveness and universality of intelligence.

2. The business nature of AI

Looking back at the development of AI, from the early concept proposal, to the later technological breakthroughs, and then to today’s widespread application, in this process, many companies blindly chase technological hotspots, but ignore the essence of business – to meet user needs and create value.

Second, because AI can produce misleading results, AI is more suitable as a service provider than a decision-maker in business. (It is impossible for current AI to replace humans in direct decision-making)

Combined with the above, at this point in time, I understand the essence of AI business: the value is reflected in the improvement of product servitization and service productization, and continues to amplify the value of services.

Observe your company, industry, or field and consider where you can provide more professional and detailed services, where AI will unlock greater value.

Some early-stage AI startups invest a lot of resources in technology research and development to create powerful AI products. However, due to a lack of deep understanding of user needs, these products often fail to truly solve user pain points and ultimately struggle to gain a foothold in the market.

Through the analysis of company scenarios and businesses, I believe that there are four major development directions for AI in business:

1. Upgrade functional efficiency, pay attention to how existing services can improve efficiency, and disassemble and optimize workflows.

Through artificial intelligence technology, various functional modules (such as production, operation, customer service, supply chain, etc.) within the enterprise are optimized and transformed, so as to achieve efficiency improvement, cost reduction, and intelligent decision-making.

For example, bill identification and reconciliation in the financial reimbursement process, automatic response to customer service dialogue, forecasting market demand, optimizing inventory management, etc.

The “black light workshop” of Tesla’s Berlin Gigafactory collects 128,000 sensor data in real time through the Internet of Things through the AI system, dynamically adjusts the production line configuration, and compresses the changeover time from 72 hours in traditional car factories to 19 minutes. This intelligent scheduling capability stems from AI’s deep learning of the manufacturing knowledge graph – the system can predict equipment wear curves and independently generate preventive maintenance plans, reducing downtime losses by 67%.

JD Logistics’ “intelligent warehouse network brain” shows disruptive value. By integrating three algorithm modules: demand forecasting, path optimization, and inventory simulation, the system increases the national warehousing turnover rate to 3.2 times the industry average.

I recommend starting with a small pilot, such as an AI tool in a department, and gradually expanding the scope of the application to accumulate data and experience.

2. New production capacity and new intelligence, open up new production capacity or intelligent scenarios, and aim at unmet production capacity needs.

In the past, with the help of AI algorithms, businesses could quickly design and produce customized products based on the individual needs of consumers. For example, by collecting consumers’ body data, preferences and style preferences, there are some so-called “guess what you like” on the e-commerce platforms we usually shop at, which are the earliest AI recommendation models.

As the population ages, the application of AI can replace some repetitive and high-intensity labor and alleviate the pressure on production capacity caused by labor shortages. For example, in manufacturing, intelligent robots can work 24 hours a day, making up for labor shortages and ensuring production continuity and stability.

Secondly, in some complex environments and high-risk areas, such as deep-sea exploration, mining and fire rescue, it is difficult for humans to operate directly. AI can replace humans for high-risk tasks, improving the safety and efficiency of tasks, and meeting the needs of special production capabilities in these fields.

Taking the clothing industry as an example, from the original intelligent recommendation, it can even help each user customize clothing through AI, which meets consumers’ needs for personalized products and also opens up new market space.

3. Product and service upgrades, stock deep cultivation and experience leap

AI drives the user experience from “usable” to “enjoyable.”

In most enterprise scenarios, AI may not be able to achieve the growth of new customers, but we can change from “traffic thinking” to “retention management”.

Let me give you an example: in the current AI era, the first thing many companies do is to build their own knowledge base and prepare for future RAG through the knowledge base. Everyone tried to adjust the division of labor between AI intelligent customer service and manual customer service through the knowledge base;

• manual customer service focuses on solving high-level users and complex problems;
• AI customer service improves the “efficiency – experience dual-dimensional value” of existing products and services, optimizes service efficiency (such as shortening response time), enhances user emotional connection (such as intelligent interaction of “understanding users”), and realizes “value digging” of existing users (such as ARPU improvement and repurchase rate growth).

4. New products and services, incremental market, and service equality

I mentioned in the first chapter of the cognitive chapter that now there are more free, low-cost models, and more advanced hardware technology, AI is no longer exclusive to Internet giants, and it has broken the boundaries of the traditional market through the concept of “service equality” and created a broad incremental market space.

Recently, Tencent and Google have launched many low-code AI products, and the combination of low-code platforms and AI tools enables small and medium-sized enterprises to achieve digital and intelligent transformation at low cost, which was previously unimaginable.

Business value key:

I want to remind everyone that AI can accomplish many tasks, but it is not omnipotent.

We should not think of AI as a panacea, but rather integrate it into overall productivity. Let AI focus on what it does well and mobilize tools to maximize overall value.

Therefore, AI needs to return to the essence of business, put user needs first, and pay attention to the input-output ratio of products to achieve real value creation.

3. The human-machine nature of AI

What is a man-machine? Let me explain first: the so-called human-machine is the relationship between humans and AI;

Many public opinion and short videos are saying that AI will completely replace humans in the future, or that a certain position is about to be replaced, causing everyone to have infinite anxiety.

Then let’s talk about the human-machine nature of AI now.

First of all, there is no doubt that AI now has a wide range of knowledge capabilities and thinking ability that surpasses most people, so this will have an impact on all knowledge workers and experts. I mentioned technology equality above, but in fact, AI has also brought knowledge equality.

However, AI is so powerful, but today’s AI lacks initiative, in other words, AI lacks internal motivation, motivation and goals, and it does not need to take responsibility.

Therefore, it needs to be led and used correctly. Therefore, the use of AI is bound to trigger changes in the labor market.

I made a superficial diagram, from the logic of labor, it has followed the pyramid structure from ancient times to the present;

In the past, “decision-making level, management, and executive level” could roughly describe all labor roles in the market;

Today, AI, as an increasingly powerful knowledge worker, will undoubtedly have an impact on all levels;

However, note that this is not necessarily a replacement relationship, and a new workforce logic will be formed in the future, and those who will lead AI will surely stand out at every level.

• The Industrial Revolution brought about changes in industrialized society. With the advent of machines, factories and standardized divisions of labor, productivity has been maximized. Those factory workers and technical backbones have become the backbone of society.
• In the computer age, the emergence of IT technology has promoted the development of computing and storage capabilities, and then realized the division of labor, efficiency and value maximization of information processing.
• In the Internet era, the core is websites and traffic, capital operation, financing investment and asset appreciation, and all fields have transformed the Internet, which has also led to the role of Internet product manager.

In the era of artificial intelligence, what is the essence of AI human-machine?

The essence of AI human-machine is actually labor change.

I will temporarily divide the labor force into two categories: the first is carbon-based biology (human), and the second is silicon-based intelligence (AI). Both are sources of labor supply to business.

By comparing the strengths and weaknesses of AI, we can see which human needs AI will replace, enhance, or increase.

Advantages of AI:

1. The level of knowledge of AI is very high, and it knows all kinds of knowledge through the reading and learning of various data on the Internet, because whether it is the breadth of human knowledge or the speed of induction, humans cannot compare with it.
2. secondly, AI has a very long-lasting ability to work, it is tireless and always online, it does not get sick, it is not emotional, and it does not need vacations; If deployed specially, it will even work without networking.
3. The most important thing is that AI has the replication and inheritance of growth, but human growth cannot be replicated. When an employee of an enterprise leaves, the newly recruited is uncertain whether they can do better than the previous one, and it takes a lot of time to hand over, learn, and remaster. Even when encountering some unreliable employees, even if they are patient and handover, it will not help.

Disadvantages of AI:

1. From the perspective of current AI, AI has not yet generated very much commercial value, which is why we do not feel the “threat” of AI for the time being.
2. AI is still trapped in the screen, its physical capabilities are low, and it is still essentially a brain that interacts with users through dialog boxes.
3. AI lacks subjectivity, it does not take the initiative, does not refuse, and is not responsible, which means that AI cannot take responsibility. If something goes wrong while performing a task, AI cannot be held accountable because they cannot take responsibility. Therefore, AI cannot be considered as a subject of business activities, that is, it cannot play the role of event leader.

Through the above comparison, we find that AI capabilities are very powerful, but let’s think about it, does having capabilities in a company mean having everything? Is ability the same as everything?

Human-machine-capability

We often think that competencies are crucial, but in fact, we are not looking for capability itself, but output and value.

Therefore, competencies must be translated into value.

Man-machine – authorization

In commercial companies, most of them have a superior-subordinate relationship, which also means supervisory work.

Therefore, the premise of ability to generate value must be empowered.

Secondly, in the real world, an employee needs to be responsible for his work, involving salary, performance, etc. (if he does a bad job, he will be criticized or even fired; If you do a good job, you will be rewarded and recognized. ），

Therefore, accountability and empowerment are prerequisites for the exercise of competence. If AI cannot be responsible, it cannot be authorized, and it will not be able to exert its capabilities to generate value.

Human-machine-trust

What is the premise of the next discussion of responsibility?

The premise of responsibility is that everyone trusts you and the boss trusts you. This stems from your daily performance, past history, life and so on.

If you use AI applications now, answer questions nine times out of ten, and draw every time you are hallucinations, as the boss of AI, will you still replace AI in your work with the employee who is familiar with his daily work?

Therefore, a value-oriented credit and responsibility system is indispensable, without trust, it cannot be authorized, and without the ability to delegate, value cannot be realized.

At this time, the human-machine relationship is very delicate,

AI must have someone accountable for it, turning irresponsible capabilities into creditworthy responsibilities.

When Xiaomi car assisted driving has an accident, many people on the Internet will point the finger at the car manufacturer, I think this is unreasonable, at least at this stage: our credit system for artificial intelligence must be jointly built by AI and people, and someone needs to take responsibility for the entire system.

After the above discussion, and then return to the pyramid model above, you may understand better.

• AI decision-maker: The person responsible for the AI’s output, he needs to determine the direction of the AI, i.e., the AI’s goals, value, and expected outcomes. These directions and results are determined by the person in charge of AI. This will greatly increase the requirements for personal business and market insights.
• AI users: Enhance their work efficiency through AI technology, with AI users aiming to improve personal efficiency and achieve personal goals. For example, help with Excel, PPT, modeling, writing documents, etc. It is concrete and independent.

So everyone, as long as we continue to run in with AI, our professions will not be replaced, and our personal value will be amplified and needed more because of AI.

In the era of AI, amplify decision-making and reduce execution;

Please do your best to become the one who controls the AI.