-
c. 3.3 million years ago · Lake Turkana region, Kenya
Stone flaking at Lomekwi 3
Excavators at Lomekwi 3, near Lake Turkana, reported large knapped stones, cores and flakes that they matched to volcanic ash layers dated to about 3.3 million years ago. That is roughly 700,000 years before the oldest tools previously accepted from the Oldowan tradition. No hominin bones were found beside them, so we do not know who made them or how they were used.
Debated. The age is inferred by correlating layers with dated ash, not by dating the stones. The maker is unknown, and some scholars question whether every piece was deliberately flaked rather than altered naturally. Lomekwi 3 holds the oldest knapped stones found so far, not necessarily the first tool use. Pounding, nut-cracking and tools made of wood or bone would rarely survive.
-
c. 1 million to 790,000 years ago (earliest claims) · Wonderwerk Cave, South Africa; Gesher Benot Ya'aqov, Israel
Early signs of controlled fire
Two sites anchor the early record. At Wonderwerk Cave in South Africa, microscopic study of cave sediments has been argued to show burning inside the cave about a million years ago. At Gesher Benot Ya'aqov in Israel, clusters of burned seeds, wood and flint from roughly 790,000 years ago suggest repeated burning in particular spots, possibly hearths.
Debated. Telling human-tended fire from lightning or grass fires is difficult in open sites and caves. Some archaeologists accept the million-year date, while others want more sites showing repeated, consistent hearth use. These are the oldest widely discussed signs that survive, not the moment fire was first used. Ash and charcoal decay quickly, so earlier use could have left nothing.
-
c. 315,000 years ago (± 34,000) · Jebel Irhoud, Morocco
Early Homo sapiens fossils at Jebel Irhoud
Fossils from Jebel Irhoud in Morocco combine a face, jaw and teeth that look strikingly modern with a braincase shaped differently from ours. Thermoluminescence dating of burned flint puts them near 315,000 years ago. The authors argued that features of our species emerged across the whole African continent over a long period, not in one tidy place at one moment.
Debated. Whether these people count as 'early Homo sapiens' or a closely related population depends on how the species is defined. The dating carries a margin of about 34,000 years. The fossils are the oldest of their kind discovered so far, not the first members of the lineage; the long-lived population that produced them must be older.
-
c. 100,000 to 75,000 years ago · Blombos Cave, South Africa
Engraved ochre at Blombos Cave
Blombos Cave on South Africa's southern coast has yielded pieces of red ochre with deliberate cross-hatched engravings, along with shell beads and worked bone. Levels containing the engraved pieces date from roughly 100,000 to 75,000 years ago. Many archaeologists read the designs as symbolic, while others suggest some marks might come from scraping ochre for powder.
Debated. Dates come from the sediment layers around the objects. Interpretation is contested: the engravings may be decorative, communicative, or a by-product of ochre processing. Blombos is among the earliest well-dated engraved objects, but older pigment use and possible symbolic items have been proposed elsewhere, and perishable decoration, such as on skin or wood, would not survive.
-
c. 65,000 years ago (disputed) · Madjedbebe rock shelter, Northern Territory, Australia
People in northern Australia at Madjedbebe
Excavations at the Madjedbebe rock shelter, on the lands of the Mirarr people, reported human occupation from about 65,000 years ago. Finds included grinding stones, ground ochre and ground-edge hatchet heads. Reaching Sahul, the landmass joining Australia and New Guinea, required crossing open water, so the site also bears on the early history of seafaring and migration out of Africa.
Debated. The age relies on optical dating of sand grains. Some researchers question whether artefacts moved between layers or whether the oldest dates are reliable; a younger arrival of about 50,000 years is still discussed. Madjedbebe gives a minimum age for the site, not for the first crossing. Rising seas have drowned the ancient coastlines where earlier landfall would have happened.
-
at least 51,200 years ago · Leang Karampuang, Sulawesi, Indonesia
Narrative painting in Sulawesi caves
In a limestone cave on Sulawesi, a painted scene shows human-like figures interacting with a pig. Researchers used laser-ablation uranium-series dating on thin mineral crusts that grew over the paint, giving a minimum age of at least 51,200 years. The scene has been described as one of the oldest known examples of figurative, story-like imagery.
Approximate. Uranium-series dates give minimum ages because they date mineral crusts covering the art, not the pigment. Whether the scene truly depicts a story is an interpretation, and other claims about older art exist. This is the oldest dated narrative-style scene known so far, not the first art; older possible markings and hand stencils have been proposed elsewhere.
-
c. 32,000 to 30,000 years ago · Chauvet-Pont d'Arc Cave, Ardèche, France
Painted animals in Chauvet Cave
Chauvet Cave holds charcoal drawings and paintings of lions, rhinoceroses, horses and bears, made with striking skill in shading and perspective. Radiocarbon dating places the images in the Aurignacian period, about 32,000 to 30,000 years ago. A rockfall sealed the entrance roughly 20,000 years ago, which helped preserve the cave until its rediscovery in 1994.
Approximate. Radiocarbon dates on charcoal and torch marks are well supported, but a few researchers have argued for younger dates for some images, so the range is not fully settled. Chauvet is among the best-preserved early galleries, which is different from being the earliest art. Sealed caves preserve images that open-air sites lose.
-
c. 23,000 to 21,000 years ago (debated) · White Sands National Park, New Mexico, USA
Ancient footprints at White Sands
Human footprints preserved in the lake-edge mud at White Sands in New Mexico have been dated to roughly 21,000 to 23,000 years ago, first using seeds and pollen in the sediment and later by radiocarbon dating of mud. If correct, people were in North America during the last glacial period, well before the traditional Clovis date of about 13,000 years ago.
Debated. Critics questioned whether seeds and pollen were reliable markers, and ask why no stone tools or camps were found nearby. Newer mud dates are consistent with the older estimate but the debate continues. Other sites are also proposed as older than Clovis, and coastal or boat routes may have left no surviving evidence because sea levels rose.
-
c. 9600 to 8200 BCE · Şanlıurfa region, southeastern Türkiye
Monumental pillars at Göbekli Tepe
On a hilltop in southeastern Türkiye, people carved and raised T-shaped limestone pillars, some over five metres tall, inside circular and rectangular enclosures. Many pillars bear animals in relief, and some show human-like arms and belts. The builders were hunter-gatherers living before farming became established in the region, and recent digs have also found domestic buildings.
Approximate. The use of the site is interpreted rather than known; ritual, feasting and gathering are common readings. Excavation covers only a small part, and the extent of permanent settlement is debated. It is among the earliest known monumental architecture, but older gathering places may exist that have not been found or did not use stone.
-
c. 11,000 to 10,000 years ago (process over millennia) · Fertile Crescent: Türkiye, the Levant, Iraq and Iran
Farming in Southwest Asia
Across Southwest Asia, people gathered, tended and then bred wild cereals and pulses, among them einkorn, emmer, barley, lentil and flax. Clearly domesticated forms of einkorn and emmer appear at sites in Türkiye by about 10,600 to 9,900 years ago, though wild plants may have been cultivated earlier. Whether domestication happened once or several times is still unresolved.
Approximate. Dates depend on how archaeologists identify domestic traits in charred seeds, and the shift took centuries to millennia. Evidence for peas and chickpeas is less clear than for cereals. The earliest clear domesticated remains are later than the earliest signs of plant cultivation; preserved plant remains are only a partial record.
-
c. 9,400 years ago and later · Shangshan culture sites, lower Yangtze valley, China
Rice cultivation on the lower Yangtze
At villages of the Shangshan culture, researchers examined microscopic silica fragments from rice and found the share showing cultivated-type features rose steadily across layers dated as far back as about 9,400 years ago. Rice husks were also used to temper pottery. The pattern suggests a gradual shift from gathering wild rice toward cultivation over many centuries.
Debated. Scholars debate how domesticated the earliest rice was, whether it was grown in paddies, and whether rice was domesticated in one region or several. Genetic studies point to complex origins for different rice varieties. The Shangshan phytoliths are the earliest evidence found so far; wild rice use and early cultivation almost certainly started before this record.
-
c. 8,700 years ago (earliest evidence found) · Central Balsas River valley, Mexico
Maize from teosinte in southwest Mexico
Maize descends from a wild grass called teosinte. Researchers working in Mexico's Central Balsas River valley found maize starch on grinding stones and maize phytoliths in layers dated to about 8,700 years ago. Maize later spread through the Americas, reaching Panama and northern South America over the following millennia as farmers kept adapting it to new climates.
Approximate. The excavators suspect the true date is earlier than the oldest dated layer. This is a 2009 report, and later studies may have refined the picture of where, when and how many times domestication occurred. The 8,700-year date marks the oldest layer where maize traces were found and dated, not the moment of first domestication.
-
c. 7,000 to 4,400 years ago · Kuk Early Agricultural Site, Western Highlands, Papua New Guinea
Wetland gardens at Kuk Swamp
In the highlands of New Guinea, Kuk Swamp preserves evidence of people working the wetlands for around 7,000 years, possibly longer. Mounds on the wetland margins, and from about 4,000 years ago drainage ditches dug with wooden tools, reflect changing farming methods. Crops such as banana, taro and yam were grown by planting cuttings and tubers rather than seeds.
Approximate. Dating the beginning of cultivation versus plant management is hard, and UNESCO gives a range of dates. Identifying which plants were truly domesticated, rather than managed, is an ongoing research question. Kuk preserves a long sequence of changing practices; wooden tools and tubers decay quickly, so earlier gardening elsewhere on the island may be lost.
-
c. 3300 to 3000 BCE · Uruk, southern Mesopotamia (modern Iraq)
Cuneiform record-keeping in Uruk
In the city of Uruk, scribes pressed marks into clay to record goods such as barley and livestock. These early tablets are mostly accounts, not stories or letters. Over centuries the pictures became wedge-shaped signs, and the script was adapted to write Sumerian, Akkadian and other languages across the Middle East for over three thousand years.
Approximate. Dates for the earliest tablets are estimates based on archaeological layers; one barley-distribution tablet from Uruk is dated by Britannica to about 3100 to 2900 BCE. Writing may have emerged in several places, and its exact origin is discussed. The surviving clay tablets are the earliest known administrative writing in Mesopotamia, but earlier counting tokens and perishable records may have existed.
-
c. 2500 BCE onward (UNESCO: flourished c. 2500 to 1500 BCE) · Mohenjo-daro, Sindh, Pakistan, and the wider Indus Valley
Planned cities of the Indus civilization
Mohenjo-daro lay on the Indus plain, with streets laid out on a grid, brick buildings, wells, public baths and covered drains. UNESCO describes it as among the best-preserved early urban settlements in South Asia. The wider Indus civilization used stamp seals bearing signs that remain undeciphered, so much about its government, beliefs and daily life is inferred from objects.
Approximate. Chronologies for the Indus cities vary between scholars, and UNESCO's 2500 to 1500 BCE span differs from some narrower urban-phase estimates. The script has not been deciphered, and claims of decipherment are not accepted. Mohenjo-daro is a well-preserved example, not the earliest Indus city; earlier towns and farming villages precede it in the region.
-
c. 1250 to 1050 BCE · Anyang, Henan, China
Oracle bone inscriptions at Anyang
Shang-dynasty diviners at Anyang wrote questions on cattle shoulder blades and turtle shells, then applied heat until the bone cracked and read the cracks as answers. They asked about harvests, weather, battles and illness. The inscriptions are the earliest surviving examples of Chinese writing and already form a developed script, though they were only recognised as writing around 1899.
Approximate. Dates follow the occupation of Anyang, usually given as roughly 1250 to 1050 BCE. Interpretations of individual inscriptions continue to be refined. These are the earliest surviving Chinese writing, not necessarily the first. Everyday writing was probably on bamboo, which does not survive in the soil.
-
c. 1000 to 800 BCE (c. 900 BCE best estimate) · San Lorenzo region, Veracruz, Mexico
The Cascajal Block and early Mesoamerican writing
A serpentine slab found in a gravel quarry near the Olmec site of San Lorenzo carries 62 incised signs, made up of 28 distinct ones. Its excavators dated it to about 900 BCE from the pottery and other objects found with it. It is the oldest known writing in Mesoamerica, though the signs remain undeciphered and have no clearly known descendants.
Debated. The slab was found without a clear archaeological layer, so its date rests on associated objects. Scholars debate how closely it is linked to later scripts, and decipherment would need more examples. This is the earliest writing-like object found so far, not proof of the earliest writing; the block's context was already disturbed when it was found.
-
c. 8th to 9th centuries CE · Baghdad, Abbasid Caliphate (modern Iraq)
Translation and scholarship in Abbasid Baghdad
Under the early Abbasid caliphs, Baghdad became a centre for translating and writing on mathematics, astronomy, medicine and philosophy, drawing on Persian, Greek, Syriac and Sanskrit works. The 'House of Wisdom' is often named as the heart of this effort, and scholars such as al-Khwarizmi are associated with this scholarly world. Historians disagree about what the institution actually was.
Debated. Some historians describe the House of Wisdom as a library and translation bureau, while others argue that popular accounts overstate it and that the translation movement was driven more widely. Founding dates and the role of individual caliphs are uncertain. Our knowledge relies mostly on later writers, so the surviving accounts are not first-hand records of the institution.
-
11 May 868 CE · Tang China; found at Dunhuang
The Diamond Sutra printed in 868
A scroll made of seven printed sheets carries a Buddhist text, a woodcut of the Buddha, and a colophon dated 11 May 868, stating it was made for free distribution. Found in 1900 in a sealed cave library near Dunhuang, it is now in the British Library. It is the earliest dated printed book yet discovered.
Well-established. The date is written on the scroll itself. The 'earliest dated' label is limited to what has been found so far, and the scale of printing at the time is unknown. Printing began earlier, probably in the eighth century; this scroll is the earliest surviving dated example, not the first printed work.
-
c. 1011 to 1021 CE · Cairo, Fatimid Caliphate (Egypt)
Ibn al-Haytham’s Book of Optics
Ibn al-Haytham, born in Basra and working in Cairo, wrote the Book of Optics in seven volumes, arguing that vision happens when light from objects enters the eye. He combined geometry with experiments, such as tests with dark chambers and mirrors. A Latin translation circulated in Europe from the late twelfth or thirteenth century, and a printed edition appeared in 1572.
Approximate. Dates for the work come from biographical sources with conflicting stories about his life. Labels such as 'father of modern optics' or 'first scientist' are modern characterisations; historians stress that experimental practice also had earlier roots. The Book of Optics is among the earliest surviving works to combine mathematical argument with systematic experiments on light; it is not the first investigation of optics.
-
c. 1025 to 1290 CE (dates debated) · East Polynesia: Society Islands, Hawai'i, Aotearoa New Zealand and more
Settling the far Pacific by voyaging canoe
Polynesian navigators crossed thousands of kilometres of open ocean using knowledge of stars, swells, winds and birds. A review of over 1,400 radiocarbon dates from 47 islands concluded that settlement of the central Society Islands came around 1025 to 1120 CE, with a rapid spread to the remaining islands, including Hawai'i and Aotearoa New Zealand, between about 1200 and 1290 CE.
Debated. Earlier estimates placed settlement much earlier. The newer dates rely on short-lived plant samples and are widely cited, but archaeologists still debate individual islands and dating practices. Radiocarbon dates show when materials were burned or left behind, not the first landfall; the exact first arrivals are likely slightly earlier than the earliest dated deposits.
-
July 1377 CE · Heungdeok Temple, Cheongju, Korea (Goryeo)
Jikji printed with metal type in Korea
In July 1377 at Heungdeok Temple in Cheongju, a collection of Buddhist teachings known as the Jikji was printed with movable metal type. The surviving second volume is now held in the Bibliothèque nationale de France. Korean nomination papers describe it as the oldest existing book printed with movable metal type.
Well-established. The date is printed in the book. Claims of 'oldest' apply to surviving books, and the nomination text is from the Korean side; earlier Korean printing with metal type is recorded in documents but not preserved. Written records suggest metal-type printing in Korea started earlier than 1377, but the Jikji is the oldest surviving example discussed in this source.
-
c. 1450 to 1455 CE · Mainz, Holy Roman Empire (modern Germany)
Gutenberg’s press in Mainz
Johannes Gutenberg combined a durable type metal, an oil-based ink and a press that applied even pressure, using these to print in Mainz around 1450. His Bible followed around 1455. Printing from movable type had already existed in China and Korea, and the Mainz achievement was a particular integrated system, which then spread quickly across Europe.
Approximate. Exact dates for the first prints are uncertain, and Gutenberg's business history is documented mainly through lawsuits and later accounts. This is not the first movable type; it is the first widely documented European press system. Earlier Chinese and Korean printing is a separate tradition.
-
c. 1760s to 1840s CE (Britannica's wider span: 1733 to 1913) · Britain, especially northern England and the Midlands
Mechanised industry in Britain
Britain's shift from handicraft and farming toward machine manufacturing brought spinning machines, steam engines and factories, with the spinning jenny and James Watt's steam engine among its famous symbols. The change brought new goods and growth, but also crowded towns and long factory hours. An 1833 Act banned factory work for children under nine and limited hours for older children.
Debated. Historians disagree about start and end dates, how revolutionary the change really was, and which causes mattered most. Dates here are approximate, and the human costs, including colonial and enslaved labour that supplied raw cotton, are part of the wider record. Mechanised textile production and steam power were developed through many inventors and earlier workshops; no single machine marks the first factory.
-
1854 CE · Soho, London, England
Mapping cholera in Soho
During London's 1854 cholera outbreak, physician John Snow mapped deaths in Soho and argued that cholera spread through contaminated water rather than bad air. He also compared customers of different water companies, using the shift of one company's intake upriver as a natural experiment. His 1855 treatise set out the case. Many contemporaries, including William Farr, favoured the bad-air explanation.
Well-established. The popular story that removing a pump handle ended the outbreak is more complicated; deaths were already declining. Snow's maps have been altered in some retellings, and others, including a local clergyman, contributed data. Filippo Pacini described the cholera organism in 1854, and Robert Koch's identification in 1884 is better known; Snow made his case without seeing the bacterium.
-
1872 CE onward · Tomioka, Gunma Prefecture, Japan
Industrialising silk at Tomioka
The Tomioka Silk Mill, founded in 1872 early in the Meiji period, was set up by the Japanese government with French machinery and technical expertise. It helped modernise raw silk production and made Japan a leading exporter of raw silk, mainly to Europe and the United States. It is a record of industrialisation shaped by state planning and technology transfer, not just a copy of Britain.
Well-established. UNESCO's English text gives a broader founding window, while its other language versions give 1872. The page does not describe working conditions, which historians have studied separately.
-
17 December 1903 · Kitty Hawk, North Carolina, USA
The Wright brothers’ powered flights
On 17 December 1903, Orville and Wilbur Wright made four powered flights near Kitty Hawk, the first lasting about 12 seconds and covering 120 feet, the last going 852 feet in 59 seconds. Their success rested on years of glider tests and research, and on an engine built in about six weeks by their machinist Charlie Taylor, who stayed in Dayton.
Well-established. Claims about who flew 'first' vary by definition (powered, controlled, sustained, public). Other inventors, including glider pioneers, contributed essential knowledge, and Taylor's role is often overlooked. Earlier powered hops and glider flights were attempted by others, so 'first' depends on the exact definition used.
-
1936 CE · Cambridge and Princeton (UK and USA)
Turing’s universal machine on paper
In a paper published in the Proceedings of the London Mathematical Society in 1936, Alan Turing described an abstract machine that follows rules to calculate, and showed that no general method can decide every mathematical statement. Alonzo Church reached a similar result independently by different means a few months earlier. The paper became a foundation for the theory of computation.
Well-established. The paper appeared in parts dated 30 November and 23 December 1936, with a correction later. Church's independent work and others' earlier ideas, such as Babbage's engines, are part of the wider story.
-
14 February 1946 · Moore School of Electrical Engineering, University of Pennsylvania, USA
ENIAC and the people who programmed it
ENIAC, announced in February 1946, was a room-sized electronic computer with about 18,000 vacuum tubes, designed by John Mauchly and J. Presper Eckert to calculate artillery trajectories. A set of calculations that took about 12 hours by hand took around 30 seconds. Six women, including Jean Jennings Bartik and Betty Holberton, programmed it by setting switches and cables, and their work was long under-credited.
Well-established. Debates over 'first computer' depend on definitions: electronic, programmable, stored-program. Other machines, such as Colossus and the Z3, also have strong claims. Other early electronic and electromechanical computers, including machines in Germany and Britain, came before or around the same time.
-
1952 CE · Eckert-Mauchly Computer Corporation, Philadelphia, USA
Hopper’s compiler and readable programming
In 1952, Grace Hopper created the A-0 compiler, a program that translated instructions into machine code and let a computer call stored subroutines automatically. She later led work on FLOW-MATIC, a business-oriented language that influenced COBOL. Her approach pushed programming toward language closer to ordinary words, and she helped standardise COBOL for government and industry.
Well-established. 'First compiler' is debated depending on definition, and earlier automatic coding efforts existed elsewhere. COBOL was designed by a committee, with Hopper as one influential contributor. Other early automatic programming systems were developed around the same time, so Hopper's A-0 is best described as an early compiler rather than the only first.
-
Summer 1956 (proposal August 1955) · Dartmouth College, New Hampshire, USA
The Dartmouth workshop names AI
In 1955 John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon proposed a summer study at Dartmouth based on the conjecture that learning and other features of intelligence could be described precisely enough for a machine to simulate them. The workshop in summer 1956 was loose and fluid, with changing attendance. It is widely treated as the moment the term 'artificial intelligence' entered the field.
Approximate. Sources give different proposal dates and varying attendee lists; the workshop's exact schedule is reconstructed from participants' notes. Earlier meetings on similar questions had already taken place. The proposal introduced the term 'artificial intelligence', but earlier work on thinking machines, including Turing's, predates the workshop.
-
4 October 1957 · Baikonur, Soviet Union (modern Kazakhstan)
Sputnik 1 enters orbit
On 4 October 1957 the Soviet Union launched Sputnik 1, a polished sphere about 58 centimetres across and weighing roughly 84 kilograms, which circled Earth about every 98 minutes. NASA describes it as the first artificial satellite and the start of the space age. The launch surprised the American public and led to Explorer 1 and to the creation of NASA in 1958.
Well-established. The date is well documented. Soviet engineers, including Sergei Korolev's team, built the satellite, but individual credit was often hidden at the time.
-
12 April 1961 · Baikonur launch; landing in the Soviet Union
Gagarin’s orbit aboard Vostok 1
On 12 April 1961 Yuri Gagarin flew aboard Vostok 1, completing one orbit of Earth in about 108 minutes and becoming the first human in space. The flight depended on rocket designers, engineers, doctors and ground controllers in a programme run largely in secret. Two days later he was welcomed in Moscow, where he appeared on the Kremlin balcony with Premier Nikita Khrushchev.
Well-established. The date and orbit are well documented. Details of the flight and landing were controlled by the state at the time, and later accounts have clarified some of them.
-
Campaign 1967 to 1980 CE · Global; last natural case in Somalia, 1977
Smallpox declared eradicated
Smallpox, caused by the variola virus, killed people for thousands of years. The World Health Organization launched an intensified eradication campaign in 1967 using mass vaccination and surveillance, building on Edward Jenner's 1796 vaccine. The last known natural case occurred in Somalia in 1977, an independent commission certified eradication in December 1979, and WHO declared it eradicated in 1980.
Well-established. The 1980 declaration is well documented. Earlier practices such as variolation existed in several regions, and the programme's success also depended on surveillance and containment. Jenner's 1796 vaccine was the first successful vaccine in this account, but inoculation methods against smallpox were used earlier in parts of Asia, Africa and the Middle East.
-
1967 to 1980 CE (key results 1972) · China (Beijing, Yunnan, Shandong, Hainan and other sites)
Artemisinin and Project 523
Project 523, begun in May 1967, was a large Chinese programme to find new malaria treatments. In 1969 Tu Youyou was asked to lead work at her institute. Her team reviewed traditional texts, including a fourth-century handbook by Ge Hong, and found that low-temperature extraction of sweet wormwood preserved the active compound. Clinical trials followed in 1972, and many colleagues refined and tested the extract.
Well-established. The programme's scale, number of institutions and individual credit are described differently in different sources. Early results were published anonymously under institutions, and Western approval came decades later. Sweet wormwood was used in traditional medicine long before; the 1972 work isolated and tested the active compound.
-
16 to 24 July 1969 · Sea of Tranquility, the Moon; launched from Florida, USA
Apollo 11 lands on the Moon
Apollo 11 launched on 16 July 1969, carrying Neil Armstrong, Buzz Aldrin and Michael Collins. Armstrong and Aldrin spent about 22 hours on the lunar surface while Collins orbited above, and the crew splashed down on 24 July. The programme employed about 400,000 people at its peak, in more than 20,000 firms and universities, with flight controllers, CapComs and backup crew.
Well-established. Mission dates and crew are well documented. Figures for workforce size vary by source and by how 'worked on Apollo' is counted.
-
29 October 1969 · UCLA and Stanford Research Institute, California, USA
The first ARPANET message
On 29 October 1969, at about 10:30 p.m., UCLA student programmer Charley Kline tried to log in to a computer at the Stanford Research Institute, in the first host-to-host message on the ARPANET. He typed 'l' and 'o' and Bill Duvall at SRI confirmed each letter by phone; when 'g' was typed, the system crashed. The only record of the exchange is an entry in a log.
Well-established. The date is widely documented. Whether this event counts as 'the birth of the internet' depends on definition, since later protocols and other networks were also essential.
-
1989 to 1993 CE · CERN, near Geneva, Switzerland
The World Wide Web at CERN
In 1989 Tim Berners-Lee, a British scientist at CERN, proposed the World Wide Web to share information among scientists at institutions around the world. The first website ran on his NeXT computer, and CERN released a line-mode browser in 1992. On 30 April 1993 CERN placed the Web software in the public domain, later using an open licence to help it spread.
Well-established. Sources differ on some intermediate milestones in 1990 and 1991. Collaborators such as Robert Cailliau and others also contributed.
-
October 1990 to April 2003; gaps closed 2022 · Laboratories in several countries, led by an international consortium
The Human Genome Project
Launched in October 1990 and declared complete in April 2003, the Human Genome Project produced the first sequence of the human genome, with an international group of researchers. The reference covered about 92 percent of the genome. In 2022 the Telomere-to-Telomere consortium published the first complete, gapless sequence, adding roughly 8 percent, most of it highly repetitive DNA that earlier methods could not read.
Well-established. The 2003 sequence was not truly complete, and the 2022 sequence came from a special cell line with two identical chromosome copies, so it does not represent all human variation. The Y chromosome was not part of that cell line.
-
2012 CE · University of Toronto, Canada
AlexNet and the deep learning turn
At the 2012 NeurIPS conference, Alex Krizhevsky, Ilya Sutskever and Geoffrey Hinton presented a deep convolutional neural network trained on GPUs to classify images from the ImageNet dataset. The network had about 60 million parameters and 500,000 neurons, and reported error rates considerably better than earlier approaches on the 2010 benchmark. Its success drew wide attention to deep learning.
Well-established. The paper's figures refer to the 2010 benchmark, and the 2012 competition result is described elsewhere. Earlier neural-network work by many researchers laid its foundations. Convolutional networks existed earlier; AlexNet is notable for scale and results rather than being the first.
-
June 2017 CE · Google (research paper, preprint on arXiv)
The Transformer architecture
In June 2017 eight researchers posted the paper 'Attention Is All You Need', proposing the Transformer, a model for sequences that relies on attention mechanisms and dispenses with recurrence and convolutions. The design is easier to train in parallel on modern hardware. Variants of the architecture now underlie many large language models and other AI systems.
Well-established. The date is the arXiv submission; the paper was also presented at a conference. Earlier work on attention in neural networks contributed to the idea.
-
May 2020 CE · OpenAI (research paper, preprint on arXiv)
GPT-3 and few-shot language models
In May 2020 researchers posted a paper describing GPT-3, a language model with 175 billion parameters. Without further training, it could perform many tasks from a few examples in a text prompt, including translation, question answering and simple arithmetic. The authors also reported weaknesses, noting that few-shot learning still struggled on some datasets and that training on web data raised methodological concerns.
Well-established. This entry records what one paper reported at one date. Performance claims, comparisons with humans and the nature of 'understanding' remain actively contested; later systems have changed quickly.