What the Information Age Left Us

Conclusion and Opening Toward the Following Period (2010-present)

Here we are at the end of sixty-five years of journey — the most accelerated of our crossing. From the CSIRAC in Sydney to the WEIZAC in Rehovot, from the ruins of Berlin to Silicon Valley garages, from M-Pesa in Kenya to TSMC in Taiwan, from the Dartmouth conference to ImageNet — six continents, and everywhere the same paradox: leaps across the abyss and falls into oblivion, frugal innovations and pharaonic projects, summers of euphoria and winters of disillusionment.

This period — from 1945 to 2010 — was the one where artificial intelligence took shape. The conceptual tools forged during the previous era became incarnated in ever more powerful machines. The Dartmouth dream — simulating every aspect of human intelligence — passed through cycles of promise and disappointment before being reborn, transformed, in deep learning. Shannon and Nakashima's circuits became microprocessors. Turing's universal machine became the personal computer, then the smartphone, then the global data center.

But this period was also one of reversals. Latecomers became pioneers. Africa invented mobile payments before the West. India leaped over hardware to become a software giant. Taiwan invented an industrial model — the pure-play foundry — that redrew the maps of technological power. The parallel paths traced for decades began to converge toward the same horizon: Asian dominance in semiconductors and the emergence of an artificial intelligence with multiple voices.

What Unites: The Common Threads of an Era of Acceleration

Five threads run through this period, weaving a fabric found from one continent to another.

The leap across the abyss.

Everywhere, nations and peoples found ways to skip the steps that others had laboriously climbed. Africa had no fixed telephone network — it jumped directly to mobile, then to payment by phone. India had missed the hardware turn — it leaped toward software services when the Y2K bug created global demand. Israel had neither electronics industry nor computing tradition — it built one of the world's first computers and became the "Startup Nation." These leaps were not shortcuts. They were inventions — new ways of solving problems that pioneers had not imagined. Technological leapfrogging is not accelerated imitation. It is creation adapted to context.

The cycles of hope and disenchantment.

Artificial intelligence experienced summers and winters. The founding summer of Dartmouth (1956), where twenty-one researchers believed they could simulate human intelligence in one generation. The first winter (1974-1980), when unfulfilled promises triggered budget cuts. The expert systems summer (1980-1987), when Japan launched its ambitious Fifth Generation project. The second winter (1987-1993), when Moore's Law made conventional machines faster than specialized architectures. And finally, from 2006, the third summer — that of deep learning, which still continues. These cycles reveal a truth about innovation: breakthroughs do not follow a linear trajectory. They advance in waves, carried by enthusiasm, then slowed by disappointment, then relaunched by those who persisted in the shadows.

Continued invisibilization.

The previous period had erased the women of Bletchley Park, the ENIAC programmers, non-Western mathematical traditions. This period continued the erasure — in other forms. The Argentine pioneers of ComIC — Clarisa Cortes, Cristina Zoltan, Liana Lew, Noemi Garcia — remained unknown outside their country. Rose Dieng-Kuntz, the first African woman admitted to Polytechnique and a pioneer of the semantic web, appears in no mainstream history of AI. Fernando Flores and the Cybersyn team were forgotten for decades after the Chilean coup. Timnit Gebru revealed that facial recognition systems erred up to thirty-five percent for dark-skinned women — proof that the biases of those who build systems are inscribed in the systems themselves.

The convergence of parallel paths.

Paths traced separately for decades began to meet. Japan moved from copying American transistors in the 1950s to dominating the global semiconductor market in the 1980s. South Korea, which had started with low-end assembly, became the world's largest memory manufacturer with Samsung. Taiwan, which had no computer industry in 1980, manufactured the majority of the planet's advanced chips by 2010. India, which had built TIFRAC in isolation, was exporting billions of dollars in software services. China, which had been cut off from the world during the Cultural Revolution, saw the emergence of Lenovo, Huawei, and the BAT giants. The parallel paths were converging toward the same horizon — and that horizon was no longer Western.

Protection as strategy.

Several nations used protection of their domestic market as a lever for technological development. Japan blocked Texas Instruments patents for years to allow its companies to copy the technology. Brazil imposed import restrictions and manufactured sixty-seven percent of its computers locally in 1982. China created the "Golden Shield" — the Great Firewall — which blocked Google, Facebook, and Twitter, allowing Baidu, Tencent, and Alibaba to prosper sheltered from foreign competition. These strategies were controversial — and effective. They raise a question that remains open: does innovation arise from openness or protection? The answer, perhaps, is: both, depending on the moment and context.

What Distinguishes: Six Facets of the Same Acceleration

If the common threads unite, each continent lived this period in its own way. Six singularities, six irreducible contributions to the history of artificial intelligence.

Africa invented frugal innovation — and revealed algorithmic biases.

M-Pesa transformed global financial inclusion from Kenya. Ushahidi created citizen mapping from Nairobi. Ubuntu Linux carried the African philosophy of "I am because we are" into global computer code. These innovations shared a characteristic: they were born of constraint, not abundance. Africa had no fixed telephone network — it invented mobile payments. It had no free media during the 2007 electoral violence — it invented participatory mapping. Frugal innovation is not bargain-basement innovation. It is innovation that solves problems that well-funded laboratories had not thought to pose. But Africa also brought a critical contribution: Timnit Gebru and Joy Buolamwini revealed that facial recognition systems, supposedly objective, were reproducing and amplifying the prejudices of those who had designed them. AI is not neutral. It bears the mark of its creators.

The Americas forged AI — and its cycles of hope and disillusionment.

Dartmouth, expert systems, deep learning: the great waves of artificial intelligence were born in North America. But the Americas also showed that other forges existed. Mexico created the first computer science master's in Latin America. Argentina developed ComIC, its own programming language. Brazil protected its industry and manufactured its own computers. And Chile dreamed of Cybersyn — a computer system to manage the economy in real time, "a sort of socialist Internet" destroyed by the 1973 coup. The Americas teach us that the history of AI is not linear. It is made of cycles, of forges extinguished by force, of interrupted paths. The Cybersyn project could have shown that another computing was possible — decentralized, at the service of workers, conceived in the Global South. We will never know what it would have become.

Asia demonstrated that parallel paths eventually converge — and that dominance can change continents.

Japan moved from copying to innovation, then from innovation to failure with the Fifth Generation project. The lesson was hard: you cannot decree a technological revolution. But Japanese robotics conquered the world — seventy percent of global production in 1980, forty-five percent still today. India leaped over hardware to become a software giant. The Y2K bug — that global panic born from a programming decision of the 1960s — became its launchpad. Bangalore now concentrates thirty-eight percent of Indian IT exports. Taiwan invented the fabless model with TSMC — a company that manufactures without designing, allowing NVIDIA, AMD, and Apple to exist without owning factories. Morris Chang, "put out to pasture" at Texas Instruments at fifty-four, created the "silicon shield" that now protects his island. China protected its market and saw its own giants emerge — Lenovo, Huawei, Baidu, Alibaba, Tencent. And Fei-Fei Li, born in Beijing, created ImageNet at Stanford — the database that launched the deep learning revolution.

Europe scuttled its future — and rebuilt itself.

The 1973 Lighthill Report triggered the first AI winter. The United Kingdom, which had invented the stored-program computer with the Manchester Baby, the first commercial computer with Ferranti, the foundations of academic AI with Donald Michie — the United Kingdom cut its own funding on the basis of a fifty-page report. The French Plan Calcul tried to create a national computer industry; it failed against IBM's dominance. But Europe rebuilt itself. Alain Colmerauer invented Prolog in Marseille — the language that would inspire the Japanese Fifth Generation project. Tim Berners-Lee created the World Wide Web at CERN — the invention that transformed the Internet into a global phenomenon. Linus Torvalds wrote Linux in Finland — the system that runs most of the world's servers. Yann LeCun laid the foundations for convolutional neural networks — the technology behind image recognition. DeepMind, founded in London, would beat the world Go champion with AlphaGo. Europe teaches us resilience — the capacity to fall, to rise again, and to start over.

The Middle East made the desert gardens bloom — and showed that necessity is the mother of invention.

Israel built the WEIZAC in 1955 — one of the world's first computers, in a six-year-old country surrounded by enemies. Unit 8200, created for military intelligence, became, without intending to, the world's greatest startup school. Check Point invented the modern firewall. ICQ invented instant messaging. Waze and Mobileye revolutionized navigation and autonomous driving. The "Startup Nation" exported eleven billion dollars in cybersecurity in 2021. Lotfi Zadeh, born in Baku, trained in Tehran, invented fuzzy logic at Berkeley — a way of representing vague concepts that humans handle intuitively. Americans were skeptical; the Japanese seized upon it. The United Arab Emirates planted the seeds of a digital ambition — Dubai Internet City, Masdar Institute — that would bloom after 2010. The Middle East teaches us that adversity can be a motor of innovation — and that gardens can grow in the desert.

Oceania proved that isolation can be a strength — and that the antipodes can invent bridges to the entire world.

The CSIRAC was the fifth stored-program computer in the world — built in Sydney, "largely independently of European and American efforts." Trevor Pearcey predicted the Internet in 1948. Graeme Clark invented the cochlear implant — more than one million people hear today thanks to him. The CSIRO developed a wireless transmission technique that became an essential component of WiFi — and won four hundred fifty million dollars against fourteen tech giants who tried to invalidate its patent. Where 2 Technologies created Google Maps from a Sydney apartment. Atlassian became Australia's first tech unicorn with ten thousand dollars of credit card debt. Oceania teaches us that geographic isolation is not a condemnation. It can force originality, oblige one to invent everything oneself, create solutions that centers of power had not imagined.

What This Period Teaches Us for AI Ethics

This crossing reveals four lessons that the previous period had sketched but that this one formulates with new urgency.

Biases are inscribed in systems.

Timnit Gebru and Joy Buolamwini demonstrated that facial recognition systems erred up to thirty-five percent for dark-skinned women — versus less than one percent for white men. This was not a bug. It was the reflection of training data, development teams, implicit priorities of those who built the systems. Artificial intelligence is not neutral. It amplifies the prejudices of its creators. And these prejudices, once encoded in algorithms deployed at scale, become infrastructures — difficult to see, even more difficult to dismantle.

Innovation can emerge from the margins.

M-Pesa was born in Kenya, not Silicon Valley. Ushahidi was born during an African electoral crisis, not in a research laboratory. The cochlear implant was born in Australia, not the United States. These innovations shared a characteristic: they responded to problems that centers of technological power had not thought to pose. Financial inclusion for the unbanked. Citizen mapping for countries without free media. Hearing for the profoundly deaf. The ethics of AI should recognize that diversity of contexts produces diversity of solutions — and that the margins can see what the center cannot see.

Protection creates ecosystems — but at what cost?

Japan protected its semiconductor industry and dominated the global market in the 1980s. China protected its digital market and saw Baidu, Alibaba, Tencent emerge. These protections were effective — but they had a cost. China's "Golden Shield" is not only an industrial policy. It is also a censorship system. The boundary between economic protection and political control is porous. The ethics of AI must recognize this ambiguity: the same tools that enable the emergence of local ecosystems can also serve surveillance and repression.

Today's decisions shape the following decades.

The 1973 Lighthill Report cut British AI funding for years. The 1973 Chilean coup destroyed Cybersyn and imprisoned Fernando Flores. Morris Chang's decision to create TSMC in 1987 redrew the maps of technological power for generations. These decisions — made by individuals, for reasons that seemed good to them at the time — had consequences that far exceeded their intentions. Who governs AI today? Who decides what data is used, what models are developed, what applications are authorized? These questions are not technical — they are political. And the answers we give them will perhaps shape the world for decades.

The Legacy for Artificial Intelligence

The period 1945-2010 left us the infrastructure of artificial intelligence — and the questions that accompany it.

From Dartmouth, we inherited the dream of simulating human intelligence. From Minsky and McCarthy, the first attempts to realize it. From expert systems, the lesson that knowledge can be encoded in rules. From the failure of the Fifth Generation project, the warning that technological hubris leads to dead ends. From Hinton, LeCun, and Bengio, the renaissance of neural networks. From Fei-Fei Li and ImageNet, the fuel that allowed deep learning to take off. This chain is direct, documented, taught.

But we also inherited what was erased or interrupted. The Cybersyn project, destroyed by a coup. The ComIC pioneers, forgotten in their own country. African researchers forced into exile. Computing traditions of countries in the Global South, marginalized by lack of resources. Rose Dieng-Kuntz, whose work on the semantic web appears in no mainstream history of AI.

More profoundly still, we inherited a concentration that is strengthening. In 1945, the centers of computing were multiple — Manchester, Princeton, Sydney, Rehovot. In 2010, a few American and Chinese companies concentrate the bulk of capabilities. The massive data needed for deep learning is held by a few giants. The data centers capable of training large models can be counted on the fingers of one hand. This concentration is not an accident — it is the product of economies of scale, network effects, political choices. Contemporary artificial intelligence inherits this concentration. And it amplifies it.

Toward the Next Period: 2010 to Today

In 2006, Geoffrey Hinton and his students published an article on deep neural networks. It was the beginning of a renaissance.

In 2009, Fei-Fei Li published ImageNet — a database of twelve million images classified into twenty-two thousand categories. It was the fuel that deep learning needed.

In 2012, Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton won the ImageNet competition with AlexNet — a convolutional neural network that surpassed all other systems. It was proof that deep learning worked.

The summer that opened would be the longest and most transformative in AI history. Large language models. Generative systems. Autonomous agents. The promises and perils of an artificial intelligence that approaches — or seems to approach — human capabilities.

But this period will inherit everything that came before. The biases inscribed in data. The concentration of power in a few hands. The continued invisibilization of contributors from the margins. The unlearned lessons of previous cycles. The parallel paths that are finally converging — toward what, we do not yet know.

The artificial intelligence we build tomorrow will depend on the choices we make today. If we tell ourselves only one story — that of Dartmouth, of Silicon Valley, of well-funded research laboratories — we will build only one type of intelligence. If we learn to listen to other stories — those of M-Pesa and Ushahidi, of Cybersyn and TSMC, of Rose Dieng-Kuntz and Timnit Gebru — perhaps we can build something else.

The Choice That Belongs to Us

This period leaves us a fundamental question: who inherits the digital revolution — and who is excluded from it?

Africa invented mobile payments before the West — but the digital divide persists. Sixty-seven percent of Latin American households have Internet access, compared to ninety-one percent in OECD countries. Indigenous peoples, those very ones whose ancestors were sent to residential schools, are among the least connected. Cuba, under embargo, did not authorize mobile Internet until 2018.

Innovation can emerge from the margins — but the margins remain margins. African talent continues to go into exile. The CSIRO patents were contested by fourteen tech giants. Chinese companies are blocked in the United States, American companies in China. Fragmentation is settling where convergence was promised.

The leap across the abyss is possible — but it is not automatic. It requires institutions, funding, political choices. M-Pesa was not born by accident. Neither was TSMC. These successes are the product of decisions — made by men and women, for reasons that seemed good to them at the time. Other decisions would have produced other results.

The Information Age leaves us this lesson: the future is not written. It is the product of the choices we make — and the choices we refuse to make.

The Information Age transformed conceptual tools into real machines, philosophical speculations into global industries, the dreams of a few researchers into planetary infrastructure. It leaves us a powerful technology and open questions: who controls it, who benefits from it, who is excluded from it.

The summer of deep learning has begun. What we build during this summer is up to us.

The journey continues.