What Had to Happen Before Artificial Intelligence?

The Road to Artificial Intelligence — Part 1 of 15

BEFORE THE MACHINE

Artificial intelligence only seems to have arrived suddenly: computers just learned to recognize images, generate prose, and solve problems once thought uniquely human.

Its deeper history, however, is much older—and less straightforward.

Long before electronic computers, humans developed ways to extend memory, coordinate action, preserve information, and repeat complex procedures. They shaped stone, controlled fire, built with wood, marked surfaces, counted goods, and recorded transactions.

These developments were never steps in a deliberate march toward AI. Their makers were solving immediate problems in particular places and circumstances. Viewed across time, however, they reveal a recurring transformation: knowledge could increasingly exist outside any single human mind.

This series follows that thread. No one in the ancient world was working toward artificial intelligence, yet human societies gradually built tools and systems that made information easier to preserve, share, and reuse.

Early tools and material technologies

The archaeological record preserves objects, features, residues, and sediments, but rarely any explanation from the people who left them.

At Lomekwi 3 in Kenya, researchers reported stone tools about 3.3 million years old and proposed naming the assemblage "Lomekwian" to set it apart from the later Oldowan tradition. The claim is important but debated, so Lomekwi cannot be called the undisputed start of technology.

Oldowan tools are better documented. At Nyayanga, Kenya, they date to between 3.03 and 2.58 million years ago and appear alongside butchered hippopotamid remains. Paranthropus teeth were found there too, but the toolmakers remain unidentified.

By about 1.75 million years ago, Acheulean toolmakers at Konso in Ethiopia were producing carefully shaped bifacial tools through repeated striking and reshaping. The artifacts are consistent with extended planning and multi-stage action, though they cannot reveal what their makers were thinking.

These tools extended their makers' ability to cut, break, scrape, and process. That is an interpretation of their function, not evidence that anyone conceived of "technology" as an abstract project.

Fire, food, and structured activity

Fire introduced another form of control.

At Gesher Benot Ya'aqov in Israel, burned materials and their spatial patterning have been interpreted as evidence that hominins controlled fire around 790,000 years ago. A related study argued that fish were cooked there about 780,000 years ago. Both conclusions rest on specific evidence and are narrower than the claim that humans "discovered cooking."

At Barnham in Britain, researchers reported heated sediment, fire-cracked flint, and locally rare pyrite in deposits around 400,000 years old. They interpreted this as deliberate fire-making. The hominins responsible are unknown.

Fire provided more than heat. It allowed materials to be transformed, food to be processed, and activity to be organized around a controlled source of energy. The evidence offers no direct record of language or social rules, but it does show increasingly structured relationships among hominins, materials, and environments.

Building beyond the moment

At Kalambo Falls in Zambia, researchers reported interlocking logs joined by an intentionally cut notch, dating to at least 476,000 years ago. The structure may have been a platform, walkway, or foundation. Its function is uncertain.

Its significance is not. Woodworking had moved beyond individual tools: large timbers were shaped and fitted into a durable arrangement.

At Schöningen in Germany, a wooden throwing stick about 300,000 years old is evidence of effective hunting technology and of carefully designed wooden weapons. The object alone does not demonstrate coordinated hunting.

Wood rarely survives. Its preservation at these sites corrects stone-centered histories, because many technologies disappeared when their materials decayed, not because they never existed.

Marks, objects, and shared meaning

Some surviving objects suggest that hominins made marks intentionally, or selected objects whose value was not simply mechanical.

Shell beads from Skhul in Israel and Oued Djebbana in Algeria date to roughly 135,000 to 100,000 years ago. Their selection and transport are consistent with intentional ornamentation. Whether they signaled identity, affiliation, status, or something else cannot be established.

At Blombos Cave in South Africa, a silcrete flake bearing a cross-hatched ochre design dates to about 73,000 years ago. The mark appears deliberate, but its meaning is unknown.

In Iberian caves, some red markings have minimum ages older than 64,000 years and may have been made by Neanderthals. The dating assumptions and the identification of the marks as deliberate art are both disputed.

Sulawesi provides further evidence of very early image-making. A composition at Leang Karampuang has a minimum age of 51,200 years, and its authors interpret it as narrative. That is an interpretation, not an uncontested description of the image's meaning.

A study published in 2026 reported a Sulawesi hand stencil with a minimum age of 67,800 years, based on dating the calcite that overlies the artwork. It is safer to call this among the oldest known dated rock art than definitively the oldest.

Together, these finds show that hominins produced durable visual marks and objects with possible social or symbolic significance. They cannot show how those marks were understood, and they do not establish a single origin for language, storytelling, or art.

From objects to records

The next transformation involved more than additional symbols. External objects began to organize practical information.

In the ancient Near East, clay objects often called "tokens" have been interpreted as early accounting devices. One influential theory held that they represented commodities and eventually contributed to the development of writing. Later scholarship emphasizes that they may have served multiple functions and should not automatically be treated as direct ancestors of writing.

The disagreement matters because artifacts do not explain themselves. A shaped clay object may have served accounting, measurement, identification, ritual, or several purposes at different times.

In southern Mesopotamia, the earliest widely accepted writing emerged around or before 3200 BCE. The surviving early texts are mostly administrative: quantities, goods, workers, and institutional transactions.

In Egypt, evidence from Tomb U-j is associated with writing around 3250 BCE. Early Egyptian writing was limited and administrative, not a complete system for recording continuous speech.

Writing changed the relationship between information and memory. A record could outlast its maker and be read by someone who had not witnessed the original transaction. This was a functional consequence of durable records, not necessarily their creators' intention.

Calculation as repeatable procedure

The Rhind Mathematical Papyrus, from around 1550 BCE, preserves about 84 problems involving fractions, multiplication, division, geometry, areas, and volumes.

Its procedures were not "algorithms" in the modern computer-science sense, but they were repeatable methods that could be applied to similar problems. Ancient mathematical texts contained no modern theory of computation. They did demonstrate a powerful idea: some reasoning can be expressed as an ordered procedure instead of being performed anew through intuition each time.

The connection to AI

Modern AI depends on several capabilities that developed across very different periods:

• tools that extend physical action
• symbols that preserve distinctions and meanings
• language and teaching that allow skills to move between people
• writing that stores information outside individual memory
• mathematics that expresses repeatable procedures
• institutions and networks that accumulate records and coordinate activity

None of these developments was an inevitable step toward artificial intelligence. Technologies disappeared, changed direction, and developed unevenly across societies, and the archaeological record is strongly biased toward materials that survive.

The connection is therefore analytical, not causal. AI belongs to a long human tradition of externalizing memory, calculation, and decision-making. Computers later automated operations over symbols and data, and machine learning added the ability to detect patterns in large collections of examples.

This history shows that AI is not simply a mysterious new kind of "mind." It extends older practices: representing information, preserving it, transforming it, and applying procedures to it.

The analogy has limits. Ancient artifacts do not show that their makers had modern concepts of information processing, and cumulative culture does not imply that increasingly capable AI is destined to become autonomous or socially transformative.

Looking ahead

Once quantities and procedures could be recorded outside the mind, a new question became possible: could a device perform some of those operations?

The next article follows the shift from counters and written methods to mechanical calculators, punched controls, and proposed programmable machines. That transition, from recording calculation to automating it, would provide part of the foundation for modern computation.

SOURCES

1. Sonia Harmand, Jason E. Lewis, Craig S. Feibel et al., "3.3-million-year-old stone tools from Lomekwi 3, West Turkana, Kenya," Nature 521, 310–315 (2015). DOI: 10.1038/nature14464.
2. Thomas W. Plummer, James S. Oliver, Emma M. Finestone et al., "Expanded geographic distribution and dietary strategies of the earliest Oldowan hominins and Paranthropus," Science 379, 561–566 (2023). DOI: 10.1126/science.abo7452.
3. Yonas Beyene, Shigehiro Katoh, Giday WoldeGabriel et al., "The characteristics and chronology of the earliest Acheulean at Konso, Ethiopia," PNAS 110, 1584–1591 (2013). DOI: 10.1073/pnas.1221285110.
4. Naama Goren-Inbar, Nira Alperson-Afil, Mordechai E. Kislev et al., "Evidence of hominin control of fire at Gesher Benot Ya'aqov, Israel," Science 304, 725–727 (2004). DOI: 10.1126/science.1095443.
5. Irit Zohar, Nira Alperson-Afil, Naama Goren-Inbar et al., "Evidence for the cooking of fish 780,000 years ago at Gesher Benot Ya'aqov, Israel," Nature Ecology & Evolution 6, 2016–2028 (2022). DOI: 10.1038/s41559-022-01910-z; author correction published November 20, 2023, DOI: 10.1038/s41559-023-02270-y.
6. Rob Davis, Claire Lucas, Simon O'Connor et al., "Earliest evidence of making fire," Nature 649, 631–637 (2025). DOI: 10.1038/s41586-025-09855-6.
7. L. Barham, G. A. T. Duller, I. Candy et al., "Evidence for the earliest structural use of wood at least 476,000 years ago," Nature 622, 107–111 (2023). DOI: 10.1038/s41586-023-06557-9.
8. Nicholas J. Conard, Jordi Serangeli, Gerlinde Bigga et al., "A 300,000-year-old throwing stick from Schöningen, northern Germany, documents the evolution of human hunting," Nature Ecology & Evolution 4, 690–693 (2020). DOI: 10.1038/s41559-020-1139-0.
9. Christopher S. Henshilwood, Francesco d'Errico, Karen L. van Niekerk et al., "An abstract drawing from the 73,000-year-old levels at Blombos Cave, South Africa," Nature 562, 115–118 (2018). DOI: 10.1038/s41586-018-0514-3.
10. Marian Vanhaeren, Francesco d'Errico, Chris Stringer et al., "Middle Paleolithic shell beads in Israel and Algeria," Science 312, 1785–1788 (2006). DOI: 10.1126/science.1128139.
11. Dirk L. Hoffmann, Christopher D. Standish, Alistair W. G. Pike et al., "U-Th dating of carbonate crusts reveals Neandertal origin of Iberian cave art," Science 359, 912–915 (2018). DOI: 10.1126/science.aap7778; João Zilhão and colleagues, critique in Journal of Human Evolution, DOI: 10.1016/j.jhevol.2018.08.004.
12. Adhi Agus Oktaviana, Renaud Joannes-Boyau, Budianto Hakim et al., "Narrative cave art in Indonesia by 51,200 years ago," Nature 631, 814–818 (2024). DOI: 10.1038/s41586-024-07541-7.
13. Adhi Agus Oktaviana, Renaud Joannes-Boyau, Budianto Hakim et al., "Rock art from at least 67,800 years ago in Sulawesi," Nature 650, 652–656 (2026), published online January 21, 2026; issue date February 19, 2026. DOI: 10.1038/s41586-025-09968-y. Supplementary Table 1 was corrected February 4, 2026; the journal states that the corrected values do not affect the study's calculations.
14. Denise Schmandt-Besserat, "The token system of the ancient Near East: Its role in counting, writing, the economy and cognition," in Iain Morley and Colin Renfrew, eds., The Archaeology of Measurement: Comprehending Heaven, Earth and Time in Ancient Societies, 27–34 (Cambridge University Press, 2010). DOI: 10.1017/CBO9780511760822.006.
15. Robert K. Englund, "Accounting in Proto-Cuneiform," in Karen Radner and Eleanor Robson, eds., The Oxford Handbook of Cuneiform Culture, 32–50 (Oxford University Press, 2011). DOI: 10.1093/oxfordhb/9780199557301.013.0002.
16. Lucy E. Bennison-Chapman, "Reconsidering 'Tokens': The Neolithic Origins of Accounting or Multifunctional, Utilitarian Tools?" Cambridge Archaeological Journal 29, 233–259 (2019). DOI: 10.1017/S0959774318000513.
17. Ilona Regulski, "The Origins and Early Development of Writing in Egypt," The Oxford Handbook of Topics in Archaeology (2016). DOI: 10.1093/oxfordhb/9780199935413.013.61.
18. Cristine H. Legare, "Cumulative cultural learning: Development and diversity," PNAS 114, 7877–7883 (2017). DOI: 10.1073/pnas.1620743114.

Continue the series →

Part 2 — From Counters to Programs