![\"\"](\"https:\/\/world.edu\/wp-content\/uploads\/2026\/02\/innovation.jpg\")
<\/p>\n<\/p>\n
In graduate school, my experimental archaeology professor told a student to create a door socket \u2013 the hole in a door frame that a bolt slides into \u2013 in a slab of sandstone by pecking at it with a rounded stone. After a couple of weeks, the student presented his results to the class. \u201cI pecked the sandstone about 10,000 times,\u201d he said, \u201cand then it broke.\u201d<\/p>\n
This kind of experience is known as individual learning. It works through trial and error, with lots of each. Also known as reinforcement learning, it is how children,\u00a0chimpanzees<\/a>,\u00a0crows<\/a>\u00a0and\u00a0AI<\/a>\u00a0often learn to do something on their own, such as making a simple tool or solving a puzzle.<\/p>\n But individual learning has limits. No matter how much someone experiments through trial and error, improvement eventually hits a ceiling. Humans\u00a0have been throwing<\/a>\u00a0javelins\u00a0for a few hundred thousand years<\/a>, yet performance has largely plateaued. At the 2024 Olympics in Paris, the gold medal javelin throw was about 5% shy of Jan \u017delezn\u00fd\u2019s 1996 record. The level of expert play in the strategy game Go was\u00a0essentially flat from 1950 to 2016<\/a>, when artificial intelligence changed the equation.<\/p>\n Throughout humanity\u2019s existence, these limits on individual learning have not applied to technology. Since IBM\u2019s Deep Blue defeated world chess champion Garry Kasparov in 1997, supercomputers have\u00a0become a million times faster<\/a>\u00a0\u2013 and now\u00a0routinely outperform humans<\/a>\u00a0in chess and many other domains.<\/p>\n Why is technological improvement so different? My work as\u00a0an anthropologist<\/a>\u00a0on cultural evolution and innovation shows that, unlike individual performance, technology advances through combination and collaboration. As more people and ideas connect, the number of possible combinations\u00a0grows superlinearly<\/a>. Technological innovation\u00a0scales with the number of collaborators<\/a>.<\/p>\n My new book with anthropologist\u00a0Michael J. O\u2019Brien<\/a>, \u201cCollaborators Through Time<\/a>,\u201d reveals these patterns across human existence. It traces how 2 million years of technological traditions progressed through collaboration among specialists, across generations and with other species.<\/p>\n Expertise\u00a0has been the key<\/a>. Because traditional communities\u00a0know who their experts are<\/a>, specialization and collaboration have consistently underpinned\u00a0human success as a species<\/a>.<\/p>\n I\u2019d summarize our insight into how technology keeps advancing as TECH: tradition, expertise, collaboration and humanity.<\/p>\n The longest technological tradition documented by paleoanthropologists was the\u00a0Acheulean hand axe<\/a>. The multipurpose stone tool was made by our hominin ancestors for\u00a0almost a million years<\/a>, including some\u00a0700,000 years at a single site in eastern Africa<\/a>. People produced Acheulean tools through techniques they learned, practiced and refined\u00a0across generations<\/a>.<\/p>\n Later, small prehistoric societies of modern humans thrived on millennia of specialized knowledge, such as\u00a0music<\/a>, thatched roofs, seed cultivation,\u00a0burying dead bodies in bogs<\/a>, and making\u00a0millet noodles<\/a>\u00a0and even\u00a0cheese suitable for interring with mummies<\/a>.<\/p>\n As early as 22,000 years ago, communities near the Sea of Galilee stored and used more than a\u00a0hundred plant species, including medicinal plants<\/a>.\u00a0Shamans<\/a>\u00a0\u2013 ritual experts in medicinal knowledge and caregiving \u2013 helped their groups survive. Archaeological evidence from burial sites suggests these specialists were widely revered across thousands of years: One shaman woman\u00a0was interred<\/a>\u00a0with tortoise shells, the wing of a golden eagle and a severed human foot in a cave in Israel.<\/p>\n Traditional expertise alone does not advance technology. Technological progress occurs when different forms of expertise are combined.<\/p>\n The wheel may have\u00a0emerged from copper-mining communities<\/a>. One expert sourced copper from the Balkans, another\u00a0transported it<\/a>, another smelted it. By about 4000 B.C., additional specialists\u00a0cast copper into an early wheel-shaped amulet<\/a>: shaping a wax model, encasing it in clay, firing it in a kiln, pouring molten metal into the mold, then breaking the mold away.<\/p>\n Transport technologies reshaped ancient product networks. As communities across Eurasia and Africa built\u00a0wheeled vehicles<\/a>\u00a0and ships, and raised\u00a0domesticated horses<\/a>\u00a0and other\u00a0pack animals<\/a>, collaboration expanded across continents. Maritime and overland trade linked blacksmiths, scribes, religious scholars, bead makers, silk weavers and\u00a0tattoo artists<\/a>.<\/p>\n Expertise was often distributed between cities and their hinterlands, with cities functioning as hubs in cross-continental product networks. In ancient Egypt, no single community could produce a mummy. Mummification experts at Saqqara drew on a\u00a0continental network<\/a>\u00a0that supplied oils, tars and resins, combining these materials with specialized techniques of antisepsis, embalming, wrapping and coffin sealing.<\/p>\n Around the world, states and empires \u2013 from the Indus Valley Civilization to the Vikings, Mongols, Mississippians and Incas \u2013 expanded these networks, serving as hubs that coordinated the exchange of raw materials, specialized knowledge and finished products. These exchanges could be highly specific: Chinese porcelain was\u00a0shipped exclusively to 12th-century palaces in Islamic Spain<\/a>\u00a0via Middle Eastern traders who added Arabic inscriptions in gold leaf.<\/p>\n The scale has changed, but the structure has not. Today, within a\u00a0global product space<\/a>, an iPhone is assembled from a distributed network of specialized expertise and facilities.<\/p>\n Today, AI may disrupt the millennia-long pattern of technological advancement through TECH. Most large language models generate statistically common responses, which can flatten culture and\u00a0dilute expertise<\/a>\u00a0and originality<\/a>. The risk grows as untapped high-quality training data \u2013 our reservoir of expertise \u2013\u00a0becomes scarcer<\/a>.<\/p>\n This creates a feedback loop: Models trained heavily on low-quality content may degrade over time, with\u00a0measurable declines in reasoning and comprehension<\/a>. Some scientists now warn that humans and large language models could become locked in a mutually reinforcing cycle of recycled, generic content, with\u00a0brain rot<\/a>\u00a0for everyone involved. The dystopian extreme is\u00a0AI model collapse<\/a>, in which systems trained heavily on their own output begin to\u00a0produce nonsense<\/a>.<\/p>\n Brain rot is one reason some AI pioneers now question whether large language models will\u00a0achieve human-level intelligence<\/a>. But that, I think, is the wrong focus. The key to continually improving AI models is the same one that has sustained human expertise for millennia: keeping human experts in the loop \u2013 the E in TECH. Thanks to a kind of \u201cpied piper\u201d effect, an\u00a0informed minority can guide<\/a>\u00a0an uninformed majority who copy their neighbors.<\/p>\n In a\u00a0classic experiment<\/a>, guppies, following their neighbors, ended up schooling behind a robotic fish that guided them toward food. A recent study showed that traffic congestion eases when autonomous vehicles make up\u00a0as little as 5% of cars on the road<\/a>. In both cases, a small, informed minority reshaped the behavior of the whole system.<\/p>\n Like humans, large language models are social learners, and the learning can go in either direction. Designers can increase the likelihood that models continue to improve by ensuring they incorporate the accumulated lessons of human expertise across history. In turn, this creates the conditions for people and models to learn from one another.<\/p>\n In the 2010s, DeepMind\u2019s AlphaGo rediscovered\u00a0centuries of accumulated human Go knowledge through individual learning<\/a>, then went beyond it by\u00a0crafting strategies no human had ever played<\/a>. Human Go masters subsequently adopted these\u00a0AI-generated strategies into their own play<\/a>.<\/p>\n Well-trained large language models can likewise\u00a0summarize vast bodies of scientific information<\/a>, help\u00a0talk people out of conspiracy thinking<\/a>\u00a0and even support collaboration itself by\u00a0helping diverse groups find consensus<\/a>. In these cases, the learning flows both ways.<\/p>\n From Acheulean hand axes to supercomputers, human innovation has always depended on tradition, expertise, collaboration and humanity. If AI is tuned to find and trust expertise rather than dilute it, it can become humanity\u2019s next great technology \u2013\u00a0on par with ancient writing, markets and early governments<\/a>\u00a0\u2013 in our long story as collaborators through time.<\/p>\nTraditions and expertise \u2013 the critical foundation<\/strong><\/h4>\n
Collaboration \u2013 knowledge spanning time and place<\/strong><\/h4>\n
Humanity \u2013 social learning<\/strong><\/h4>\n
![\"\"](\"https:\/\/world.edu\/wp-content\/uploads\/2026\/02\/AItrainingdata.jpg\")
<\/p>\n