Chapter 18: To You Beyond Age 20: Reasoning as a Science

A Letter from Mr. Pallas's Cat
Dear Piglet and Little Seal:

The flame trees are blooming. Their vivid red blossoms blaze like fire across the campus of Sun Yat-sen University. Spring has passed completely; summer has arrived.

This is the final letter.

We've talked about what reasoning is, and about the toolbox of reasoning. Now, I want to talk with you about the future — not the future of technology, but the future of value.

This book is called To Future Reasoning Scientists. But a "scientist" is not merely someone who masters technology — they are also a guardian of value.

We've spoken many times about the democratization of reasoning. But democratization is not only "enabling more people to use it" — it is also making technology serve more people.

When reasoning becomes the privilege of a few, it intensifies inequality. When reasoning becomes the slave of capital, it distorts humanity. When reasoning serves only efficiency, it forgets meaning.

Today, I want to talk with you about: as a science, how does reasoning preserve human warmth?

Not some distant, sci-fi future, but the future beyond your age of 20. The future that you yourselves will help shape.

Don't rush. Let's speak slowly. This is the last "take it slow."

---

The Issues You Care About Most

Piglet, I remember you asked: "Professor, will AI take our jobs?"

Little Seal, you asked: "Professor, will AI have consciousness?"

These are all good questions. But today, I want to invite you to ask deeper ones:

Efficiency, alignment, and how to give AI the kind of nuanced logic that humans possess — behind these issues, what truly matters?

Let me share three issues I care about most:

1. The Myth of Efficiency

We live in an era that worships efficiency. Faster, cheaper, more.

But Piglet, do you remember the "greedy algorithm" we discussed? At each step, choose what looks best right now — but not necessarily globally optimal.

The trap of efficiency is: we optimize the local and may sacrifice the whole.

For example: making AI generate content faster, but quality drops. Making recommendation algorithms more precise, but filter bubbles worsen. Making autonomous driving more efficient, but safety decreases.

The real question is not "how to be more efficient," but "efficiency for what."

2. The Dilemma of Alignment

Little Seal, you've studied the AI alignment problem — how to align AI's goals with human values.

But there is a deeper dilemma here: whose human values?

My values? Your values? Chinese values? American values? The values of the rich? The values of the poor?

The challenge of alignment is: humanity itself hasn't yet aligned.

We are still debating what justice is, what the good life is, what kind of future is worth pursuing.

For AI to align with humanity, the prerequisite is that humanity first finds consensus.

3. Nuanced Logic

Piglet, you've noticed that AI sometimes lacks "common sense." Little Seal, you've noticed it sometimes lacks "empathy."

But what I want to say is: nuanced logic is more than common sense and empathy.

It is an ability to understand context. Knowing when to be rigorous and when to be flexible. It is an ability to grasp proportion. Knowing when to persist and when to compromise. It is an ability to perceive subtlety. Hearing what is unsaid, understanding what is implied.

This nuance comes from life experience, from cultural immersion, from being with other people.

What AI finds hardest to learn may not be logical rules, but this subtlety of life.

Core Concept: Algorithms Are the Vehicle, but Empathy and Truth Are the Ultimate Heading of Reasoning

Let me share a view that may surprise you:

The ultimate goal of reasoning is not correctness, but understanding.

Correctness matters. 2+2=4, Paris is the capital of France — these factual truths must be correct.

But understanding matters more. Understanding why 2+2=4, understanding why Paris became the capital, understanding the meaning behind these facts.

Algorithms Are the Vehicle

Little Seal, you've studied the history of computing. From the Turing machine to the Transformer, algorithms keep evolving.

But algorithms are only the vehicle — just as the pen is the vehicle for writing, the car is the vehicle for transportation.

The vehicle matters. Without a good pen, you can't write well. Without a good car, you can't travel far.

But the vehicle is not the purpose. We don't write for the sake of using a pen; we don't travel for the sake of driving a car.

Similarly, we don't reason for the sake of using algorithms; we use algorithms to understand the world.

Empathy Is the Bridge

Piglet, you might think empathy is "soft," not as "hard" as algorithms.

But let me tell you: empathy is one of the hardest reasoning tools there is.

Because reasoning is not an isolated intellectual game. Reasoning happens between people, within social contexts.

To understand why someone made a certain decision, you need empathy. To design a human-friendly system, you need empathy. To make technology truly serve humanity, you need empathy.

Empathy is the bridge connecting "correctness" and "meaningfulness."

Truth Is the Heading

Little Seal, you've studied philosophy. You know "truth" is a complex concept.

But the truth I mean is simple: do not deceive yourself; do not deceive others.

In reasoning, truth means:

• Acknowledging the limits of data
• Acknowledging the assumptions of models
• Acknowledging the uncertainty of conclusions
• Acknowledging one's own ignorance

Truth is the beacon of reasoning. It doesn't guarantee we reach truth, but it guarantees we don't lose ourselves in illusion.

---

Key Takeaways: Future Research Directions in Reasoning Science

Based on these reflections, let me share a few research directions I consider important:

1. Explainable Reasoning Systems

Current AI systems are mostly black boxes. Input a question, output an answer — the intermediate process is invisible.

The future direction is: make the reasoning process transparent.

Not simply "attention visualization," but truly traceable reasoning chains. Let users see: why did the system reach this conclusion? What premises were used? What reasoning steps were taken?

This requires combining:

• The rigor of symbolic reasoning
• The flexibility of neural networks
• The friendliness of human-computer interaction

2. Value-Sensitive Reasoning

AI systems inevitably embody some set of values. The question is: whose values are being embodied?

The future direction is: make values a discussable, adjustable parameter of the system.

Don't preset a single set of "correct" values. Instead, design mechanisms that allow different values to:

• Be explicitly expressed
• Engage in rational debate
• Be dynamically adjusted
• Coexist peacefully

This requires combining:

• Theoretical frameworks from ethics
• Deliberative mechanisms from political philosophy
• Implementation techniques from computer science

3. Cross-Cultural Reasoning Patterns

Little Seal, you've studied differences between Eastern and Western thinking. The East emphasizes holism, the West analysis; the East emphasizes relationships, the West the individual.

The future direction is: not to pursue a unified mode of reasoning, but to understand diverse modes of reasoning.

Enable AI to:

• Recognize different reasoning styles
• Adapt to different cultural contexts
• Translate between different thinking frameworks
• Facilitate dialogue between different perspectives

This requires combining:

• Field methods from anthropology
• Analytical tools from linguistics
• Modeling capabilities from artificial intelligence

4. Lifelong Learning in Reasoning

Piglet, you learn quickly. But once AI is trained, it's very difficult for it to learn new knowledge.

The future direction is: give AI the capacity for lifelong learning.

Not through constant retraining, but by enabling it to:

• Continuously absorb new information
• Integrate new and old knowledge
• Correct erroneous beliefs
• Develop new understandings

This requires combining:

• Learning theories from cognitive science
• Plasticity research from neuroscience
• Continual learning algorithms from machine learning

---

Reflection Question: What Human Qualities Does "True Reasoning" Still Need?

Piglet, Little Seal:

This is the final reflection question. I want to hear your answers.

But let me first share my thoughts:

1. A Sense of Humor

Does it sound strange? But humor is a high-level cognitive ability.

Humor requires:

• Understanding conventional expectations
• Creating unexpected twists
• Grasping the right proportion
• Sharing mutual joy

A reasoning system without a sense of humor may be correct, but it will certainly be dull.

2. Aesthetic Judgment

Isn't beauty subjective? But aesthetic judgment requires complex pattern recognition.

Judging what is beautiful requires:

• Perceiving formal relationships
• Understanding cultural symbols
• Experiencing emotional resonance
• Making value judgments

A reasoning system without aesthetic capacity may be practical, but it will certainly be impoverished.

3. Moral Courage

Reasoning is not only an intellectual activity — it is also a moral practice.

Standing for truth requires courage, because:

• Truth may be unpopular
• Persistence may exact a price
• Loneliness may be hard to bear
• Doubt may continue to torment

A reasoning system without moral courage may be clever, but it will certainly be cowardly.

4. The Capacity to Love

This is the most "unscientific" quality — but perhaps the most important.

Because reasoning is ultimately for:

• Understanding those we love
• Solving the problems we love
• Creating the world we love
• Becoming the self we love

A reasoning system without love may be powerful, but it will certainly be cold.

To You Beyond Age 20

Piglet, Little Seal:

The flame trees are in full bloom. Summer is the season of growth.

Writing this letter, I imagine what you will be like beyond age 20.

Piglet, you may become an engineer, solving real problems with algorithms. But I hope you remember: algorithms are the vehicle, not the purpose. What truly matters are the problems you solve, the people you help.

Little Seal, you may become a scholar, researching the theoretical foundations of reasoning. But I hope you remember: theory must connect to practice. The most profound theories often arise from the most concrete care.

And what you share together is the most precious gift of this era: the capacity to reason.

But capacity is responsibility. The capacity to reason is the responsibility to understand the world, to improve the world, to create a better future.

But most importantly: reasoning must serve human well-being.

The true meaning of the democratization of reasoning is not making reasoning technology more widespread — it is returning reasoning to human purposes.

When reasoning serves only efficiency, it creates more efficient inequality. When reasoning serves only profit, it creates more sophisticated exploitation. When reasoning serves only power, it creates more intelligent control.

Reasoning must serve people — serve the dignity of ordinary people, serve justice for the vulnerable, serve the good life for all.

Otherwise, reasoning will become the arrogance of things: technology self-appreciating, capital infinitely accumulating, systems constantly optimizing — while human happiness is forgotten, human meaning is dissolved, human value is instrumentalized.

A few final exhortations:

1. Stay critical: critique technology, critique theory, critique yourself. But let critique be constructive, not destructive.

2. Stay connected: connect with different fields, different cultures, different classes. Isolated reasoning is impoverished.

3. Stay hopeful: the world has many problems, but also many possibilities. Reasoning is a tool for solving problems, and also a tool for creating possibilities.

4. Stay human: in the age of AI, what may be most precious is humanity — those qualities that algorithms struggle to imitate: humor, aesthetic sense, courage, love.

Most importantly: enjoy the pleasure of reasoning, shoulder the responsibility of reasoning, share the fruits of reasoning.

Our journey of taking it slow ends here.

But your journey has only just begun.

Take your toolbox, your questions, your hope — and set forth.

The flame trees bloom for you.

Yours, Mr. Pallas's Cat Black Stone House, Sun Yat-sen University In the season of flame trees The end of the journey, and the beginning

---

Mr. Pallas's Cat's Final Question

Practical Directions (for Piglet)

1. Your age-20 plan: imagine your life beyond 20. What problems do you want to solve with reasoning? What value do you want to create?
2. Tech ethics practice: choose a technology domain you're interested in. Research its ethical issues and propose your solutions.
3. Cross-disciplinary project: collaborate with a friend in a non-technical field on a project. What problems can you help them solve with reasoning tools?

Theoretical Explorations (for Little Seal)

1. Your academic map: imagine your academic path beyond 20. What theoretical questions do you want to research? What contributions do you want to make?
2. Intellectual history: choose a thinker (Chinese or Western). Research their mode of reasoning and its lessons for today.
3. Futures thinking: based on historical trends, predict the future of reasoning science. What will it look like in 50 years?

Shared Reflection

1. The future of humanity: in the age of AI, what is humanity's unique value? Is reasoning capacity still our advantage?
2. The revolution of education: how should future education teach reasoning? Starting from what age? Using what methods?
3. The design of society: how do we design a society that encourages rational reasoning, embraces diverse perspectives, and promotes sincere dialogue?
4. Personal cultivation: as reasoners, how do we keep our minds open, our hearts warm, our morals firm?

Piglet's note: The Professor's letter made me cry. Not from sadness, but from being moved. I realized that reasoning is not just an intellectual game — it's a bridge connecting us to the world. Beyond 20, I want to use reasoning to make the world a little better, even just a little.

Little Seal's note: The final letter, like a graduation address. I realized that the process of learning to reason is also the process of learning to be human. Rigorous yet gentle, critical yet embracing, independent yet connected — these are the character of a reasoner, and the character of a person.

Mr. Pallas's Cat's closing words: For the final words of this book, I want to say something even more fundamental.

We spent 18 chapters exploring the science of reasoning. From the simplest Boolean logic to the most complex Transformer. But all these technologies, algorithms, models — they are only tools.

Tools need purpose.

The democratization of reasoning is not only about enabling more people to master the tools of reasoning. It is about making reasoning serve people — not capital, not power, not the self-appreciation of technology.

When reasoning does not serve people, reasoning becomes the self-appreciation of things. It will optimize efficiency, but not happiness. It will grow wealth, but not justice. It will create intelligence, but not wisdom.

This will be the arrogance of things, not human wisdom.

The arrogance of things says: faster, more, stronger is good. Human wisdom asks: faster for whom? More for whom? Stronger for whom?

The arrogance of things pursues limitless growth. Human wisdom knows boundaries and balance.

The arrogance of things turns the world into computable resources. Human wisdom sees incommensurable values: love, beauty, justice, dignity.

This is the true aspiration of the democratization of reasoning that I pursue: to return reasoning to the human scale, to serve human well-being, to enrich human meaning.

Reasoning is light — but it must illuminate the human path. Reasoning is a bridge — but it must connect human hearts. Reasoning is love — it must understand human pain and hope.

May you become bearers of light, builders of bridges, lovers of humanity.

Take it slow — understanding is what matters most. Then, carrying human warmth, move forward with courage.