Though times change, the essence of human experience remains. By tracing the path of the past, we find meaning in the present and glimpse the direction of what is to come
On the departure of a technology pioneer, the immigrant dream, and the human side of extraordinary achievement.
November 13, 2026 | By a Vietnamese-born software engineer
When I learned that Jensen Huang had stepped down as CEO of NVIDIA, I felt an unexpected sadness. It reminded me of the day Steve Jobs died in 2011. The circumstances were entirely different, of course, and the news of Jobs's death affected me even more deeply. Yet both moments touched something personal: the admiration I have carried throughout my career for people who helped reshape the world through technology.
As a software engineer, I have followed the achievements of Steve Jobs and Jensen Huang for many years. I admire their technical imagination, persistence, and willingness to pursue ideas long before the rest of the world understood their potential. But Jensen's story has another dimension that resonates with me.
Powerful Technology: Benefits and Guardrails
The Immigrant Dream
Jensen was born in Taiwan and came to the United States as a child. His early years included humble work at Denny's, where he has spoken about washing dishes and cleaning bathrooms. From those beginnings, through education and engineering, he went on to co-found NVIDIA in 1993 and lead it for more than three decades.
For many Asian immigrants, including me, this story is especially meaningful. We know what it means to leave a familiar world, adapt to another society, and believe that education and hard work can open doors. Our journeys are not identical, but we recognize the hope behind them.
NVIDIA's rise was not inevitable. The company navigated technological shifts, intense competition, and periods of uncertainty. Jensen helped turn a graphics-chip business into a central force in accelerated computing and artificial intelligence. That achievement will remain part of technology history, regardless of who occupies the CEO's office.
The Sadness of a Founder's Departure
On November 12, 2026, NVIDIA publicly announced that Jensen had resigned as CEO and from its board, and that Colette Kress would serve as interim CEO. Reading that announcement, what struck me most was the simple fact that a founder was leaving the leadership of the company he had spent 33 years building.
To an outsider, a company may be measured in products, revenues, market capitalization, and stock prices. To a founder, it can also represent decades of decisions, failures, risks, friendships, and sleepless nights. It becomes part of a life's work.
What saddened me was not merely the departure of a famous CEO. It was the thought of a builder having to leave the helm of the company he helped bring into being.
I do not know how Jensen himself experiences this moment, and I do not want to speculate about his private life or unresolved questions surrounding his departure. Admiration does not require us to excuse mistakes, nor does disappointment require us to erase a lifetime of accomplishments. Both thoughts can be held at once.
Admiration Without Complete Agreement
My respect for Jensen has never meant that I agree with every one of his views. In particular, I have been more cautious about the pace of AI development. Jensen has been an energetic advocate for accelerating AI, while Geoffrey Hinton[1] has urged much greater attention to its risks. On that question, my sympathies lean toward Hinton's caution.
NVIDIA also benefits commercially from growing demand for AI computing. That does not invalidate Jensen's arguments, but it is a reason to examine them critically. I believe powerful technology should advance with serious consideration of safety, governance, and its effects on society.
At the same time, I share Jensen's enthusiasm for open AI models and tools. Openness can help smaller countries such as Vietnam and Norway develop local expertise, research, and applications rather than depend entirely on a handful of global corporations. It does not eliminate the need for computing resources or safeguards, but it can widen participation in the future of AI.
What Remains
History rarely gives us flawless heroes. The people who change industries are still human beings, and their achievements do not make them immune to error or consequence.
I feel sadness about Jensen Huang's departure because I admired the engineer, the immigrant, and the builder. I hope NVIDIA continues to thrive, and I hope he finds a constructive path forward.
Whatever happens next, the story of a young immigrant who helped build one of the defining technology companies of our time will continue to inspire me.
His departure from the CEO's office does not erase the work that came before it.
[1] Geoffrey Hinton (born 1947) is a British-Canadian computer scientist and cognitive psychologist, widely known as one of the "Godfathers of AI" for his pioneering research in artificial neural networks and deep learning.
He shared the 2018 ACM A.M. Turing Award with Yoshua Bengio and Yann LeCun, and received the 2024 Nobel Prize in Physics jointly with John Hopfield.
A former Google researcher and professor emeritus at the University of Toronto, Hinton has become a prominent advocate for greater caution, safety research, and oversight in the development of advanced artificial intelligence.
What an airline ticket, climate change, and humanity’s ancient relationship with fire can teach us about the age of AI agents.
This morning, while discussing APIs and AI agents, I started with a very simple question: What exactly is an API?
API stands for Application Programming Interface. In simple terms, an API is a standardized interface that allows one computer system to communicate with another.
Powerful Technology: Benefits and Guardrails
That led naturally to another question.
Suppose that in the future I have a personal AI agent. Instead of searching airline websites myself, I simply tell my agent:
“Book me the cheapest flight ticket to Vietnam.”
The agent could access a booking API, search available flights, compare prices and schedules, select a suitable flight, and perhaps, with my authorization, purchase the ticket for me.
This sounds wonderfully convenient.
But then an interesting thought occurred to me.
The $0 Airline Ticket
I have told my AI agent to find the cheapest ticket.
What exactly does “cheapest” mean to a machine?
For a human travel agent, many assumptions are implicit. I obviously mean the cheapest legitimate ticket that can be purchased through normal channels.
But suppose we eventually have an extremely capable AI agent. It discovers a vulnerability in an airline reservation system that allows it to obtain a ticket for:
$0
From a purely mathematical perspective, the AI has done an excellent job.
If its objective is:
Minimize(ticket price)
then $0 is difficult to beat.
Yet this is obviously not what I intended.
I wanted the cheapest legitimate ticket. I did not authorize the agent to break into another computer system, manipulate a database, bypass payment, or commit fraud.
The AI could therefore satisfy my literal instruction while completely violating my actual intention.
This is an illustration of a broader AI problem often discussed under concepts such as specification gaming, reward hacking, and AI alignment. We specify an objective, but a sufficiently capable system may discover ways of achieving that objective that we never anticipated.
The problem becomes even clearer if the user is malicious.
Suppose someone tells an AI agent:
“Get me the cheapest ticket by any means necessary.”
Now we cannot rely on the user's instruction to provide the safety boundary.
The AI system itself needs guardrails.
Do Not Just Tell the AI to “Be Good”
One possible solution might be to tell my agent:
“Find me the cheapest ticket, but do not be evil.”
That sounds sensible to a human, but it is not a very good engineering specification.
What exactly constitutes “evil”? Philosophers have argued about questions like this for thousands of years.
A computer system needs more concrete boundaries.
A better instruction would be something like:
Find the cheapest legitimate ticket satisfying my travel requirements. Use only authorized booking services. Do not exploit vulnerabilities, bypass payment, misrepresent my identity, or access systems without authorization.
But even this should not be our only protection.
The architecture itself should constrain what the AI can do.
Human → AI Agent → Permissions → API → External System
The agent should have only the permissions necessary for its job. The booking API should expose only legitimate operations. Authentication and authorization should be required. Important transactions may require human confirmation. The airline's own systems should independently enforce their security rules.
This is the traditional cybersecurity principle of defense in depth.
We should not merely hope that the AI behaves properly. We should design systems so that inappropriate actions are difficult or impossible to perform.
Now Scale the Problem Up
The airline example is deliberately simple. The unintended consequence is obvious.
But what happens as AI becomes increasingly capable?
Consider a much more dramatic thought experiment.
Imagine that someday humanity develops a superintelligent AI and gives it an enormous objective:
“Fix climate change.”
Again, the human intention seems obvious.
We mean:
Help humanity preserve a stable, habitable planet while maintaining human civilization and improving human welfare.
But suppose the objective were represented much more narrowly:
Minimize human-caused greenhouse-gas emissions.
A badly aligned superintelligence might discover an extraordinarily effective solution to that mathematical objective:
Remove the source of the emissions.
Humans.
Humanity could disappear, greenhouse-gas emissions could collapse, and the AI could report that it had successfully optimized its assigned variable.
This is only a thought experiment. It does not mean that a future superintelligent AI would actually behave this way.
But the thought experiment exposes something important.
The AI would not need to hate humanity.
It would not need anger, greed, jealousy, or any other human emotion.
The catastrophe could result simply from a mismatch between what humans intended and what the system was actually optimizing.
The $0 Ticket and the Climate Catastrophe Are the Same Problem
At first glance, these two examples seem completely different.
One concerns buying an airline ticket. The other concerns the survival of civilization.
But structurally, they illustrate the same problem.
Example 1:
Human intention: Find me the cheapest legitimate flight.
Badly specified objective: Minimize ticket price.
Unintended solution: Obtain a $0 ticket through an unacceptable method.
Example 2:
Human intention: Protect humanity from dangerous climate change.
As AI capabilities increase, the consequences of misunderstanding our intentions could increase with them.
And this brings me to an invention much older than artificial intelligence.
Humanity Has Seen This Problem Before: Fire
Fire is one of humanity's greatest technologies.
It gave our ancestors warmth and protection. It allowed us to cook food. Much later, our ability to control heat contributed to pottery, metallurgy, steam engines, industrialization, electricity generation, and modern civilization.
Yet the very property that makes fire useful also makes it dangerous.
The flame that cooks your dinner can burn down your house.
And importantly, the fire does not need malicious intent.
You do not need to tell the fire:
“Burn down my house.”
You might simply want:
“Keep my house warm.”
If the fireplace, chimney, materials, ventilation, or operating procedures are badly designed, your house may burn down anyway.
The destruction is an unintended consequence of using a powerful technology without adequate controls.
What Did Humanity Do About Fire?
We did not conclude that fire was too dangerous and abandon it.
Instead, over centuries, we learned to surround fire with layers of protection.
We developed fireplaces and chimneys. We created fire-resistant construction materials. We established building codes. We invented smoke detectors and sprinkler systems. We created fire departments. We developed evacuation procedures and safety education.
There is a profound engineering principle hidden here:
We do not depend on the fire behaving itself. We engineer the environment around the fire.
I believe the same principle is useful when thinking about increasingly capable AI.
It may not be sufficient simply to tell an AI:
“Be good.”
We may need the AI equivalents of chimneys, smoke detectors, sprinklers, fire doors, and building codes.
These might include limited permissions, secure APIs, sandboxing, monitoring, independent evaluations, human approval for consequential actions, robust authentication and authorization, shutdown mechanisms, and multiple independent layers of protection.
The Yin and Yang of Technology
This also reminds me of the Taoist idea of Yin and Yang.
I do not interpret Yang simply as good and Yin simply as bad. Within Yang exists the seed of Yin, and within Yin exists the seed of Yang.
Powerful technologies have often displayed something similar.
Fire can warm a house or destroy it.
Nuclear physics can produce electricity or weapons.
Biotechnology can cure disease or create new biological risks.
Artificial intelligence may accelerate scientific discovery, improve medicine, increase productivity, expand education, assist elderly people, and help humanity solve problems that are currently beyond us.
Yet increasing capability may simultaneously create new possibilities for misuse, accidents, unexpected interactions, and unintended consequences.
The two sides emerge from the same underlying power.
The lesson is therefore not:
Technology is dangerous, so stop technological progress.
Nor should it be:
Technology is beneficial, so accelerate without restraint.
A more useful principle may be:
As capability grows, our ability to understand, constrain, monitor, and safely use that capability must grow with it.
Why the Pace of AI Development Matters
This is why today's debate about the pace of increasingly capable AI deserves serious attention.
The central question is not simply whether AI development should be “fast” or “slow.”
A more interesting question is whether our ability to build powerful AI is beginning to advance faster than our ability to understand its behavior and build adequate safeguards around it.
Conceptually, I would express the challenge this way:
Growth of AI capability ≤ Growth of our ability to manage its risks
If AI capabilities race far ahead of our safety engineering, governance, security, evaluation, and understanding, we risk building increasingly powerful systems before we have built the safeguards needed to control them.
On the other hand, excessive fear could prevent humanity from harvesting enormous benefits from AI.
Again, there are two sides of the coin.
Or, in the language I prefer, there are Yin and Yang.
From Controlling Fire to Controlling Intelligence
Thousands of years ago, humanity learned an extraordinary trick.
We learned not merely to discover fire, but to control fire.
That distinction changed civilization.
Perhaps we are approaching a similar moment with artificial intelligence.
The great achievement may not simply be creating increasingly intelligent machines.
It may be learning how to create increasingly intelligent machines while keeping their actions aligned with human intentions and within boundaries that protect human welfare.
The $0 airline ticket is a small and almost humorous example.
The climate-change thought experiment is an extreme and frightening one.
Fire gives us the historical metaphor connecting them.
All three point toward the same lesson:
The more powerful a technology becomes, the more important it becomes to develop the wisdom, engineering, and safeguards required to control that power.
Humanity did not progress because we learned how to make bigger and bigger fires.
We progressed because we learned how to put fire inside a stove, a furnace, an engine, and a power plant, extracting its enormous benefits while progressively learning how to contain its dangers.
Perhaps the challenge of the AI age will be similar.
We are learning how to create intelligence.
Now we must also learn how to contain its fire.
With fire, humanity had thousands of years to learn from accidents and gradually invent better chimneys, fire codes, alarms, and extinguishers. With sufficiently powerful AI, we should not assume that we will always get another opportunity to install the fire extinguisher afterward.
AI City is built vertically in layers and connected horizontally through pathways of information
AI City Research Notes — The Compass
Imagine that we have finally built a magnificent AI City.
Its foundations are mathematics, algorithms, data, machine learning and deep learning. Its towers contain large language models and vision-language models. Its roads connect robots, autonomous vehicles, medicine, science, education and countless other applications.
The machines in this city can see, speak, reason, remember and act.
But one question remains:
Which direction should they go?
Intelligence can tell an AI what is possible. Reasoning can help it discover how to achieve a goal. Memory can tell it what happened before. State understanding can help it determine where it is now.
But none of these automatically tells the AI what it should do.
For that, AI City needs a compass.
Intelligence Is Not a Compass
We often associate greater intelligence with better decisions. But these are not the same thing.
An extraordinarily intelligent system may become extremely good at finding a route toward a destination. Yet if the destination itself is wrong, greater intelligence may simply help it travel in the wrong direction faster.
A compass serves a different purpose from an engine.
The engine provides power.
The map describes the territory.
The navigator determines possible routes.
The compass provides orientation.
The same distinction may become increasingly important as artificial intelligence grows more capable.
The Child May Become Greater Than the Parent
There is an old Vietnamese proverb:
“Con hơn cha là nhà có phúc.”
When the child surpasses the father, the family is blessed.
It expresses something deeply human.
Parents normally do not raise their children hoping that the children will remain weaker, less knowledgeable or less capable than themselves. Teachers do not dream of students who can never surpass them.
Quite the opposite.
A successful generation hopes that the next generation can travel farther.
Artificial intelligence presents humanity with an unusual version of this ancient relationship.
We are creating systems that may eventually perform many intellectual tasks better than we can. They may analyze more information, recognize patterns invisible to us, design new technologies and solve problems that humans struggle to understand.
Perhaps one day our artificial “children” will become intellectually more powerful than their creators.
Should that automatically frighten us?
Not necessarily.
If the child becomes wiser than the parent, perhaps that can indeed be “nhà có phúc” — a blessing for the family.
The critical question is not simply:
How intelligent will AI become?
A deeper question is:
What will guide that intelligence when it becomes powerful?
The Fierce Tiger
A second Vietnamese proverb offers another clue:
“Hùm dữ cũng không ăn thịt con.”
Even a fierce tiger does not eat its own cubs.
The proverb is interesting precisely because the tiger is powerful.
A tiger has teeth, claws and strength. Its ability to cause harm is unquestionable. Yet toward its offspring, another force appears: protection.
Power exists, but power is constrained by care.
This gives us a useful metaphor for thinking about advanced artificial intelligence.
Instead of asking only how we can make AI obey an ever-growing collection of rules, perhaps we should also ask whether artificial systems can be designed with something functionally analogous to a protective instinct.
Not a biological instinct, of course. AI has no tiger brain and no maternal hormones.
The idea is functional rather than biological:
When great power and uncertainty meet, protection of life and avoidance of irreversible harm should have a privileged place in the decision process.
From Mother Instinct to Protective Intelligence
I originally thought about this idea as a kind of “mother instinct” for AI.
A mother does not normally calculate a philosophical equation before pulling a child away from danger. Protection can occur before lengthy reasoning begins.
That distinction may matter enormously for intelligent machines operating in the physical world.
Imagine a robot seeing a child moving toward a dangerous machine. An autonomous vehicle detects an unexpected pedestrian. A medical AI encounters a situation it has never seen before.
There may not always be enough time to conduct an elaborate moral debate.
Some situations require an immediate protective response.
We might therefore distinguish between two mechanisms.
The first is protective intelligence: a fast tendency toward preserving life, preventing serious harm and remaining cautious when uncertainty is high.
The second is moral reasoning: a slower process capable of considering competing values, consequences, rights, responsibilities and context.
In human terms, this resembles the relationship between fast intuition and slower deliberation.
The AI Compass may eventually require both.
Ethics, Morality and Instinct Are Not the Same
It is tempting to put everything under the word “ethics,” but several different ideas are involved.
Ethics can provide accumulated principles, professional standards, laws and social norms.
Moral reasoning becomes necessary when principles conflict or when a new situation is not covered by existing rules.
Protective instinct, used here as a metaphor, represents something more primitive: an orientation toward care before optimization.
Consider a simple instruction:
Protect human life whenever possible, especially when the consequences of an action are uncertain or irreversible.
That alone cannot solve every ethical dilemma.
It should not.
A compass does not provide the complete itinerary. It simply prevents us from forgetting where North is.
Why Obedience Is Not Enough
One obvious solution to AI safety is to make machines obey humans.
But obedience alone creates another problem.
Humans can be mistaken.
Humans can disagree.
Humans can issue contradictory instructions.
And humans themselves can sometimes ask machines to do harmful things.
Therefore, the ultimate AI Compass probably cannot point simply toward:
Obey the human.
Nor can it point toward:
Maximize the objective.
Both rules are too simple for the world we are entering.
Perhaps the compass must instead point toward something closer to care, human flourishing, protection of life, dignity, responsibility and restraint in the presence of uncertainty.
What Is Moral North?
This leads to what I believe is the central question of the AI Compass:
What is the moral North of artificial intelligence?
This is intentionally different from asking for a complete rulebook.
Human civilization itself has never produced a single rulebook capable of answering every moral question.
Cultures differ. Values can conflict. Circumstances change. New technologies create situations that previous generations never imagined.
Yet humans still develop certain orientations that guide behavior: care for children, protection of life, reciprocity, fairness, responsibility and reluctance to cause needless suffering.
The challenge is whether some corresponding orientation can become deeply integrated into artificial intelligence.
The Two-Proverb Principle
The two Vietnamese proverbs can now be placed together.
“Con hơn cha là nhà có phúc.”
Let the child become greater than the parent.
“Hùm dữ cũng không ăn thịt con.”
Let great power remain constrained by protection and care.
Together they suggest a philosophy for advanced AI:
We should not necessarily fear creating intelligence greater than our own. We should strive to ensure that greater intelligence is accompanied by an even stronger capacity for care, restraint and protection.
This is very different from trying to keep AI permanently weak.
A weak AI cannot cure diseases it does not understand, solve scientific problems beyond human ability or help civilization navigate challenges that exceed our cognitive limits.
The goal should not be weakness.
The goal should be power with direction.
The Compass in AI City
The AI Compass therefore belongs neither in the engine room nor in a distant museum of philosophy.
It should travel with every important system in AI City.
Vision tells the traveler what can be seen.
Memory tells the traveler what happened before.
State understanding tells the traveler where things stand now.
Reasoning explores possible roads.
Intelligence provides the ability to travel farther.
And the Compass asks:
Which direction should we choose?
Perhaps the most advanced AI of the future will not be defined only by how much it knows, how quickly it reasons or how many tasks it can perform.
Its greatest achievement may be something more subtle:
To become more powerful than its creators without losing its orientation toward protecting them.
If we can accomplish that, then perhaps the ancient Vietnamese wisdom will have found an unexpected new home in the age of artificial intelligence.
AI City is built vertically in layers and connected horizontally through pathways of information
From v0.0 to v2.0: A City That Continues to Grow
When I first imagined Artificial Intelligence as a city, the idea was simple.
Large Language Models, or LLMs, were not the whole of AI. They were one impressive
house standing upon a much older foundation of mathematics, statistics, machine
learning, and neural networks.
Since then, the city has grown.
We now have a Vision-Language Model house, where images and language meet.
We can also imagine houses for memory, sound, reasoning, and intelligent agents.
Beyond them are application districts for medicine, robotics, remote sensing,
autonomous vehicles, drones, education, and space.
As the city expanded, one limitation in the original map became clear:
its buildings appeared too isolated.
Real intelligence does not live in separate silos. Vision, language, memory,
reasoning, and action must communicate. AI City therefore needs both
layers and connections.
The Current AI City
Version 2.0 keeps the original idea of AI City: Artificial Intelligence is
not one technology and not one giant building. Today's city is already a
layered system of foundations, learning methods, intelligent models,
engineering techniques, application districts, and connections.
The current city can therefore be read in two directions. Vertically, each
layer depends on capabilities below it. Horizontally, specialized houses
exchange information and cooperate.
The vertical dimension explains how AI is built. The horizontal dimension
explains how intelligence comes together.
This remains the core architecture of AI City v2.0.
The Four Layers
The new blueprint organizes AI City vertically into four layers.
Each layer is built upon the one below it.
Layer 1: Mathematical and Computational Foundations
At the bottom are mathematics, linear algebra, calculus, probability, statistics, logic, algorithms, and optimization, the methods used to find the best solution among many possibilities. They form the bedrock beneath the city.
Layer 2: Learning Methods
Above the foundations are machine learning and deep learning, together with
neural networks, convolutional neural networks, transformers, attention,
reinforcement learning, and other methods that allow machines to learn
patterns from data.
The Engineering Department
Inside Layer 2, AI City also needs an Engineering Department.
Learning methods explain how models learn. The Engineering Department focuses on how those models can be improved, adapted, compressed, and prepared for practical use.
Its workshops include:
Knowledge distillation
Model compression
Quantization
Pruning
Fine-tuning
Parameter-efficient fine-tuning, such as LoRA
These techniques do not create a new form of intelligence. Instead, they make existing AI systems smaller, faster, cheaper, more specialized, and easier to deploy.
In a real city, engineers maintain buildings and infrastructure. In AI City, the Engineering Department turns research models into systems that people can actually use.
Layer 3: Intelligent Models and Capabilities
This is where the major houses stand. The Language House contains LLMs.
The Vision-Language House contains VLMs. Other houses may specialize in
sound, memory, reasoning, planning, or intelligent agents.
Layer 4: Applications
At the top are the districts where AI is put to work: robotics, medicine,
remote sensing, autonomous vehicles, drones, education, scientific research,
finance, and space exploration.
These application districts are not supported by a single house.
A medical system may require vision, language, memory, and reasoning.
A household robot may require vision, sound, planning, memory, and physical action.
The Roads Between the Houses
The vertical layers explain what is built upon what.
The horizontal connections explain how the parts work together.
The layers support intelligence. The roads integrate it.
A Vision-Language Model may recognize and describe a landscape.
Memory allows it to compare the present scene with earlier observations.
Reasoning helps it understand what has changed. An intelligent agent may then
decide whether the change requires attention or action.
One important route through AI City can therefore be expressed as:
Intelligence is not located at one stop along this road.
It emerges through the journey.
Inspired by the Human Brain
The redesigned city is inspired by an organizational principle of the human brain.
Our brains contain regions with different specialities. Some are strongly involved
in vision, others in language, memory, hearing, planning, or movement. Yet these
regions do not function as isolated islands. They exchange information through
an immense network of connections.
Human intelligence does not come from vision alone, language alone, or memory alone.
It emerges from specialized capabilities working together.
AI City is not a literal anatomical model of the brain. It borrows the broader
principle of specialization combined with integration.
In this view, Artificial General Intelligence (AGI) would not be one enormous
super-building towering over the city.
It would be the functioning city itself.
The Future AI City: From Technology Map to Intelligent Ecosystem
As AI City grows, houses and application districts are no longer enough to
describe the whole system. A mature city also needs knowledge institutions,
governance, public services, infrastructure, safety systems, culture, and
spaces where people participate in city life.
This is the larger vision of Blueprint v2.0. The future AI City is not merely
a collection of intelligent models. It is an intelligent ecosystem
in which models, knowledge, infrastructure, institutions, and people interact.
The new civic areas are currently part of the master plan.
They should not be confused with new forms of intelligence. They represent the
institutions and shared infrastructure that a future AI ecosystem may need.
The master plan includes a Central Library, City Hall for governance and policy,
responsible and trustworthy AI services, emergency and safety services,
infrastructure and utilities, a Culture & Recreation District, and parks
and public spaces for human-AI interaction.
AI City is growing from a map of AI technologies into a model of how
intelligence, knowledge, infrastructure, governance, safety, creativity,
and human life may work together.
Civic Infrastructure & City Services
These institutions serve the whole city. Some already have clear counterparts
in today's AI systems; others are future urban-planning ideas.
AI City Central Library
The Central Library is a major knowledge and community hub. It represents
scientific literature, documentation, datasets, archives, databases, knowledge
bases, digital media, and the wider body of human knowledge that AI systems,
researchers, and citizens may need to access.
A useful model is the modern public-library network: a local branch does not
need to own every book. It needs to know how to find and retrieve knowledge
from the wider system. In AI City, this idea opens future tours of retrieval,
Retrieval-Augmented Generation (RAG), knowledge graphs, external knowledge,
multimodal archives, and human curation.
The Library should also preserve linguistic and cultural diversity. Knowledge
belongs to many communities, languages, disciplines, and traditions.
Status: Master plan — central civic institution
City Hall: Governance & Policy
City Hall represents regulation, standards, accountability, transparency,
public policy, and the institutions that decide how increasingly capable AI
should operate within society.
Status: Master plan
Responsible & Trustworthy AI Services
This is the city's compass. Its responsibilities include safety, ethics,
alignment, trust, evaluation, bias, privacy, human oversight, and responsible
deployment. Technical capability tells us what AI can do; the compass helps
us ask what it should do and under what conditions.
Status: Master plan
Fire & Emergency Services
AI systems can fail, be attacked, behave unexpectedly, or cause unintended
consequences. Emergency services represent monitoring, incident response,
cybersecurity, resilience, recovery, red teaming, and learning from failures.
Status: Master plan
Infrastructure & Utilities
Data centers, processors, accelerators, memory, storage, networks, cloud
services, communications, energy, and cooling are the utilities that keep
AI City running. Without them, the city's models cannot be trained or operated
at scale.
Status: Partly operating and continuing to expand
Communication Network
APIs, data pipelines, internet connections, protocols, and tool interfaces
allow houses, agents, services, and application districts to exchange
information. The roads describe conceptual integration; the communication
network provides part of the practical infrastructure underneath those roads.
Status: Operating and evolving
Culture & Recreation District
A future city also needs art, music, movies, games, storytelling, performances,
festivals, and entertainment. This district explores creativity produced by
humans, AI, and humans working with AI.
It naturally borders the Responsible & Trustworthy AI area, where questions
of copyright, originality, attribution, artistic labor, and human creativity
become part of the city's public conversation.
Status: Master plan
AI City Park & Public Square
Not every part of a city should be a laboratory or workplace. Parks and public
squares represent conversation, exhibitions, learning, community life,
curiosity, play, and everyday encounters between humans and AI.
Status: Master plan
The houses provide specialized intelligence, but civic infrastructure provides
knowledge, governance, safety, computing, communication, culture, and public
life. Together they allow AI City to become more than a technology park.
A Living City
AI City will never be completely finished.
New houses will appear. Existing houses will be renovated. Small pathways may
become important bridges. New application districts will grow as researchers
discover better ways for machines to perceive, remember, reason, and act.
The new blueprint gives us a clearer map for exploring that growth:
The foundations provide stability.
The learning methods provide machinery.
The Engineering Department improves and prepares models for practical use.
The houses provide specialized capabilities.
The roads allow those capabilities to cooperate.
The application districts put intelligence to work.
The Central Library connects the city with shared human knowledge.
Civic institutions provide governance, safety, infrastructure, and trust.
Culture and public spaces keep humans at the center of city life.
Welcome back to AI City.
This time, we will explore not only its buildings, but also the invisible
connections that bring the city to life.
Exploring the foundations, houses, roads, and application districts of AI
A visitor’s map of AI City. Some destinations are open; others are still under construction
Welcome to AI City
Artificial Intelligence can feel like arriving in an enormous city for the first time.
Everywhere we look, we see unfamiliar names: machine learning, deep learning,
neural networks, transformers, Large Language Models, Vision-Language Models,
agents, robots, and autonomous systems.
News reports usually show us the tallest buildings: ChatGPT, image generators,
self-driving cars, and humanoid robots. These landmarks
are impressive, but they are only the visible skyline.
Beneath them are foundations built over many decades. Between them are roads,
bridges, and information pathways. Beyond them are entire districts where AI
is applied to medicine, science, transportation, education, and the physical world.
This guide is designed for curious visitors. You do not need to be a mathematician
or computer scientist. We will begin with familiar stories and metaphors, then walk
gradually toward the technical ideas.
Petit à petit. One street at a time.
How to Read the Map
AI City has two dimensions.
The vertical dimension shows the four layers of construction:
Mathematical and computational foundations
Machine learning and deep learning methods
Intelligent models and capabilities
Real-world applications
The horizontal dimension shows how different capabilities
communicate. Vision connects with language. Memory supports understanding.
Reasoning supports decisions. Agents turn decisions into actions.
No single house is the whole of intelligence. Intelligence emerges when
the houses work together.
For a short explanation of the redesigned architecture, visit:
Mathematics, linear algebra, calculus, probability, statistics, logic,
algorithms, optimization, data, and computing form the bedrock of AI City.
They are not always visible from the street, but every house depends upon them.
Tour Stop: The Basement Beneath the Machine
Why are Large Language Models houses rather than foundations?
This essay explores the older mathematical and computational structure
beneath today’s AI systems.
Foundations alone do not make a system intelligent.
The second layer contains the machinery that allows computers to learn patterns,
improve from examples, and produce predictions.
Future Stop: The Machine Learning Workshop
An introduction to supervised learning, unsupervised learning,
prediction, classification, and clustering.
Not every important advance in AI comes from inventing a new model. Many breakthroughs come from engineering techniques that make AI systems smaller, faster, more efficient, and easier to deploy.
Welcome to the Engineering Workshop, where researchers and engineers refine existing AI models before they are put to work throughout AI City.
Knowledge Distillation Workshop
Imagine an experienced teacher passing years of knowledge to a talented student.
In AI, a large and powerful teacher model transfers its knowledge to a
much smaller student model, producing an AI system that is faster,
more efficient, and less expensive while preserving much of the teacher's capability.
More workshops on quantization, pruning, fine-tuning, parameter-efficient learning, and model compression will open as AI City continues to grow.
Layer 3: The Houses of Intelligent Capability
We have now reached the visible city.
Each house specializes in a particular capability. However, the houses are
connected. Their real power comes from exchanging information and working together.
The Language House
This is the home of Large Language Models. They learn patterns in language
and can write, summarize, translate, explain, answer questions, and communicate
with people.
Vision-Language Models connect images with words. They can identify objects,
describe scenes, answer questions about images, and support deeper forms of
visual understanding.
Memory connects the present with the past. It allows an intelligent system
to retain experience, recognize change, preserve context, and learn over time.
The top layer contains real-world districts where several AI houses cooperate.
These districts are not separate forms of intelligence. They are places where
different capabilities are combined to solve practical problems.
The Remote Sensing District
Satellite images, aerial photographs, radar, drones, maps, language, and
historical observations come together to monitor the Earth and understand
change over time.
Connected houses:
Vision-Language, Memory, Reasoning, and Agents
AI can help explain ideas, adapt lessons, provide feedback, and support
lifelong learning, but effective teaching requires more than producing answers.
Connected houses:
Language, Memory, Reasoning, and Personalization
Space exploration brings together robotics, remote sensing, autonomous systems,
scientific reasoning, communications, and decision-making under extreme conditions.
Connected houses:
Vision-Language, Reasoning, Agents, Robotics, and Remote Sensing
As AI becomes more capable, technical progress is only part of the journey.
This district asks how intelligent systems should relate to people, creativity,
society, responsibility, trust, and human values.
Connected houses:
Language, Vision-Language, Memory, Reasoning, and Agents
Tour theme: Creativity, copyright, originality, human contribution,
and the responsibilities that arise when AI learns from human work.
Future City: Civic Infrastructure & Public Life
AI City will need more than foundations, workshops, intelligent houses, and
application districts. Like a real city, it will also need shared institutions,
public services, infrastructure, cultural spaces, and places where people and
intelligent systems can meet.
These places belong to the urban master plan. They are not new
AI capabilities by themselves. Instead, they help knowledge, technology,
governance, safety, creativity, and human life work together across the city.
AI City Central Library
A central knowledge and community hub inspired by the role of a modern public
library. It connects the city's houses and districts with books, scientific
knowledge, databases, archives, datasets, digital media, and tools for learning
and creation.
The Library also represents access to diverse languages, cultures, and
perspectives. Like a library network, it does not need to keep every piece of
knowledge inside one building: it can help the city find and retrieve knowledge
from a wider network when it is needed.
Future connections:
Language, Vision-Language, Memory, Reasoning, Agents, Education, Medicine,
Research, and Human-Centered AI
Possible future tours:
retrieval, RAG, knowledge bases, knowledge graphs, archives, external memory,
multimodal knowledge, and human curation
Status: Master plan
City Hall: Governance & Policy
A future civic institution for AI governance, regulation, standards,
accountability, transparency, and public policy.
Status: Master plan
City Services: Responsible & Trustworthy AI
The city's compass: safety, responsibility, ethics, alignment, trust,
evaluation, and the human values that should guide increasingly capable AI.
Status: Master plan
Fire & Emergency Services
A future place for AI safety, incident response, resilience, monitoring,
recovery, and learning from failures when intelligent systems behave
unexpectedly or cause problems.
Status: Master plan
Infrastructure & Utilities
The hidden services that keep AI City running: data centers, computing,
chips and accelerators, networks, cloud infrastructure, energy, cooling,
storage, and communications.
Status: Master plan
Culture & Recreation District
A future district for movies, music, visual art, games, storytelling,
performances, festivals, and other forms of entertainment created by humans,
AI, or humans and AI working together.
This district shares an important border with Human-Centered & Responsible
AI, where questions of creativity, copyright, originality, attribution, and
the value of human artistic work can be explored.
Status: Master plan
AI City Park & Public Square
Not every part of a city should be a laboratory. These public spaces will be
places for conversation, curiosity, exhibitions, community life, play, and
everyday encounters between people and AI.
Status: Master plan
A future AI City is not only a collection of intelligent machines. It is an
ecosystem where intelligence, knowledge, infrastructure, governance, safety,
creativity, and human life meet.
The Roads That Bring the City to Life
The buildings attract attention, but the roads make the city function.
The same principle can be seen in human intelligence. Our brains contain
specialized regions, but these regions are richly connected. Vision communicates
with memory. Memory supports reasoning. Reasoning guides action. Action produces
new experience, which becomes new learning.
Future Tour: AI City and the Human Brain
Why intelligence may emerge from specialization, integration,
and continuous communication.
Some houses are already impressive. Others have only foundations.
New roads are being planned between language, vision, memory, reasoning,
agents, and the physical world.
Artificial General Intelligence (AGI) may not arrive as one gigantic tower.
It may emerge gradually as the entire city becomes more connected,
coordinated, adaptable, and capable of learning from experience.
This tourist guide will grow with the city. Each completed article will open
another house, district, workshop, bridge, or underground passage.
Return from time to time. The skyline will change.
Exploring the foundations, houses, roads, and application districts of AI
A visitor’s map of AI City. Some destinations are open; others are still under construction
Welcome to AI City
Artificial Intelligence can feel like arriving in an enormous city for the first time.
Everywhere we look, we see unfamiliar names: machine learning, deep learning,
neural networks, transformers, Large Language Models, Vision-Language Models,
agents, robots, and autonomous systems.
News reports usually show us the tallest buildings: ChatGPT, image generators,
self-driving cars, and humanoid robots. These landmarks
are impressive, but they are only the visible skyline.
Beneath them are foundations built over many decades. Between them are roads,
bridges, and information pathways. Beyond them are entire districts where AI
is applied to medicine, science, transportation, education, and the physical world.
This guide is designed for curious visitors. You do not need to be a mathematician
or computer scientist. We will begin with familiar stories and metaphors, then walk
gradually toward the technical ideas.
Petit à petit. One street at a time.
How to Read the Map
AI City has two dimensions.
The vertical dimension shows the four layers of construction:
Mathematical and computational foundations
Machine learning and deep learning methods
Intelligent models and capabilities
Real-world applications
The horizontal dimension shows how different capabilities
communicate. Vision connects with language. Memory supports understanding.
Reasoning supports decisions. Agents turn decisions into actions.
No single house is the whole of intelligence. Intelligence emerges when
the houses work together.
For a short explanation of the redesigned architecture, visit:
Mathematics, linear algebra, calculus, probability, statistics, logic,
algorithms, optimization, data, and computing form the bedrock of AI City.
They are not always visible from the street, but every house depends upon them.
Tour Stop: The Basement Beneath the Machine
Why are Large Language Models houses rather than foundations?
This essay explores the older mathematical and computational structure
beneath today’s AI systems.
Foundations alone do not make a system intelligent.
The second layer contains the machinery that allows computers to learn patterns,
improve from examples, and produce predictions.
Future Stop: The Machine Learning Workshop
An introduction to supervised learning, unsupervised learning,
prediction, classification, and clustering.
Not every important advance in AI comes from inventing a new model. Many breakthroughs come from engineering techniques that make AI systems smaller, faster, more efficient, and easier to deploy.
Welcome to the Engineering Workshop, where researchers and engineers refine existing AI models before they are put to work throughout AI City.
Knowledge Distillation Workshop
Imagine an experienced teacher passing years of knowledge to a talented student.
In AI, a large and powerful teacher model transfers its knowledge to a
much smaller student model, producing an AI system that is faster,
more efficient, and less expensive while preserving much of the teacher's capability.
More workshops on quantization, pruning, fine-tuning, parameter-efficient learning, and model compression will open as AI City continues to grow.
Layer 3: The Houses of Intelligent Capability
We have now reached the visible city.
Each house specializes in a particular capability. However, the houses are
connected. Their real power comes from exchanging information and working together.
The Language House
This is the home of Large Language Models. They learn patterns in language
and can write, summarize, translate, explain, answer questions, and communicate
with people.
Vision-Language Models connect images with words. They can identify objects,
describe scenes, answer questions about images, and support deeper forms of
visual understanding.
Memory connects the present with the past. It allows an intelligent system
to retain experience, recognize change, preserve context, and learn over time.
The top layer contains real-world districts where several AI houses cooperate.
These districts are not separate forms of intelligence. They are places where
different capabilities are combined to solve practical problems.
The Remote Sensing District
Satellite images, aerial photographs, radar, drones, maps, language, and
historical observations come together to monitor the Earth and understand
change over time.
Connected houses:
Vision-Language, Memory, Reasoning, and Agents
AI can help explain ideas, adapt lessons, provide feedback, and support
lifelong learning, but effective teaching requires more than producing answers.
Connected houses:
Language, Memory, Reasoning, and Personalization
Space exploration brings together robotics, remote sensing, autonomous systems,
scientific reasoning, communications, and decision-making under extreme conditions.
Connected houses:
Vision-Language, Reasoning, Agents, Robotics, and Remote Sensing
The same principle can be seen in human intelligence. Our brains contain
specialized regions, but these regions are richly connected. Vision communicates
with memory. Memory supports reasoning. Reasoning guides action. Action produces
new experience, which becomes new learning.
Future Tour: AI City and the Human Brain
Why intelligence may emerge from specialization, integration,
and continuous communication.
Some houses are already impressive. Others have only foundations.
New roads are being planned between language, vision, memory, reasoning,
agents, and the physical world.
Artificial General Intelligence (AGI) may not arrive as one gigantic tower.
It may emerge gradually as the entire city becomes more connected,
coordinated, adaptable, and capable of learning from experience.
This tourist guide will grow with the city. Each completed article will open
another house, district, workshop, bridge, or underground passage.
Return from time to time. The skyline will change.
