YOUNG WEIDAR WONG IN-HOON: "FIVE LAYERS OF CAKE" BY AI

Original title:AI Is a Five-Layer Cake

Original by Nvidia

Original by Peggy, Block Beats

 

COMPOSERS PRESS: ARTIFICIAL INTELLIGENCE IS EVOLVING FROM A CUTTING-EDGE TECHNOLOGY TO AN INFRASTRUCTURE THAT SUPPORTS THE FUNCTIONING OF THE MODERN ECONOMY. IN ITS FIRST LONG ARTICLE, PUBLISHED IN ITS OFFICIAL ACCOUNT, YING WEIDA TRIED TO SYSTEMATIZE THE STRUCTURE OF AI'S INDUSTRY FROM THE FIRST PRINCIPLE: FROM ENERGY AND CHIPS TO THE DATA CENTRE INFRASTRUCTURE, TO MODELS AND APPLICATIONS, IT CONSTITUTES A COMPLETE FIVE-STOREY TECHNOLOGY WAREHOUSE。

THE ARTICLE NOTES THAT AI IS NOT JUST A COMPETITION FOR SOFTWARE OR MODELS, BUT A GLOBAL INDUSTRIAL CONSTRUCTION INVOLVING ENERGY, COMPUTING, MANUFACTURING AND APPLICATIONS, THE SCALE OF WHICH COULD BE ONE OF THE LARGEST INFRASTRUCTURE EXPANSIONS IN HUMAN HISTORY. THROUGH THIS PERSPECTIVE OF FIVE-STOREY CAKES, WE TRY TO SHOW THAT THE REAL MEANING OF AI IS NOT JUST SMARTER SOFTWARE, BUT AN INFRASTRUCTURE REVOLUTION ON A SCALE COMPARABLE TO ELECTRICITY AND THE INTERNET。

The following is the original text:

Artificial intelligence is one of the most powerful forces shaping the world today. It is not a smart application or a single model, but an infrastructure, as important as electricity and the Internet。

AI OPERATES ON REAL HARDWARE, REAL ENERGY AND REAL ECONOMIC SYSTEMS. IT TRANSFORMS RAW MATERIALS INTO "SMART" FOR SCALE PRODUCTION. EVERY COMPANY WILL USE IT, EVERY COUNTRY WILL BUILD IT。

IT WOULD BE HELPFUL TO UNDERSTAND WHY AI WOULD PROCEED IN THIS WAY, STARTING FROM THE FIRST PRINCIPLE, TO SEE WHAT FUNDAMENTAL CHANGES HAVE TAKEN PLACE IN THE FIELD OF COMPUTING。

From "prefabricated software" to "real-time-generated intelligence."

IN MOST OF THE HISTORY OF COMPUTER DEVELOPMENT, SOFTWARE IS "PREFABRICATED". HUMANS FIRST DESCRIBE AN ALGORITHM, THEN COMPUTERS FOLLOW INSTRUCTIONS. THE DATA MUST BE CAREFULLY STRUCTURED, PLACED IN THE TABLE AND REMOVED THROUGH PRECISE QUERIES. SQL IS INDISPENSABLE BECAUSE IT MAKES THIS WHOLE SYSTEM WORK。

AND AI BROKE THIS PATTERN。

For the first time, we have a computer that understands unstructured information. It can look at images, read texts, hear voices and understand what they mean; it can reason context and intent. More importantly, it produces intelligence in real time。

Each response is new. Every answer depends on the context you provide. This is no longer a software search of existing commands from a database, it is a real-time reasoning of the software and generating intelligence as needed。

Just as intelligence is created in real time, the entire computing vault that supports it must also be recreated。

AI AS INFRASTRUCTURE

IF YOU LOOK AT AI FROM AN INDUSTRY PERSPECTIVE, IT CAN ACTUALLY BREAK DOWN INTO A FIVE-TIERED STRUCTURE。

Energy

The bottom is energy。

real-time generation of intelligence requires real-time generation of electricity. each token generation means that electronics are moving, heat is being managed, energy is being converted into computing power。

UNDER THIS LAYER, THERE IS NO ABSTRACTION. ENERGY IS THE FIRST PRINCIPLE OF AI INFRASTRUCTURE AND THE FUNDAMENTAL CONSTRAINT IN DETERMINING HOW MUCH INTELLIGENCE THE SYSTEM CAN PRODUCE。

Chips

Energy above the chip. These processors are designed to transform energy into computing capacity under highly efficient, large-scale conditions。

AI TASK LOADS REQUIRE SIGNIFICANT PARALLEL COMPUTING CAPACITY, HIGH BANDWIDTH MEMORY AND HIGH SPEED INTERCONNECTION. THE PROGRESS OF THE CHIP LAYER DETERMINES THE SPEED AT WHICH AI EXPANDS, AND HOW SMART CAN EVENTUALLY BECOME CHEAP。

Infrastructure

The chip is the infrastructure. This includes land, power transmission, heat-dispersion systems, construction works, network systems, and movement control systems that organize tens of thousands of processors into one machine。

THESE SYSTEMS ARE ESSENTIALLY AI FACTORIES. THEY ARE NOT DESIGNED TO STORE INFORMATION, BUT TO CREATE INTELLIGENCE。

Models

INFRASTRUCTURE IS A MODEL. AI MODELS UNDERSTAND ALL TYPES OF INFORMATION: LANGUAGE, BIOLOGY, CHEMISTRY, PHYSICS, FINANCE, MEDICINE, AND THE REAL WORLD ITSELF。

LANGUAGE MODELS ARE ONLY ONE SUCH CATEGORY. ONE OF THE MOST TRANSFORMATIVE TASKS IS TAKING PLACE IN THE FOLLOWING AREAS: PROTEIN AI, CHEMISTRY AI, PHYSICAL SIMULATION, ROBOTICS, AUTONOMOUS SYSTEMS

Application

The top layer is the application layer, where economic values really come from. For example, drug discovery platforms, industrial robots, law Copilot, auto-driving cars。

AN AUTO-DRIVING CAR IS ESSENTIALLY AN "AI APPLICATION CARRIED BY A MACHINE" AND A HUMAN TYPE ROBOT IS AN "AI APPLICATION CARRIED BY A BODY". THE BOTTOM VAULT IS THE SAME, BUT THE FINAL PATTERN IS DIFFERENT。

SO THIS IS THE FIVE-TIERED STRUCTURE OF AI: ENERGY CHIPS, INFRASTRUCTURE, MODEL APPLICATIONS. EACH SUCCESSFUL APPLICATION WILL BRING DOWN ALL LEVELS TO THE BOTTOM OF THE POWER PLANT THAT POWERS IT。

An infrastructure building still in the early stages

We have just begun this construction. The current level of investment was just hundreds of billions of dollars, while in the future it was still necessary to build trillions of dollars in infrastructure。

GLOBALLY, WE ARE SEEING CHIPS FACTORIES, COMPUTER ASSEMBLY PLANTS, AI FACTORIES。

It was built on an unprecedented scale. This is becoming one of the largest infrastructure developments in human history。

AI AGE LABOUR NEEDS

The labour force required to sustain this construction is enormous。

AI PLANT NEEDS: ELECTRICIANS, PLUMBERS, PLUMBERS, STEEL STRUCTURAL WORKERS, NETWORK TECHNICIANS, EQUIPMENT INSTALLERS, TRANSPORTERS

These are highly skilled and well-paid jobs and are currently in a critical shortage. Participation in this transition does not necessarily require a PhD in computer science。

AT THE SAME TIME, AI IS PROMOTING PRODUCTIVITY GAINS IN THE KNOWLEDGE ECONOMY. TAKE RADIOLOGY, FOR EXAMPLE. AI HAS BEGUN TO INTERPRET AUXILIARY MEDICAL IMAGES, BUT THE DEMAND FOR RADIOLOGISTS CONTINUES TO GROW。

It's not a contradiction。

THE REAL DUTY OF RADIOLOGISTS IS TO TAKE CARE OF PATIENTS, AND READING FILMS IS ONLY ONE OF THEM. WHEN AI TAKES OVER MORE AND MORE REPETITIVE TASKS, DOCTORS CAN INVEST MORE TIME IN JUDGEMENT, COMMUNICATION AND TREATMENT。

Hospitals are more efficient and serve more patients and therefore require more staff. Productive capacity and capacity to generate growth。

What has changed in the past year

IN THE PAST YEAR, AI HAS CROSSED A CRITICAL THRESHOLD。

Models are good enough to make a real difference in large-scale scenarios。

:: Significant improvement in reasoning

• Considerable decrease in hallucinations

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

FOR THE FIRST TIME, AN AI-BASED APPLICATION BEGAN TO CREATE REAL ECONOMIC VALUE。

Prominent product market matching has emerged in the areas of drug research and development, logistics, passenger service, software development, manufacturing

These applications are pressing the entire technology vault at the bottom。

Role of open source models

OPEN SOURCE MODELS PLAY A KEY ROLE. MOST AI MODELS IN THE WORLD ARE FREE OF CHARGE. RESEARCHERS, START-UPS, FIRMS AND EVEN COUNTRIES AS A WHOLE RELY ON OPEN-SOURCE MODELS TO COMPETE IN ADVANCED AI。

When open source models reach the technological frontier, they not only change software but also activate the demand for the entire technology stack。

DeepSeek-R1 is a typical example. By making a robust reasoning model widely available, it has contributed to the rapid growth of the application layer, while increasing the demand for training algorithms, infrastructure, chips and energy。

What does that mean

When you see AI as infrastructure, everything becomes clear. AI may start with Transformer and large language models, but it's much more than that。

It is an industrial transformation that will reshape:

Energy production and consumption patterns

The way in which factories are built

Organization of work

Patterns of economic growth

AI FACTORIES ARE BUILT BECAUSE INTELLIGENCE CAN NOW BE GENERATED IN REAL TIME. THE CHIP HAS BEEN REDESIGNED BECAUSE EFFICIENCY DETERMINES THE SPEED OF INTELLIGENT EXPANSION. ENERGY IS CENTRAL BECAUSE IT DETERMINES HOW MUCH INTELLIGENCE THE SYSTEM CAN PRODUCE AT BEST. THE APPLICATION BROKE OUT BECAUSE THE MODEL HAD FINALLY CROSSED THE THRESHOLD OF "WORKABLE SIZE"。

Each layer is reinforcing the other。

That is why the construction is so large, why it affects so many industries at the same time, and why it is not limited to one country or another。

EACH COMPANY WILL USE AI。

EVERY COUNTRY WILL BUILD AI。

We are still at an early stage。

A significant amount of infrastructure has not yet been built, a large proportion of the labour force has not yet been trained and opportunities have not been realized。

But the direction is very clear。

Artificial intelligence is becoming the basic infrastructure of the modern world。

And the choices we have made today, the speed of construction, the breadth of participation and the responsibility for deployment will determine what this era will eventually become。