By the beginning of the personal computer era, humanity had already crossed an extraordinary technological threshold.

Machines had become programmable. Information could be translated into digital form, and computers could calculate, store records, manipulate text, process images and perform entirely different tasks depending on the software running on them. Yet there remained a significant limitation: a computer was still most useful for the information available to that computer.

A document stored on one machine remained there unless someone deliberately transferred it. A program existed wherever a copy had been installed. An organisation might have an extraordinarily powerful computer system, but much of its value remained contained within the boundaries of that institution.

The next transformation would begin when computers learned to communicate.

Connecting machines seems almost obvious today because networking has become part of the invisible infrastructure of modern life. We send messages without considering the route they take, open websites hosted thousands of kilometres away, transfer money between institutions, stream video from remote data centres and use software that may not actually be running on the device in our hands.

For most of computing history, none of that was inevitable. The internet changed the world because it transformed computing from an activity performed by individual machines into an activity performed across networks. The World Wide Web then made those networks navigable by ordinary people, while smartphones and cloud computing eventually made them almost continuously accessible.

The result was larger than faster communication. Once computers became connected, markets became connected, information became globally distributable, software became remotely accessible and human activity began producing enormous quantities of data. The digital economy emerged from the network.

Computers First Had to Learn How to Talk

The earliest computers were extraordinarily expensive machines, and researchers naturally wanted ways to use them more efficiently. During the 1960s, computer scientists began investigating how machines in different locations might communicate and share resources. One of the central problems was that traditional telephone networks had been designed primarily around continuous connections, with a communication path effectively reserved while two people spoke.

Computer communication behaved differently. Machines often needed to send bursts of data rather than maintain a continuous conversation, which meant that reserving an entire communications channel while a computer was momentarily silent would waste capacity.

The emerging solution was packet switching. Instead of treating a message as one continuous transmission, information could be divided into smaller packets that travelled through a network and were then reassembled at their destination. Work on the underlying concepts emerged from several researchers and institutions during the 1960s, including efforts associated with Leonard Kleinrock, Paul Baran and Donald Davies.

The principle was powerful because network capacity could be shared. Different computers could transmit packets across the same infrastructure without requiring a dedicated circuit for every communication, and that idea would become one of the foundations of modern digital networking.

ARPANET Turned Networking Into a Working System

In the late 1960s, the United States Advanced Research Projects Agency funded the development of a computer network that became known as ARPANET. By the end of 1969, four host computers were connected to the initial network. They were located at research institutions rather than spread across a vast public system, but something important had happened: computers in different places could communicate through a packet switched network.

The significance was not simply that one machine could send information to another. Researchers could increasingly imagine computing resources as part of a connected environment rather than as isolated installations.

That change soon produced an unexpected lesson. People wanted to communicate with people just as much as computers needed to communicate with computers. Electronic mail became one of ARPANET’s most important early applications. In 1972, Ray Tomlinson’s work on network email helped establish a practical way of exchanging messages between users on different host computers, and email quickly became one of the network’s dominant uses.

This pattern would repeat throughout internet history. Engineers built networks to connect machines, but some of their most transformative applications emerged from connecting human beings.

A Network Was Useful but a Network of Networks Was Revolutionary

ARPANET proved that computer networking could work, but another problem soon became apparent. There would not be only one network. Universities could build networks, governments could operate networks, companies could create internal networks, and different technologies could use different technical standards.

The greater challenge was therefore not simply connecting computers within a network. It was connecting different networks to one another.

During the 1970s, researchers including Vint Cerf and Robert Kahn developed the architecture that evolved into TCP/IP, the protocol suite that became fundamental to the modern internet. The idea was elegant: individual networks would not need to become identical before they could communicate. Each could operate internally according to its own design while common protocols allowed information to pass between them.

TCP helped manage the transmission and reassembly of information, while IP handled addressing and routing across interconnected networks. On 1 January 1983, ARPANET formally transitioned from its earlier Network Control Protocol to TCP/IP, helping establish the architecture of an internet in the literal sense of the word, a system capable of connecting networks to other networks.

The distinction is important. A network connects computers, while the internet connects networks. That architecture allowed the system to grow without requiring one organisation to design and control every machine that might eventually join it.

The Internet Needed an Address Book

As the network expanded, another ordinary sounding problem became increasingly difficult. People needed a practical way to find computers. Numerical internet addresses could identify machines, but remembering long strings of numbers was hardly a sensible way to build a system intended to grow.

The Domain Name System emerged during the early 1980s to solve this problem. Instead of requiring users to remember numerical addresses, the system could translate human readable domain names into the numerical addresses computers used to locate one another. The initial specifications for DNS were published in 1983 and were later refined into the system that still underpins internet naming today.

This may seem like a minor administrative development beside inventions such as the computer or the microprocessor, but infrastructure becomes useful when complexity disappears from the experience of the person using it. Drivers do not need to understand combustion chemistry to operate a car, just as people switching on a light do not need to understand the electrical grid.

Likewise, ordinary internet users could eventually navigate a global network without understanding packet routing or numerical addressing. The technical machinery could disappear behind names.

Research Networks Began Becoming Infrastructure

For much of its early history, computer networking remained closely connected to government, defence and academic research. That gradually changed.

In 1986, the United States National Science Foundation launched NSFNET to connect researchers with supercomputer centres. The network expanded rapidly and became an important backbone for the growing internet. According to the National Science Foundation, NSFNET connected roughly 2,000 computers in 1986 and more than two million by 1993.

The important point is not the particular number of machines but the direction of travel. Networking was escaping its experimental origins and becoming infrastructure. Researchers could communicate across institutions, universities could connect their local networks and information could move between organisations that had previously operated separate computing environments.

Yet even at this stage, using networked information could remain technically demanding. The internet existed, but what ordinary people needed was a much easier way to navigate it.

The Web Made the Internet Understandable

The internet and the World Wide Web are often treated as though they are the same thing. They are not. The internet is the underlying network infrastructure through which computers communicate, while the Web is one system built on top of that infrastructure.

That distinction helps explain why Tim Berners-Lee’s work at CERN was so consequential. In 1989, Berners-Lee proposed a system that could help researchers manage and share information across different computer systems. His solution combined several ideas, including hypertext documents, web addresses and a communications protocol that allowed a user to request information from another computer. The result became the World Wide Web.

Its brilliance lay partly in the hyperlink. A document did not have to stand alone; it could point towards another document, which could point to another. Information could form a web of relationships that users could navigate by following links rather than knowing where each underlying file physically existed.

The structure felt natural because it mirrored something human beings already understood. One idea leads to another, and the Web converted that intellectual relationship into an architecture for digital information. In 1993, CERN made the Web software available on a royalty free basis, removing an important barrier to its spread.

The network now had an information system capable of expanding with it.

The Browser Opened the Door

The early Web remained a technical environment, but graphical browsers began changing who could use it. NCSA Mosaic, released in 1993, was particularly important because it provided a comparatively approachable graphical interface and was made available on widely used computer platforms. Images could appear alongside text, links could be followed visually and browsing began to feel less like operating a technical system and more like exploring information.

That shift mattered because technologies become socially transformative when ordinary people can use them without understanding the machinery underneath. A user did not need to know TCP/IP; the user needed to click. A user did not need to understand a server’s physical location; the user needed an address.

The complexity of global networking began disappearing behind the browser. As that happened, businesses started asking an inevitable question: could the network become a marketplace?

Commerce Discovered That Distance Had Changed

Traditional commerce is constrained by geography. A physical shop can display only the goods that fit within its space, its potential customers are heavily influenced by where the shop is located, and reaching more customers normally requires opening more premises or distributing goods through other retailers.

The Web altered that equation. In 1995, Amazon opened its online bookstore to customers on the Web. Its catalogue could offer vastly more titles than a conventional physical bookstore because the website was not constrained by retail shelf space in the same way. Within its first weeks, Amazon was already reporting orders from customers in dozens of countries.

That same year, Pierre Omidyar launched the service that became eBay, allowing strangers to buy and sell goods through an online marketplace.

These companies represented more than new retailers. They demonstrated new economic structures. The internet could aggregate demand across enormous geographic areas, so a specialised seller no longer needed enough customers within one town if potential buyers could be reached across an entire country or, eventually, the world. A marketplace did not necessarily need to own the goods being sold if it could connect buyers and sellers.

The network itself could become part of the business model.

The Digital Economy Began Rewarding Connection

Industrial economics had often rewarded scale because larger factories could manufacture enormous quantities of products efficiently. The network economy introduced another kind of scale because connection itself could create value.

A telephone network is more useful when more people can be reached through it. An online marketplace can become more useful when it has more buyers because sellers gain access to greater demand, while more sellers can make the marketplace more attractive to buyers. Similar dynamics later became important in search, social networks, payment systems, app stores and many other digital businesses.

The computer had already lowered the cost of processing information. The internet lowered the cost of moving it, and that combination changed what businesses could coordinate.

Companies could communicate with suppliers electronically, customers could compare products without travelling between shops, financial transactions could increasingly occur through networks, software could be distributed digitally and entire services could be delivered through a screen.

The industrial economy had organised factories around flows of materials. The digital economy increasingly organised businesses around flows of information.

Search Became Necessary Because Abundance Created a New Problem

The early digital world had been constrained by scarcity because there was not enough information available through networks. The Web eventually created the opposite problem: there was too much.

As websites multiplied, simply knowing that information existed somewhere on the internet became less useful. People needed ways to find the particular information that mattered to them. Directories initially attempted to organise the Web manually, while search engines increasingly automated the process by indexing pages and ranking possible results.

This represented another subtle transformation. The internet had made information accessible, while search made abundance usable. That difference would later become economically enormous because controlling the path between a question and the information that answered it created opportunities for advertising, recommendation and eventually detailed observation of user behaviour.

The network was no longer simply carrying information. It was beginning to organise attention.

The Internet Escaped the Desk

For much of the Web’s early history, going online was something people deliberately did. They sat at a computer, connected to a network, opened a browser and entered the digital world.

Mobile computing changed that relationship. Mobile phones had been developing internet capabilities before the modern smartphone era, but the convergence of powerful processors, mobile data networks, touch interfaces and sophisticated browsers eventually made the internet genuinely practical as a constant companion.

The arrival of devices such as Apple’s iPhone in 2007 illustrated the transformation clearly. Apple explicitly presented the device not simply as a phone but as a handheld internet communications device capable of email, web browsing, search and maps.

This altered the geography of computing. The internet was no longer primarily somewhere you went; it increasingly travelled with you. A connected device could know where it was, capture photographs, send messages, browse the Web, access maps and eventually run an expanding ecosystem of applications linked to remote services.

The distinction between online life and ordinary life began to weaken.

Computing Moved Back Into Enormous Machines

There is an interesting irony in the history of computing. The personal computer revolution moved computing power away from enormous central systems and placed it on individual desks, while the internet eventually helped move much of that computation back into enormous centralised facilities.

The difference was that users no longer needed to own the machines. Cloud computing allowed storage and processing capacity to be delivered remotely through networks.

In 2006, Amazon Web Services launched services including S3 for remote storage and EC2 for remotely accessible computing capacity. Rather than purchasing and maintaining all of their own physical server infrastructure, organisations could increasingly rent computing resources when required.

This changed the economics of building digital businesses. A company that once needed substantial capital to buy servers before launching a service could increasingly acquire computing capacity as demand grew.

Software itself also began changing. Instead of purchasing a program, installing it on one machine and periodically replacing it with a newer version, users could access applications through the internet while developers updated the underlying service remotely.

The network was beginning to absorb the computer.

Every Click Began Leaving Data Behind

The internet’s most consequential product may not have been the website, the email or even online commerce. It may have been data.

Once human activity moved onto digital networks, actions that had previously disappeared into everyday life could leave machine readable traces. A search generated data, a purchase generated data, opening a webpage generated data, sending a message generated data, using a map generated data and watching a video generated data. Carrying a connected smartphone could additionally generate information about location, movement and patterns of use.

Companies discovered that these records could be analysed. They could reveal what customers bought, what they searched for, what they ignored, where they stopped reading, which advertisements they clicked and how groups of people behaved across enormous digital systems.

Information had become valuable during the computer revolution. The internet made information continuous.

The digital economy therefore evolved beyond selling products through computers. Increasingly, understanding activity on the network became an economic resource in its own right.

The Internet Created the Conditions for the Next Revolution

This is where the history of the internet begins merging with the history of artificial intelligence. Modern AI did not emerge merely because researchers developed better algorithms; it required an entire technological environment.

The computer revolution supplied programmable machines. Semiconductor development supplied enormous and increasingly affordable computational power. Digitisation turned text, photographs, sound, video and records into data. The internet connected machines across the world, and billions of people and organisations then began generating still more digital information through those networks. Cloud computing concentrated enormous computing resources in data centres that could be accessed remotely.

By the early twenty first century, humanity had created something historically unusual: a civilisation in which vast amounts of human activity were being represented digitally, stored computationally and transmitted through global networks.

The infrastructure required for modern artificial intelligence was beginning to exist.

The Network Became Part of Society

It is tempting to think of the internet as another invention alongside the steam engine, automobile or telephone, but that description no longer captures what it became. The internet evolved into infrastructure.

Businesses depend upon it, as do banks, governments, universities, media and transport systems. Families, friendships and increasingly ordinary social relationships operate partly through it.

This is the mark of a genuinely transformative technology. Eventually we stop noticing the technology itself and begin noticing only what happens when it fails. The electrical grid disappeared into the background of industrial civilisation because electricity became expected, and the internet is undergoing the same transformation.

It is becoming less a product people consciously use and more an environment within which modern economic and social life takes place. That transformation also means its weaknesses have become society’s weaknesses.

Cybercrime can cross borders, false information can move at extraordinary speed, digital platforms can accumulate enormous economic power, surveillance can operate at unprecedented scale and network failures can disrupt services far beyond the technology sector.

The same connectivity that creates enormous opportunity also creates dependence. Industrial revolutions have always worked this way because they solve old constraints and create new ones.

The World Had Become Ready for Machines That Learn

The story that began with isolated computers had reached an entirely different scale. First, machines learned to process information. Then information became digital. Computers were connected, the Web made their information navigable, commerce turned networks into markets, smartphones made connectivity continuous and cloud computing made enormous computational resources remotely available. Every stage produced more data than the one before it.

The Third Industrial Revolution had therefore done more than create computers and the internet. It had created an information environment in which humanity possessed vast computing capacity, enormous stores of digital data and global networks capable of moving information almost instantly.

Those three developments would converge in the next technological transformation. For decades, computer scientists had been trying to build machines capable of doing something more ambitious than following explicit instructions. They wanted machines that could recognise patterns, improve from examples and perform tasks associated with intelligence.

For much of computing history, those ambitions repeatedly collided with limitations in processing power, available data and algorithmic capability. The digital revolution gradually weakened those constraints.

The world had connected its computers, digitised its information and generated unprecedented quantities of data. Now machines were about to begin learning from it.

Next in The Trend Report

Machines Learned to Learn

Artificial Intelligence Emerged From Decades of Computing

Resources

This article draws principally on the Internet Society’s histories of ARPANET, packet switching, TCP/IP, email and the Domain Name System; the United States National Science Foundation’s history of NSFNET and the commercial internet; CERN’s historical archive on Tim Berners-Lee and the creation of the World Wide Web; the National Center for Supercomputing Applications and Computer History Museum material on Mosaic and the early Web; archival material from Amazon and eBay on early electronic commerce; Apple historical material on the emergence of smartphone based internet access; and Amazon Web Services material on the emergence of large scale commercial cloud computing.