--- title: "Technology tree · how virtual worlds became possible · Simulation Hypothesis" url: https://simulation-hypothesis.com/trees/technology description: "The inventions behind virtual worlds." license: https://creativecommons.org/licenses/by/4.0/ credit: "Ross Allan Dahl, Simulation Hypothesis (simulation-hypothesis.com)" --- # Technology tree · how virtual worlds became possible A virtual world needs two things: a way to fool the senses, and a machine that computes what they should receive. For most of history these developed separately. They meet at the head-mounted display. You can open any node. Lines show what each one built on. Dashed nodes stalled because a prerequisite was missing. Future nodes are published forecasts, each cited. A forecast’s date range covers the sources quoted, and they may define the milestone differently. **0** of 82 read unread read idea stalled forecast bridge [Read it as text](#text) Whole tree Forecasts → Illusion Virtual worlds Computation 51,200 years ago Painted scenes before 300 BC Abacus 628 CE · Idea Zero as a number c. 1011–1021 Camera obscura explained 11th century Shadow theatre 9th century Self-playing flute (Baghdad) c. 820s CE · Idea Algorithms (al-Khwārizmī) c. 1415–1425 Linear perspective c. 1659 Magic lantern 1703 · Idea Binary arithmetic 1642–1644 Mechanical calculator 1832 Motion from stills 1787–1788 The panorama c. 1826–1827 Photography 1804 Punched-card loom 1822 · Stalled Difference Engine 1838 Stereoscope 1837 · Stalled Analytical Engine 1881 Stereo sound 1895 Projected moving pictures 1843 · Idea Note G 1847 · Idea Boolean logic 1904–1907 Vacuum tube 1929 Flight simulator 1922 · Stalled Stereoscopic film 1943–44 Colossus 1941 Zuse Z3 1936–37 · Idea Universal machine 1937–38 · Idea Logic in circuits 1945–46 Electronic computer 1945 · Idea The stored-program design 1957 Cochlear implant 1947–48 Transistor 1945 Stored-program computer 1962 · Stalled Sensorama 1960 Telesphere Mask 1964 · Idea Simulacron-3 1963 Interactive graphics 1958 First video games 1958–59 Integrated circuit 1965 · Idea The Ultimate Display 1965 · Idea Moore’s law 1968 Head-mounted display 1971–1980 Realistic shading 1972 Arcade video game 1971 Microprocessor 1969 Packet network 1975–76 Colossal Cave Adventure 1973–74 First-person networked world 1978–79 Aspen Movie Map 1985–1987 Data glove; “virtual reality” 1978–1980 Shared text world 1982–84 · Idea Cyberspace 1983 The Internet 1992 CAVE 1986–88 Graphical online world 1989–1991 World Wide Web 1993–1996 Real-time 3D worlds 1995–96 · Stalled Virtual Boy 1992 · Idea The Metaverse 1997–2003 Persistent online worlds 2004–2006 Brain-computer interface 1999 Graphics processor 2006–2007 General-purpose GPU computing 2007 Smartphone 2012–13 Modern VR headset 2012 Deep learning on GPUs 2015–2016 Room-scale tracking 2017 Transformer architecture 2018 Learned world models 2020–2023 Neural rendering 2018–2019 Standalone headsets 2020 Diffusion models 2024–2025 Generated interactive worlds 2029–2034 · Forecast Real-time photorealism 2027–2040 · Forecast Reversible computing 2010–2030s · Forecast Molecular nanotechnology 2027–2047 · Forecast Human-level AI c. 2250 or later · Forecast Star-scale computer 2050–2075 · Bridge Whole-brain emulation 2029–2030s · Forecast Full-immersion VR 2027–2045+ · Forecast Superintelligence Where it starts Show ▴ Start here ## Where it starts Every node with no line into it is a beginning. These lines started on their own and only later fed into each other. The oldest are painted scenes , the first rendered worlds, and the abacus , the oldest calculating device whose design survives. Your progress stays in this browser only. If you clear site data, the tree starts over. The whole tree as text · every node in order, for screen readers, text browsers and machine readers. Same nodes as the map Dates are years of the common era unless marked BC or BCE. Evolutionary dates are in years ago. Each node has a stable address, shown after it. Forecast nodes list published forecasts only, each cited. ## Illusion 1. **Painted scenes** (at least 51,200 years ago) A red-ochre painting in Leang Karampuang cave, Sulawesi, shows human-like figures and a wild pig. Uranium-series dating of the calcite over it puts it at least 51,200 years old. It’s the oldest known scene that tells a story. So people have been painting imagined or remembered worlds for a very long time. Led to: [Linear perspective](#node-perspective) Source: [Oktaviana et al., ‘Narrative cave art in Indonesia by 51,200 years ago’, Nature 631 (2024)](https://www.nature.com/articles/s41586-024-07541-7) Address: 2. **Camera obscura explained** (c. 1011–1021) China recorded earlier that light through a small hole makes an upside-down image. The Later Mohist Canon (compiled between the late 4th and mid-3rd century BCE) explains the flip by rays crossing at the hole. Shen Kuo’s Dream Pool Essays (1088) said it again, calling the inversion after the small hole a normal principle. Ibn al-Haytham wrote his Kitāb al-Manāẓir (Book of Optics) in Cairo in the early eleventh century. He tested the hole image by experiment. He also argued that we see by receiving light, not by sending it out from the eye. His Optics later fed into Renaissance theories of perspective. Led to: [Linear perspective](#node-perspective) ; [Magic lantern](#node-magic-lantern) ; [Photography](#node-photography) Source: [Encyclopaedia Britannica, ‘Kitab al-manazir’ (and ‘Ibn al-Haytham’)](https://www.britannica.com/topic/Kitab-al-manazir) Address: 3. **Shadow theatre** (11th century (securely documented)) In shadow theatre, flat figures behind a lit screen act out stories. It’s documented in Java by the 11th-century poem Arjunawiwaha (c. 1035), which compares the world to a puppet screen. It’s documented in China by Northern Song accounts (960–1127). Earlier claims are legend, or only name a performance without describing it. Examples are the Han story of Emperor Wu and the Jaha and Wukajana inscriptions of c. 840–907\. So the earliest date isn’t secure. Source: [Istari, Wukayana inscription of Balitung (907 CE), Berkala Arkeologi (BRIN); Wikipedia, ‘Wayang’ for the Jaha/Arjunawiwaha dates; Hanshu 97A tale as discussed by Gao Cheng](https://ejournal.brin.go.id/berkalaarkeologi/article/download/4742/3505/13391) Address: 4. **Linear perspective** (c. 1415–1425) Filippo Brunelleschi showed a geometric way to paint a scene as it looks from one spot. Alberti wrote the method down in De pictura (1435). The rule is that parallel lines all meet at a vanishing point. Real-time graphics software uses the same projection on every frame. Built on: [Painted scenes](#node-cave-painting) ; [Camera obscura explained](#node-camera-obscura) Led to: [The panorama](#node-panorama) ; [Photography](#node-photography) ; [Stereoscope](#node-stereoscope) Source: [Encyclopaedia Britannica, ‘Linear perspective’](https://www.britannica.com/art/linear-perspective) Address: 5. **Magic lantern** (c. 1659) In 1659 Christiaan Huygens drew a set of skeleton images to be projected by a lantern with convex lenses. That’s the earliest surviving record of the device. It was the first widely used way to throw a painted or photographic image on a wall for an audience. For over two centuries it was the standard tool for public image shows. Built on: [Camera obscura explained](#node-camera-obscura) Led to: [Projected moving pictures](#node-cinema) Source: [Magic Lantern Society (UK), ‘A brief history of the Magic Lantern’](https://magiclantern.org.uk/history/) Address: 6. **The panorama** (1787–1788) Robert Barker patented the panorama in 1787 and showed the first one, a view of Edinburgh, in 1788\. Later rotundas held round paintings about 18 metres across or more. Viewers stood in the middle of a painted horizon that never broke. It was an early commercial try at swapping a visitor’s surroundings for a built scene. Built on: [Linear perspective](#node-perspective) Source: [Encyclopaedia Britannica, ‘Robert Barker’ and ‘Panorama’](https://www.britannica.com/biography/Robert-Barker) Address: 7. **Photography** (c. 1826–1827) Nicéphore Niépce made the earliest surviving camera photograph around 1826–27\. It was a heliograph on a pewter plate, the view from his window at Le Gras, after an exposure of many hours. It fixed an optical image for good. So for the first time an image resembled the world through a machine, not through a hand. Built on: [Camera obscura explained](#node-camera-obscura) ; [Linear perspective](#node-perspective) Led to: [Projected moving pictures](#node-cinema) Source: [Harry Ransom Center, University of Texas, ‘The First Photograph’ (Niépce heliograph)](https://www.hrc.utexas.edu/niepce-heliograph/) Address: 8. **Motion from stills** (1832) Joseph Plateau in Ghent and Simon Stampfer in Vienna each came up with the phenakistiscope in 1832\. It’s a slotted disc of drawings in sequence. Spin it and look in a mirror, and the drawings seem to move. It showed that the eye blends a fast series of still images into smooth motion. That’s the principle behind every film and video frame. Led to: [Projected moving pictures](#node-cinema) Source: [Encyclopaedia Britannica, ‘Phenakistoscope’](https://www.britannica.com/science/phenakistoscope) Address: 9. **Stereoscope** (1838) In 1838 Charles Wheatstone showed that two slightly different flat drawings, one for each eye, merge into a single image with real-looking depth. He built a mirror stereoscope to prove it. The difference between the two eyes’ views became something you could control. Every headset since still uses it. Built on: [Linear perspective](#node-perspective) Led to: [Stereoscopic film](#node-3d-film) ; [Sensorama](#node-sensorama) ; [The Ultimate Display](#node-ultimate) ; [Telesphere Mask](#node-telesphere) Source: [Wheatstone, ‘Contributions to the physiology of vision, Part I’, Phil. Trans. R. Soc. 128 (1838), 371–394](https://doi.org/10.1098/rstl.1838.0019) Address: 10. **Stereo sound** (1881) At the 1881 Paris International Exposition of Electricity, Clément Ader’s system sent live performances from the Opera over telephone lines to visitors. It used paired transmitters and receivers, which gave a two-ear, spatial effect. A subscription service followed in 1890\. It was an early sign that a sense of place could be sent by electricity to a listener far away. Led to: [Sensorama](#node-sensorama) Source: [Bob’s Old Phones / Telephone Collecting, ‘Clément Ader’](https://telephonecollecting.org/Bobs%20phones/Pages/Essays/Ader/Ader.htm) Address: 11. **Projected moving pictures** (1895) On 28 December 1895 the Lumière brothers showed the Cinématographe to a paying crowd at the Grand Café in Paris. It was a camera, printer and projector in one. Showing film to a shared audience made recorded motion something people experienced together. That’s the template for later shared screens and rooms. Built on: [Photography](#node-photography) ; [Motion from stills](#node-phenakistiscope) ; [Magic lantern](#node-magic-lantern) Led to: [Stereoscopic film](#node-3d-film) ; [Sensorama](#node-sensorama) ; [Aspen Movie Map](#node-aspen) ; [Telesphere Mask](#node-telesphere) Source: [Encyclopaedia Britannica, ‘Cinématographe’](https://www.britannica.com/technology/Cinematographe) Address: 12. **Stereoscopic film** (1922; boom 1952–54; revival 2009 · Stalled) Stereoscopic cinema took Wheatstone’s idea and put it on film. Anaglyph features were shown in the early 1920s. The polarised Natural Vision process set off the 1952 boom with Bwana Devil. Audiences lost interest within about a year. The format came back with digital cinema in 2009. Why it stalled: It stalled because the picture was fixed. Every viewer saw the same two views no matter where they looked. Glasses and projector alignment were annoying. And the image had no way to respond to the viewer. Built on: [Projected moving pictures](#node-cinema) ; [Stereoscope](#node-stereoscope) Source: [Encyclopaedia Britannica, ‘3-D’ (film process) and ‘Natural Vision’](https://www.britannica.com/technology/3-D) Address: 13. **Flight simulator** (1929) Edwin Link built a cockpit driven by air pressure. It pitched, rolled and yawed as the pilot moved the controls, so trainees could practise instrument flying on the ground. The Allies used Link’s Blue Box in the Second World War, and more than 500,000 pilots are reported to have trained on such devices. It showed that a believable feel of flight could stand in for a real aircraft in training. Source: [ASME Landmark #210, ‘Link C-3 Flight Trainer’](https://www.asme.org/about-asme/engineering-history/landmarks/210-link-c-3-flight-trainer) Address: 14. **Cochlear implant** (1957; multichannel 1977–84) Djourno and Eyries in Paris reported in 1957 that they had electrically stimulated the auditory nerve of a deaf patient. House and Doyle implanted a single-channel device in 1961\. Multichannel devices (Hochmair, Clark) were implanted in 1977–78 and approved by the FDA for adults in the mid-1980s. Sound reaches the brain without a working ear. It was the first clinical case of a sense supplied by direct electrical stimulation of the nervous system. Led to: [Brain-computer interface](#node-bci) Source: [Lasker Foundation, ‘Modern cochlear implant’ (2013 Clinical Medical Research Award)](https://laskerfoundation.org/winners/modern-cochlear-implant/) Address: 15. **Telesphere Mask** (1960) Morton Heilig patented a head-worn viewer that gave each eye its own image, with stereo sound. That was years before any computer drove such a display. It played prerecorded film and didn’t track the head. Built on: [Stereoscope](#node-stereoscope) ; [Projected moving pictures](#node-cinema) Led to: [Head-mounted display](#node-damocles) Source: [Heilig, US Patent 2,955,156, “Stereoscopic-television apparatus for individual use” (1960)](https://patents.google.com/patent/US2955156A/en) Address: 16. **Sensorama** (1962 (patent; prototype 1957–62) · Stalled) Morton Heilig’s Sensorama Simulator (US patent 3,050,870, filed 1961, granted 1962) was an arcade booth. It played pre-filmed stereoscopic colour footage with stereo sound, vibration, wind and scent. It was the first try at engaging several senses at once in one machine. It’s the usual starting point for the history of virtual reality. Why it stalled: It stalled because everything was pre-recorded on film. No computer made or changed the scene, so the viewer could look but never change anything. Built on: [Projected moving pictures](#node-cinema) ; [Stereoscope](#node-stereoscope) ; [Stereo sound](#node-theatrophone) Source: [US Patent 3,050,870, M. L. Heilig, ‘Sensorama Simulator’ (filed 10 Jan 1961, granted 28 Aug 1962)](https://patents.google.com/patent/US3050870A/en) Address: ## Virtual worlds 1. **First video games** (1958; 1962) William Higinbotham’s Tennis for Two (October 1958, Brookhaven) drew a ball’s flight on an oscilloscope under analog-computer control. Spacewar! (1962) was a two-player space combat game written for MIT’s PDP-1 digital computer. Together they showed that a machine could host a small, shared world with rules that players act in. Built on: [Stored-program computer](#node-stored-program) Led to: [Arcade video game](#node-pong) Source: [Computer History Museum, ‘Spacewar!' (see also CHM timeline 1958 for Tennis for Two)](https://www.computerhistory.org/pdp-1/spacewar/) Address: 2. **Interactive graphics** (1963) Ivan Sutherland’s doctoral thesis at MIT (1963) described Sketchpad. It ran on the TX-2 and let a user draw and edit line diagrams with a light pen, with constraints and reusable master drawings. It made the interactive display a way to work on a model held in a computer. Built on: [Stored-program computer](#node-stored-program) Led to: [The Ultimate Display](#node-ultimate) ; [First-person networked world](#node-maze-war) ; [Realistic shading](#node-shading) Source: [I. E. Sutherland, ‘Sketchpad: a man-machine graphical communication system’, MIT PhD thesis (1963), University of Cambridge reprint, TR-574](https://www.cl.cam.ac.uk/techreports/UCAM-CL-TR-574.pdf) Address: 3. **Simulacron-3** (1964 · Idea) Daniel F. Galouye’s novel Simulacron-3 (1964; UK title Counterfeit World) describes a city simulated on a computer. Its people don’t know they’re programs. It’s one of the earliest novels about a world run inside a computer. It’s also a common reference point in writing on the simulation argument. Built on: [Stored-program computer](#node-stored-program) Source: [The Encyclopedia of Science Fiction, ‘Galouye, Daniel F.'](https://sf-encyclopedia.com/entry/galouye%5Fdaniel%5Ff) Address: 4. **The Ultimate Display** (1965 · Idea) In his 1965 paper to the IFIP Congress, Sutherland proposed a computer-driven display so complete that it could control matter in the room. He wrote that a chair displayed there would be good enough to sit in. It set the goal of virtual reality as a world you couldn’t tell from reality in any sense. Built on: [Interactive graphics](#node-sketchpad) ; [Stereoscope](#node-stereoscope) Led to: [Head-mounted display](#node-damocles) Source: [I. E. Sutherland, ‘The Ultimate Display’, Proc. IFIP Congress 1965, vol. 2, pp. 506–508 (scan)](https://worrydream.com/refs/Sutherland%5F1965%5F-%5FThe%5FUltimate%5FDisplay.pdf) Address: 5. **Head-mounted display** (1968) In 1968 Sutherland and Bob Sproull built a head-mounted display. It showed a user simple wireframe 3D graphics, which a computer redrew as the user’s head moved. The optics hung from the ceiling (the ‘Sword of Damocles’). It was the first system to combine a head-worn display, computer-generated images and head tracking. Built on: [The Ultimate Display](#node-ultimate) ; [Telesphere Mask](#node-telesphere) Led to: [Data glove; “virtual reality”](#node-vpl) ; [CAVE](#node-cave) ; [Virtual Boy](#node-virtualboy) ; [Modern VR headset](#node-rift) Source: [I. E. Sutherland, ‘A head-mounted three dimensional display’, AFIPS Fall Joint Computer Conference (1968)](https://doi.org/10.1145/1476589.1476686) Address: 6. **Realistic shading** (1971–1980) Henri Gouraud (1971) and Bui Tuong Phong (1973–75) devised smooth shading and a lighting model for polygon surfaces. Turner Whitted (1980) showed recursive ray tracing with reflection and refraction. Each step moved computer images closer to how light behaves. The techniques became the basis of film and game rendering. Built on: [Interactive graphics](#node-sketchpad) Led to: [CAVE](#node-cave) ; [Real-time 3D worlds](#node-doom) Source: [Gouraud, ‘Continuous shading of curved surfaces’, IEEE Trans. Computers C-20 (1971); also Phong, CACM 18 (1975), and Whitted, CACM 23 (1980)](https://doi.org/10.1109/T-C.1971.223313) Address: 7. **Arcade video game** (1972) Atari’s Pong was designed by Allan Alcorn from a concept by Nolan Bushnell. It was first placed in a Sunnyvale bar in late 1972 and sold in large numbers. It was built from logic chips, not a microprocessor. It made the video game a commercial public pastime. It also funded the industry that later drove graphics hardware. Built on: [First video games](#node-tennis-for-two) ; [Integrated circuit](#node-ic) Led to: [Cyberspace](#node-cyberspace) Source: [The Strong National Museum of Play, ‘Pong’](https://www.museumofplay.org/games/pong/) Address: 8. **First-person networked world** (1973–74) Maze was written at NASA Ames in 1973 by Steve Colley, Greg Thompson and Howard Palmer on Imlac PDS-1 computers. It showed the view down corridors from the player’s own position. It started with two linked machines, then added shooting and bots. It was an early first-person shared 3D space. It’s a likely ancestor of the networked first-person shooter. Built on: [Packet network](#node-arpanet) ; [Interactive graphics](#node-sketchpad) The sources leave this date or claim uncertain. Source: [DigiBarn Computer Museum, Steve Colley, ‘Maze War’ history](https://digibarn.com/history/04-VCF7-MazeWar/stories/colley.html) Address: 9. **Colossal Cave Adventure** (1975–76) Will Crowther wrote this text game in 1975, and Don Woods extended it in 1976\. A player explored a cave described in prose by typing commands. It set up the explorable, rule-bound world described in text. The first multi-user worlds built on it. Built on: [Stored-program computer](#node-stored-program) Led to: [Shared text world](#node-mud) Source: [Encyclopaedia Britannica, “Adventure”](https://www.britannica.com/topic/Adventure-electronic-game-by-Crowther-c-1975) Address: 10. **Shared text world** (1978–1980) Roy Trubshaw began MUD at the University of Essex in autumn 1978, and Richard Bartle joined the work. Once Essex was linked to ARPANET in spring 1980, outsiders could connect. It was a lasting multi-user world of rooms and objects that several people occupied at once. It set the pattern for later online worlds. Built on: [Packet network](#node-arpanet) ; [Colossal Cave Adventure](#node-adventure) Led to: [Graphical online world](#node-habitat) ; [Persistent online worlds](#node-mmo) Source: [Richard Bartle, ‘Early MUD history’](https://mud.co.uk/richard/mudhist.htm) Address: 11. **Aspen Movie Map** (1978–79) MIT’s Architecture Machine Group filmed Aspen street by street in 1978–79\. A user could drive through the stored footage on a touch-screen laserdisc system, turning left, turning right or stopping at will. It was an early interactive, navigable picture of a real place. It was a forerunner of street-level mapping. Built on: [Projected moving pictures](#node-cinema) ; [Stored-program computer](#node-stored-program) Source: [MIT Media Lab, ‘The Aspen Movie Map beat Google Street View by 34 years’](https://www.media.mit.edu/articles/the-aspen-movie-map-beat-google-street-view-by-34-years) Address: 12. **Cyberspace** (1982–84 · Idea) William Gibson coined the word ‘cyberspace’ in the story ‘Burning Chrome’ (Omni, July 1982). He made it popular in the novel Neuromancer (1984), as a shared graphical space of data that you can move around in. It gave a name and an image to the idea of an online world, and engineers later cited it. Built on: [Arcade video game](#node-pong) Led to: [The Metaverse](#node-metaverse) Source: [The Encyclopedia of Science Fiction, ‘Cyberspace’](https://sf-encyclopedia.com/entry/cyberspace) Address: 13. **Data glove; “virtual reality”** (1985–1987) Jaron Lanier founded VPL Research in 1985\. Its DataGlove (1987) and EyePhone headset were the first commercial VR products. Britannica credits Lanier with coining the term ‘virtual reality’ in 1987\. The company put body-tracking input and a head display within reach outside military and academic labs. Built on: [Head-mounted display](#node-damocles) Source: [Encyclopaedia Britannica, ‘Jaron Lanier’](https://www.britannica.com/biography/Jaron-Lanier) Address: 14. **Graphical online world** (1986–88) Habitat was built by Chip Morningstar and Randy Farmer at Lucasfilm Games for the Commodore 64, over the Quantum Link service. It ran as a pilot in 1986\. Users showed up as cartoon figures called avatars in a shared graphical world. It tested at scale what happens when thousands of people live in a designed virtual place. The builders published the lessons. Built on: [Shared text world](#node-mud) Led to: [The Metaverse](#node-metaverse) ; [Persistent online worlds](#node-mmo) Source: [Morningstar and Farmer, ‘The Lessons of Lucasfilm’s Habitat’ (1990)](https://www.fudco.com/chip/lessons.html) Address: 15. **CAVE** (1992) The CAVE was described in 1992 by Cruz-Neira, Sandin, DeFanti and colleagues at the University of Illinois at Chicago. It projected stereo images on the walls and floor of a room. Several people could stand inside a shared virtual scene wearing light glasses. It showed an immersive alternative to the head-worn display. Built on: [Head-mounted display](#node-damocles) ; [Realistic shading](#node-shading) Source: [Cruz-Neira, Sandin and DeFanti et al., ‘The CAVE: audio visual experience automatic virtual environment’, CACM 35(6) (1992)](https://doi.org/10.1145/129888.129892) Address: 16. **The Metaverse** (1992 · Idea) Neal Stephenson’s novel Snow Crash (1992) coined the word ‘metaverse’. It means a shared 3D online world where people appear as avatars. Technology firms revived the term decades later as a name for the next stage of online worlds. Built on: [Cyberspace](#node-cyberspace) ; [Graphical online world](#node-habitat) Source: [Encyclopaedia Britannica, ‘Snow Crash’](https://www.britannica.com/topic/Snow-Crash) Address: 17. **Real-time 3D worlds** (1993–1996) Id Software’s Doom (December 1993) put networked players in texture-mapped first-person levels on ordinary PCs. Quake (1996) drew true polygonal 3D with real-time lighting. Fast first-person 3D on consumer hardware created the market and the techniques for later virtual worlds. Built on: [Realistic shading](#node-shading) ; [Microprocessor](#node-microprocessor) Led to: [Persistent online worlds](#node-mmo) Source: [The Strong National Museum of Play, ‘Quake’ (with Encyclopaedia Britannica, ‘Doom’)](https://www.museumofplay.org/games/quake/) Address: 18. **Virtual Boy** (1995–96 · Stalled) Nintendo’s Virtual Boy (Japan July 1995, North America August 1995) was a table-mounted stereoscopic console. It showed red, wireframe-style images. It was pulled within a year. Nintendo’s own Iwata later called it a commercial failure. Why it stalled: It stalled because in 1995 there was no affordable compact full-colour display and no head tracking. The red single-colour image, fixed viewing position and awkward fit tired users quickly. Built on: [Head-mounted display](#node-damocles) ; [Integrated circuit](#node-ic) Source: [Iwata Asks, ‘Shigeru Miyamoto talks Virtual Boy’, Nintendo](https://www.nintendo.com/en-gb/Iwata-Asks/Iwata-Asks-Nintendo-3DS/Vol-1-And-That-s-How-the-Nintendo-3DS-Was-Made/2-Shigeru-Miyamoto-Talks-Virtual-Boy/2-Shigeru-Miyamoto-Talks-Virtual-Boy-229419.html) Address: 19. **Persistent online worlds** (1997–2003) Ultima Online (1997), EverQuest (1999) and Second Life (2003) put very large numbers of people into lasting graphical worlds. The worlds kept going when any player logged off. They had economies and social structures of their own. They’re the best evidence we have of how people behave in long-lived virtual places. Built on: [Graphical online world](#node-habitat) ; [Shared text world](#node-mud) ; [Real-time 3D worlds](#node-doom) ; [World Wide Web](#node-web) Source: [Encyclopaedia Britannica, ‘Ultima Online’ and ‘Online gaming’](https://www.britannica.com/topic/Ultima-Online) Address: 20. **Brain-computer interface** (2004–2006) In 2006 Hochberg and colleagues reported on a man paralysed from the neck down. He could move a computer cursor and run a simple robotic hand using signals from a microelectrode array implanted in his motor cortex. It showed that voluntary neural activity could be read out as a control signal. Built on: [Cochlear implant](#node-cochlear) ; [Microprocessor](#node-microprocessor) Led to: [Full-immersion VR](#node-fulldive) ; [Whole-brain emulation](#node-wbe) Source: [Hochberg et al., ‘Neuronal ensemble control of prosthetic devices by a human with tetraplegia’, Nature 442 (2006)](https://doi.org/10.1038/nature04970) Address: 21. **Modern VR headset** (2012–13 (developer kit); 2016 (consumer)) Oculus raised $2.4 million on Kickstarter in 2012\. It shipped the Rift DK1 developer kit from late March 2013\. The consumer Rift followed on 28 March 2016\. It paired phone-class displays and sensors with a GPU and low-latency head tracking. That made a wide-view headset cheap enough for developers and consumers. Built on: [Head-mounted display](#node-damocles) ; [Graphics processor](#node-gpu) ; [Smartphone](#node-smartphone) Led to: [Room-scale tracking](#node-roomscale) Source: [Road to VR, ‘3 years ago the Oculus Rift DK1 shipped, here’s a quick look back’](https://roadtovr.com/3-years-ago-the-oculus-rift-dk1-shipped-heres-a-quick-look-back/) Address: 22. **Room-scale tracking** (2015–2016) HTC and Valve announced the Vive in March 2015 and shipped it on 5 April 2016\. It used Valve’s Lighthouse system, which sweeps infrared lasers across a play area. A headset and two hand controllers could be tracked precisely over about four metres. Walking and reaching inside a virtual space, not just turning your head, became standard. Built on: [Modern VR headset](#node-rift) Led to: [Standalone headsets](#node-standalone) Source: [Road to VR, ‘HTC, Valve and SteamVR: a 2015 retrospective and CES 2016 preview’](https://roadtovr.com/htc-valve-and-steamvr-a-2015-retrospective-and-ces-2016-preview/) Address: 23. **Standalone headsets** (2018–2019) Oculus Quest (21 May 2019) was a self-contained headset. It used four cameras to track the headset and both controllers from the inside, with no PC or base stations. Earlier standalone units (Oculus Go, Lenovo Mirage Solo, Vive Focus, 2018) tracked at most the head position. The Quest made full tracking portable. Built on: [Room-scale tracking](#node-roomscale) ; [Smartphone](#node-smartphone) Led to: [Full-immersion VR](#node-fulldive) Source: [Road to VR, ‘Oculus announces Quest, the high-end standalone headset starting at $400’](https://roadtovr.com/oculus-announces-quest-high-end-standalone-headset-starting-400/) Address: 24. **Neural rendering** (2020–2023) Mildenhall et al.’s Neural Radiance Fields (2020) learned a scene from photographs and drew new views of it. 3D Gaussian Splatting (Kerbl et al., 2023) did it in real time. Photo-real 3D scenes you can explore can now be captured, not modelled by hand. Built on: [Deep learning on GPUs](#node-deep-learning) ; [Graphics processor](#node-gpu) Led to: [Generated interactive worlds](#node-world-models) ; [Real-time photorealism](#node-photoreal) Source: [Mildenhall et al., ‘NeRF: Representing scenes as neural radiance fields for view synthesis’, ECCV 2020, arXiv:2003.08934](https://arxiv.org/abs/2003.08934) Address: 25. **Generated interactive worlds** (2024–2025) In 2024, research systems such as Google DeepMind’s Genie and the GameNGen diffusion model generated each frame of a playable environment from the player’s input. They had no hand-written game engine. GameNGen simulated the game Doom at over 20 frames per second. By 2025 the demonstrations ran for minutes at interactive speeds. They’re still research prototypes. Built on: [Transformer architecture](#node-transformer) ; [Neural rendering](#node-neural-rendering) ; [Learned world models](#node-learned-worlds) ; [Diffusion models](#node-diffusion) Source: [Valevski et al., ‘Diffusion Models Are Real-Time Game Engines’ (GameNGen), arXiv:2408.14837 (2024)](https://arxiv.org/abs/2408.14837) Address: 26. **Real-time photorealism** (2029–2034 · Forecast) Industry figures have forecast real-time images you can’t tell from photographs. In 2024 one major vendor stated a window of 5 to 10 years. A 2013 forecast of ten years for the same milestone came due in 2023\. No one claimed it had been met. Built on: [Neural rendering](#node-neural-rendering) Published forecasts: * **Tim Sweeney (Epic Games)** (2013): within ten years of 2013 (c. 2023). [MCV/Develop conference report, 11 July 2013](https://www.mcvuk.com/business-news/publishing/epic-founder-expects-photorealistic-games-within-ten-years/) * **Jensen Huang (NVIDIA CEO)** (2024): 5–10 years (2029–2034) for real-time AI-generated pixels. [Press Q&A after GTC 2024 keynote, reported by Wccftech, 21 March 2024](https://wccftech.com/nvidia-ceo-believes-were-just-10-years-away-from-ai-generating-every-pixel-in-a-game-in-real-time/amp/) * **Jensen Huang (NVIDIA CEO)** (2025): Qualifier: no fully AI-generated graphics; hybrid rendering stays. [CES 2025 remarks, PCWorld, 7 January 2025](https://www.pcworld.com/article/2570615/nvidias-huang-video-games-will-never-be-entirely-rendered-using-ai.html) Address: 27. **Full-immersion VR** (2029–2030s · Forecast) One named futurist has published dates for VR that works at the level of the nervous system. He puts it in the late 2020s to the 2030s, by way of nanobots in the brain. We found no expert survey or independent forecast of this milestone. No such system has been demonstrated. Built on: [Brain-computer interface](#node-bci) ; [Standalone headsets](#node-standalone) ; [Molecular nanotechnology](#node-nanotech) Published forecasts: * **Ray Kurzweil** (2002): 2029\. [Remarks of 26 March 2002, Kurzweil Library page](https://www.writingsbyraykurzweil.com/?p=370709) * **Ray Kurzweil** (2015): 2030s. [Webinar reported by Peter Diamandis, Singularity Hub, 12 Oct 2015](https://singularityhub.com/2015/10/12/ray-kurzweils-wildest-prediction-nanobots-will-plug-our-brains-into-the-web-by-the-2030s/) * **Ray Kurzweil** (2024): ‘within our lifetimes’ (no year given on page). [The Singularity Is Nearer, Viking/Penguin, June 2024 (publisher description)](https://penguin.co.uk/books/462759/the-singularity-is-nearer-by-kurzweil-ray/9781529937503) Address: 28. **Whole-brain emulation** (2050–2075 · Bridge) Also in the [Simulations tree](https://simulation-hypothesis.com/trees/simulations?n=f-wbe) and [Becoming synthetic](https://simulation-hypothesis.com/trees/synthetic?n=s-wbe). The 2008 FHI roadmap judged that emulation was possible before mid-century only if models at the electrophysiological level are enough. Finer detail would need later dates. A 2014 Monte Carlo model gives a median of 2064, if it arrives at all. The Metaculus community median is about 2073. Built on: [Human-level AI](#node-agi) ; [Brain-computer interface](#node-bci) Published forecasts: * **Sandberg & Bostrom** (2008): before mid-century if electrophysiological models suffice; animal models one to two decades earlier; table of earliest dates by detail level (spiking network 2019–2042, electrophysiology 2033–2068, metabolome 2044–2087). [Whole Brain Emulation: A Roadmap, Technical Report 2008–3, Future of Humanity Institute, Oxford, 2008](https://www.futureofhumanityinstitute.org/s/2008-3.pdf) * **Anders Sandberg** (2014): 50% by 2064 (25th percentile 2053, 75th 2080), conditional on WBE ever arriving; 2.4% before 2040\. [Monte Carlo model of brain emulation development, FHI working paper 2014–1](https://web.archive.org/web/20210125040505/https://www.fhi.ox.ac.uk/reports/2014-1.pdf) * **Metaculus community (question 2813, first whole human brain emulated)** (2026): median about November 2073\. [Metaculus community forecast, as shown in search results on 16 October 2026 (check the live question)](https://www.metaculus.com/questions/2813/date-of-first-human-whole-brain-emulation) * **Robin Hanson** (2016): within perhaps a century of 2016\. [The Age of Em, Oxford University Press, 2016 (ISBN 978–0198754626)](https://ageofem.com/) * **Dario Amodei** (2024): outside the 5–10 years after powerful AI. [‘Machines of Loving Grace’, Oct 2024](https://darioamodei.com/machines-of-loving-grace) Address: 29. **Star-scale computer** (c. 2250 or later · Forecast) Dyson (1960) proposed that advanced civilisations would capture starlight. Bradbury (1997–99) extended this to nested shells that run computers on the star’s waste heat. The only dated estimate we found is Bradbury’s. It’s about 2250 if current silicon production trends continued, and within a few thousand years at most. Built on: [Moore’s law](#node-moores-law) ; [Molecular nanotechnology](#node-nanotech) Published forecasts: * **Freeman Dyson** (1960): no date given. [‘Search for Artificial Stellar Sources of Infra-Red Radiation’, Science 131(3414), 3 June 1960](https://doi.org/10.1126/science.131.3414.1667) * **Robert J. Bradbury** (1999): by 2250 from current silicon trends; ‘within at most a few thousand years’; once begun, \~20 years for minor bodies, 10–1000 years for gas giants. [‘Matrioshka Brains’, manuscript dated 21 July 1999 (not peer reviewed; concept from 1997 discussions)](https://www.gwern.net/doc/ai/scaling/hardware/1999-bradbury-matrioshkabrains.pdf) * **Stuart Armstrong & Anders Sandberg** (2013): no date; ‘in the foreseeable future’. [‘Eternity in six hours’, Acta Astronautica 89, 2013](https://doi.org/10.1016/j.actaastro.2013.04.002) * **Nick Bostrom** (2003): no date; capability figure only: \~10^42 operations per second for a planetary-mass computer. [‘Are You Living in a Computer Simulation?', Philosophical Quarterly 53(211), 2003](https://simulation-argument.com/simulation.pdf) Address: ## Computation 1. **Abacus** (uncertain; Mesopotamian counting boards of the 3rd millennium BC (disputed); oldest surviving board c. 300 BC) A counting board, or abacus, is a surface where counters stand for place values. By moving them you can add, subtract, multiply and divide. Mesopotamian examples are often claimed, but none survives and the claim is disputed. The oldest surviving board is the Salamis tablet (c. 300 BC), though some see it as a gaming board. It shows that calculation can be moved into a physical state that a mechanism, and later a circuit, can hold. Led to: [Mechanical calculator](#node-pascaline) The sources leave this date or claim uncertain. Source: [Encyclopaedia Britannica, ‘Abacus’](https://www.britannica.com/technology/abacus-calculating-device) Address: 2. **Zero as a number** (628 CE (Brahmagupta); placeholder dot earlier, date disputed · Idea) Brahmagupta’s Brāhmasphuṭasiddhānta (628 CE) is the earliest surviving text to treat zero as a number with its own rules. Adding or subtracting zero leaves a number unchanged. A number times zero is zero. His rules for dividing by zero were incomplete, and his 0 ÷ 0 = 0 is wrong. A dot used as a placeholder appears earlier in the Bakhshali manuscript. But radiocarbon dates for its folios run from as early as the 3rd–4th century CE to the 9th–10th. So the date of its zero is disputed. Led to: [Binary arithmetic](#node-binary) ; [Algorithms (al-Khwārizmī)](#node-algorithm) Source: [O’Connor and Robertson, ‘Brahmagupta’, MacTutor History of Mathematics (rules); Bodleian Libraries / University of Oxford, Bakhshali manuscript carbon dating release (2017) (dating caveat)](https://mathshistory.st-andrews.ac.uk/Biographies/Brahmagupta/) Address: 3. **Algorithms (al-Khwārizmī)** (c. 820s CE (Arabic original lost); Latin 12th century · Idea) In early 9th-century Baghdad, al-Khwārizmī wrote a treatise on the Hindu place-value numerals, including zero, and on calculating with them. The Arabic text is lost. The Latin translation, Algoritmi de numero Indorum, is known to differ a lot from it. Through it, and the later algorismus texts, the Hindu-Arabic numerals and their written arithmetic reached Europe. The word ‘algorithm’ comes from the Latinized form of his name. Built on: [Zero as a number](#node-zero) Led to: [Mechanical calculator](#node-pascaline) Source: [O’Connor and Robertson, ‘Abu Ja’far Muhammad ibn Musa Al-Khwarizmi’, MacTutor History of Mathematics (see also Encyclopaedia Britannica, ‘al-Khwarizmi’ and ‘algorismus’)](https://mathshistory.st-andrews.ac.uk/Biographies/Al-Khwarizmi/) Address: 4. **Self-playing flute (Baghdad)** (9th century) The Banū Mūsā brothers wrote their Book of Ingenious Devices in 9th-century Baghdad. It describes a flute that plays by itself. A water-driven mechanism pushes air to sound pipes under a revolving cylinder, and swapping the cylinder changes the tune. Koetsier (2001) counts it among the earliest programmable machines. Al-Jazari’s 1206 book describes later musical automata. But no documented line runs from either work to the European pinned barrel or to the Jacquard loom. Source: [Koetsier, ‘On the prehistory of programmable machines: musical automata, looms, calculators’, Mechanism and Machine Theory 36(5), 2001, pp. 589–603; Hill (transl.), The Book of Ingenious Devices (Reidel, 1979)](https://doi.org/10.1016/S0094-114X%2801%2900005-2) Address: 5. **Mechanical calculator** (1642–1644) From 1642 Blaise Pascal built a geared adding machine to help his father’s tax work. He made about fifty. Britannica calls it the first calculator made in quantity and actually used. Arithmetic moved from hands to gears. Built on: [Abacus](#node-abacus) ; [Algorithms (al-Khwārizmī)](#node-algorithm) Led to: [Analytical Engine](#node-analytical-engine) ; [Difference Engine](#node-difference-engine) Source: [Encyclopaedia Britannica, ‘Pascaline’](https://www.britannica.com/technology/Pascaline) Address: 6. **Binary arithmetic** (1703 (printed 1705) · Idea) In a manuscript dated 15 March 1679 (De progressione dyadica), Leibniz set out arithmetic in base two. He also sketched a calculating machine without wheels. In it, balls drop through holes that sliding slots open and close, and that does the sums. He never built it, though a working model was made at the Deutsches Museum in the early 2000s. His paper was read to the Paris Academy in 1703 and printed in 1705\. It also links the 0–1 notation to the hexagram figures of the Yijing, attributed to Fu Xi, which Joachim Bouvet had sent him from Beijing (received April 1703). An earlier Indian scheme appears in Piṅgala’s Chandaḥśāstra (date uncertain, about the 3rd to 2nd century BCE). It lists metres of short and long syllables in a way later readers analysed as binary. Built on: [Zero as a number](#node-zero) Led to: [Zuse Z3](#node-z3) Source: [Leibniz, ‘Explanation of Binary Arithmetic’ (1703), English translation](https://www.leibniz-translations.com/binary.htm) Address: 7. **Punched-card loom** (1804) Joseph Marie Jacquard’s 1804 loom used a chain of punched cards to pick which warp threads to lift. To change the pattern, you changed the cards. Bouchon (1725) and Falcon (1728) had used punched paper or cards before him. But Jacquard’s version was widely adopted, and Babbage borrowed its card-reading idea. Led to: [Analytical Engine](#node-analytical-engine) Source: [Science Museum Group, ‘Model of Falcon’s loom’ (Jacquard refined Falcon’s method in 1804)](https://collection.sciencemuseumgroup.org.uk/objects/co44927/model-of-falcons-loom-looms) Address: 8. **Difference Engine** (1822 (proposed); built 1991 · Stalled) Babbage designed a machine to compute and print mathematical tables by the method of finite differences. It would take human error out of the arithmetic. A small working section was shown. The full engine wasn’t built in his lifetime. Why it stalled: Work halted in 1833 after a dispute with his engineer, Joseph Clement. The Science Museum finished Difference Engine No. 2 from his drawings in 1991\. That showed the design itself was sound. Built on: [Mechanical calculator](#node-pascaline) Led to: [Analytical Engine](#node-analytical-engine) Source: [Science Museum, “Charles Babbage’s Difference Engines and the Science Museum”](https://www.sciencemuseum.org.uk/objects-and-stories/charles-babbages-difference-engines-and-science-museum) Address: 9. **Analytical Engine** (1837 (conceived 1834) · Stalled) Charles Babbage designed a programmable mechanical computer. It had a store, an arithmetic mill, conditional branching and instructions on punched cards. It anticipated the layout of later computers by a century. Only a trial piece was built before he died in 1871. Why it stalled: It stalled because it needed thousands of parts, and 1840s Britain couldn’t supply the precision engineering, steady funding and settled design that took. The Computer History Museum lists runaway costs, high precision demands, a dispute with his engineer and patchy financing. Built on: [Punched-card loom](#node-jacquard) ; [Mechanical calculator](#node-pascaline) ; [Difference Engine](#node-difference-engine) Led to: [Note G](#node-lovelace) Source: [Computer History Museum, ‘Babbage: a brief history’](https://www.computerhistory.org/babbage/history/) Address: 10. **Note G** (1843 · Idea) Ada Lovelace translated Menabrea’s paper and added notes. Note G is a worked table of operations for computing Bernoulli numbers on the Analytical Engine. She also remarked that the engine might act on things other than number, such as musical composition, if their relations could be expressed. That second claim is the first clear statement that a computing machine handles symbols, not just quantities. Built on: [Analytical Engine](#node-analytical-engine) Source: [Menabrea and Lovelace, ‘Sketch of the Analytical Engine’, Taylor’s Scientific Memoirs (1843), text at Fourmilab](https://www.fourmilab.ch/babbage/sketch.html) Address: 11. **Boolean logic** (1847; 1854 · Idea) George Boole’s Mathematical Analysis of Logic (1847) and An Investigation of the Laws of Thought (1854) treated logical statements as an algebra with two values. That meant reasoning could be calculated. Shannon later showed that circuits carry out this algebra. So Boole’s system became the logical basis of digital hardware. Led to: [Universal machine](#node-turing) ; [Logic in circuits](#node-shannon) Source: [George Boole, An Investigation of the Laws of Thought (1854), Project Gutenberg](https://www.gutenberg.org/ebooks/15114) Address: 12. **Vacuum tube** (1904–1907) John Ambrose Fleming’s two-electrode valve (1904) turned radio signals into a usable form. Lee de Forest added a grid to make the three-electrode Audion (1906–07), which could amplify and switch electrical signals. The triode was the working part of radio, telephony and the first electronic computers. Led to: [Electronic computer](#node-eniac) ; [Transistor](#node-transistor) ; [Colossus](#node-colossus) Source: [Encyclopaedia Britannica, ‘Thermionic valve’](https://www.britannica.com/technology/thermionic-valve) Address: 13. **Universal machine** (1936–37 · Idea) Alan Turing’s ‘On Computable Numbers’ (read in 1936, published 1936–37) defined an abstract machine. It showed that one universal machine can do the work of any other, given that other’s description as input. That’s the theoretical reason one general computer can run any simulated world. It also sets the limits on what such a simulation can compute. Built on: [Boolean logic](#node-boole) Led to: [The stored-program design](#node-edvac) Source: [A. M. Turing, ‘On Computable Numbers, with an Application to the Entscheidungsproblem’, Proc. London Math. Soc. s2–42 (1937)](https://doi.org/10.1112/plms/s2-42.1.230) Address: 14. **Logic in circuits** (1937–38 · Idea) Claude Shannon’s master’s thesis at MIT (1937, published 1938) showed that networks of relays and switches can be analysed and designed with Boolean algebra. It tied abstract logic to physical circuits. Engineers now had a systematic way to design digital machines. Built on: [Boolean logic](#node-boole) Led to: [Electronic computer](#node-eniac) Source: [C. E. Shannon, ‘A Symbolic Analysis of Relay and Switching Circuits’, MIT thesis (1937)](https://dspace.mit.edu/handle/1721.1/11173) Address: 15. **Zuse Z3** (1941) Konrad Zuse completed the Z3 in Berlin in 1941\. It had about 2,400 relays, a binary number system and 22-bit floating-point words. The program was read in sequence from punched tape. The IEEE Computer Society’s history site calls it the first fully working program-controlled electromechanical digital computer in the world. It had no conditional branch, and it kept its program outside the machine’s memory. Built on: [Binary arithmetic](#node-binary) Source: [IEEE Computer Society, ‘Konrad Zuse’, Computer Pioneers; Rojas, ‘Konrad Zuse’s Legacy: The Architecture of the Z1 and Z3’, IEEE Annals of the History of Computing 19(2), 1997](https://history.computer.org/pioneers/zuse.html) Address: 16. **Colossus** (1943–44) Tommy Flowers’s electronic machines at Bletchley Park read cipher text from paper tape. They used valves (2,500 in the Mark 2) to find the wheel settings of the German Lorenz cipher. They were electronic and configurable, but built for one task. They were kept secret for many years after the war. Built on: [Vacuum tube](#node-vacuum-tube) Source: [The National Museum of Computing, “Colossus”](https://www.tnmoc.org/colossus) Address: 17. **Electronic computer** (1945–46) ENIAC was completed at the University of Pennsylvania in 1945 and announced publicly in February 1946\. It used about 18,000 vacuum tubes. You could rewire it to solve a wide range of numerical problems. It showed that electronic speed was practical for general computing. But reprogramming took days. Built on: [Vacuum tube](#node-vacuum-tube) ; [Logic in circuits](#node-shannon) Led to: [The stored-program design](#node-edvac) Source: [Encyclopaedia Britannica, ‘ENIAC’](https://www.britannica.com/technology/ENIAC) Address: 18. **Stored-program computer** (1945 (design); 1948 (first run)) On 21 June 1948 the Manchester Small-Scale Experimental Machine (‘Baby’), built by Williams and Kilburn, ran its first program from its own electronic memory. The program was 17 instructions long and searched for the highest proper factor of a number. Keeping instructions and data in the same memory is what lets software change at electronic speed. It’s also what lets simulations be loaded. Built on: [The stored-program design](#node-edvac) Led to: [Packet network](#node-arpanet) ; [Interactive graphics](#node-sketchpad) ; [Simulacron-3](#node-simulacron) ; [First video games](#node-tennis-for-two) ; [Aspen Movie Map](#node-aspen) ; [Colossal Cave Adventure](#node-adventure) Source: [University of Manchester, ‘Manchester Baby computer’ (Computer 50 archive)](https://curation.cs.manchester.ac.uk/computer50/www.computer50.org/mark1/new.baby.html) Address: 19. **The stored-program design** (1945 · Idea) John von Neumann’s First Draft of a Report on the EDVAC described a computer that keeps its program in the same memory as its data. Almost every computer since has used this layout. Built on: [Electronic computer](#node-eniac) ; [Universal machine](#node-turing) Led to: [Stored-program computer](#node-stored-program) Source: [von Neumann, First Draft of a Report on the EDVAC (30 June 1945), Internet Archive](https://archive.org/details/firstdraftofrepo00vonn) Address: 20. **Transistor** (1947–48) John Bardeen and Walter Brattain demonstrated a point-contact transistor at Bell Laboratories in December 1947\. William Shockley developed the junction form soon after. The three shared the 1956 Nobel Prize in Physics. It was a solid-state amplifier and switch with no heater and no vacuum. Computers became smaller, faster and far more reliable. Built on: [Vacuum tube](#node-vacuum-tube) Led to: [Integrated circuit](#node-ic) Source: [Nobel Prize in Physics 1956, summary](https://www.nobelprize.org/prizes/physics/1956/summary/) Address: 21. **Integrated circuit** (1958–59) Jack Kilby at Texas Instruments built a working integrated circuit in September 1958\. Robert Noyce at Fairchild devised a planar silicon version that could be manufactured in 1959\. Kilby shared the 2000 Nobel Prize in Physics. Putting many parts on one piece of material made complex electronics cheap. That set up the scaling of computing power. Built on: [Transistor](#node-transistor) Led to: [Moore’s law](#node-moores-law) ; [Microprocessor](#node-microprocessor) ; [Arcade video game](#node-pong) ; [Virtual Boy](#node-virtualboy) Source: [Nobel Prize in Physics 2000, Jack S. Kilby facts](https://www.nobelprize.org/prizes/physics/2000/kilby/facts/) Address: 22. **Moore’s law** (1965 (revised 1975) · Idea) In an April 1965 article in Electronics, Gordon Moore observed that the number of components per integrated circuit had doubled about every year. He projected the trend for a decade. In 1975 he changed the rate to about every two years. The projection held roughly for five decades. It set the pace at which simulated worlds could grow richer. Built on: [Integrated circuit](#node-ic) Led to: [Star-scale computer](#node-matrioshka) ; [Molecular nanotechnology](#node-nanotech) ; [Reversible computing](#node-reversible) Source: [Computer History Museum, ‘Moore’s Law predicts the future of integrated circuits’](https://www.computerhistory.org/siliconengine/moores-law-predicts-the-future-of-integrated-circuits/) Address: 23. **Packet network** (1969) ARPANET sent its first host-to-host message from UCLA to SRI on 29 October 1969\. Minicomputers called IMPs routed the packets. It grew into a national research network. Shared virtual spaces need machines that swap data in real time. ARPANET was the first large network built on packet switching. Built on: [Stored-program computer](#node-stored-program) Led to: [First-person networked world](#node-maze-war) ; [Shared text world](#node-mud) ; [The Internet](#node-internet) Source: [Computer History Museum, ‘October 29, 1969: Happy 40th birthday to a radical idea’](https://computerhistory.org/blog/october-29-1969-happy-40th-birthday-to-a-radical-idea/) Address: 24. **Microprocessor** (1971) Intel’s 4004 came out in November 1971\. Hoff, Mazor, Faggin and Shima designed it for a Busicom calculator. It put a programmable central processor on a single chip. General computing became cheap enough for personal machines, consoles and, eventually, phones. Built on: [Integrated circuit](#node-ic) Led to: [Graphics processor](#node-gpu) ; [Smartphone](#node-smartphone) ; [Real-time 3D worlds](#node-doom) ; [Brain-computer interface](#node-bci) Source: [Encyclopaedia Britannica, ‘Intel 4004’](https://www.britannica.com/technology/Intel-4004) Address: 25. **The Internet** (1983) In January 1983 the ARPANET switched to the TCP/IP protocols. They’re the common language that lets separate networks connect. Shared online worlds run over this network of networks. Built on: [Packet network](#node-arpanet) Led to: [World Wide Web](#node-web) Source: [Computer History Museum, “Internet History of 1980s”](https://www.computerhistory.org/internethistory/1980s/) Address: 26. **World Wide Web** (1989–1991) Tim Berners-Lee proposed a linked hypertext system at CERN in March 1989\. By the end of 1990 he had the first server and browser running there. He announced the software publicly in August 1991\. It made the Internet usable for non-specialists. It became the common way to reach shared online spaces. Built on: [The Internet](#node-internet) Led to: [Persistent online worlds](#node-mmo) Source: [CERN, ‘A short history of the Web’](https://home.cern/science/computing/the-birth-of-the-web/short-history-web/) Address: 27. **Graphics processor** (1999) Nvidia’s GeForce 256 (October 1999) put transformation and lighting on a single chip. It was marketed as the first graphics processing unit. That moved per-pixel and per-vertex work off the CPU. Cheap parallel graphics hardware is what put real-time 3D worlds within reach. Later it did the same for large-scale neural computing. Built on: [Microprocessor](#node-microprocessor) Led to: [General-purpose GPU computing](#node-gpgpu) ; [Modern VR headset](#node-rift) ; [Neural rendering](#node-neural-rendering) Source: [NVIDIA press release, ‘NVIDIA launches the world’s first graphics processing unit: GeForce 256’ (1999)](https://nvidianews.nvidia.com/news/nvidia-launches-the-world-s-first-graphics-processing-unit-geforce-256) Address: 28. **General-purpose GPU computing** (2006–2007) Nvidia announced CUDA in 2006 and released it in June 2007\. It let programmers write ordinary C-like code for a GPU’s thousands of parallel threads. Graphics hardware became a general parallel computer. That’s what later made training large neural networks practical. Built on: [Graphics processor](#node-gpu) Led to: [Deep learning on GPUs](#node-deep-learning) Source: [Nickolls, Buck, Garland and Skadron, ‘Scalable Parallel Programming with CUDA’, ACM Queue 6(2) (2008)](https://doi.org/10.1145/1365490.1365500) Address: 29. **Smartphone** (2007) Apple announced the iPhone in January 2007 and released it in June. It put a touch display, a capable processor and motion sensors in a pocket device, and mass production of those parts followed. The result was a supply of cheap high-density displays, motion sensors and mobile chips. That made consumer headsets possible. Built on: [Microprocessor](#node-microprocessor) Led to: [Modern VR headset](#node-rift) ; [Standalone headsets](#node-standalone) Source: [Apple Newsroom, ‘Apple reinvents the phone with iPhone’ (9 January 2007)](https://www.apple.com/newsroom/2007/01/09Apple-Reinvents-the-Phone-with-iPhone/) Address: 30. **Deep learning on GPUs** (2012) In September 2012 AlexNet won the ImageNet image-classification contest by a wide margin. It was a convolutional network that Krizhevsky, Sutskever and Hinton trained on two GPUs. The result showed that large neural networks trained on parallel hardware beat hand-built methods at perception tasks. Built on: [General-purpose GPU computing](#node-gpgpu) Led to: [Transformer architecture](#node-transformer) ; [Neural rendering](#node-neural-rendering) ; [Learned world models](#node-learned-worlds) ; [Diffusion models](#node-diffusion) Source: [Krizhevsky, Sutskever and Hinton, ‘ImageNet classification with deep convolutional neural networks’, NeurIPS 25 (2012)](https://papers.nips.cc/paper/2012/hash/c399862d3b9d6b76c8436e924a68c45b-Abstract.html) Address: 31. **Transformer architecture** (2017) Vaswani and colleagues at Google introduced the Transformer in 2017\. It’s a network built on attention. It handles sequences in parallel and scales well with data and computing power. It’s the architecture behind large language models, which can produce open-ended, conversational behaviour. Built on: [Deep learning on GPUs](#node-deep-learning) Led to: [Generated interactive worlds](#node-world-models) ; [Human-level AI](#node-agi) Source: [Vaswani et al., ‘Attention Is All You Need’, NeurIPS (2017), arXiv:1706.03762](https://arxiv.org/abs/1706.03762) Address: 32. **Learned world models** (2018) Ha and Schmidhuber trained a neural network to model a game environment. Then they trained an agent inside the model’s own imagined version of it. A world could now be learned from watching, not written by hand. Built on: [Deep learning on GPUs](#node-deep-learning) Led to: [Generated interactive worlds](#node-world-models) Source: [Ha and Schmidhuber, “World Models”, arXiv:1803.10122 (2018)](https://arxiv.org/abs/1803.10122) Address: 33. **Diffusion models** (2020) Ho, Jain and Abbeel showed that a network trained to remove noise step by step can generate high-quality images from random input. The method sits under most later image, video and interactive-world generators. Built on: [Deep learning on GPUs](#node-deep-learning) Led to: [Generated interactive worlds](#node-world-models) Source: [Ho, Jain and Abbeel, “Denoising Diffusion Probabilistic Models”, arXiv:2006.11239 (2020)](https://arxiv.org/abs/2006.11239) Address: 34. **Molecular nanotechnology** (2010–2030s · Forecast) Forecasters in the 1990s–2000s dated programmable molecular assembly and brain-entering nanobots to 2010–2029\. Diamondoid assembly was forecast for 2010–2015 and hasn’t been demonstrated. So this is a record of dates that passed. Built on: [Moore’s law](#node-moores-law) Led to: [Full-immersion VR](#node-fulldive) ; [Star-scale computer](#node-matrioshka) Published forecasts: * **Robert J. Bradbury, citing Zyvex’s Jim Von Ehr and Drexler 1998** (1999): 2010–2015 (Zyvex: by 2010). [‘Matrioshka Brains’ manuscript, 1999](https://www.gwern.net/doc/ai/scaling/hardware/1999-bradbury-matrioshkabrains.pdf) * **Ray Kurzweil** (2002): by 2029\. [Remarks 26 March 2002, Kurzweil Library](https://www.writingsbyraykurzweil.com/?p=370709) * **Ray Kurzweil** (2015): 2030s. [Singularity Hub report, 12 Oct 2015](https://singularityhub.com/2015/10/12/ray-kurzweils-wildest-prediction-nanobots-will-plug-our-brains-into-the-web-by-the-2030s/) Address: 35. **Human-level AI** (2027–2047 · Forecast) Forecasts for machines that beat humans at every task run from 2026–2030, in statements by AI-lab leaders, to a 2047 median in the 2023 survey of 2,778 researchers. Surveys have moved earlier over time. They gave 2040–2050 in 2012–13, 2060 in 2022 and 2047 in 2023. Built on: [Transformer architecture](#node-transformer) Led to: [Whole-brain emulation](#node-wbe) ; [Superintelligence](#node-asi) Published forecasts: * **Grace et al. (AI Impacts), survey of 2,778 AI researchers, run October 2023** (2024): HLMI: 10% by 2027, 50% by 2047; full automation of all occupations 10% by 2037, 50% by 2116\. [arXiv:2401.02843 ‘Thousands of AI Authors on the Future of AI’ (Jan 2024; v3 Oct 2025)](https://arxiv.org/abs/2401.02843) * **Müller & Bostrom, survey of four expert groups (2012–13)** (2016): 50% chance of high-level machine intelligence around 2040–2050; 90% by 2075\. [‘Future progress in artificial intelligence: A survey of expert opinion’, in Fundamental Issues of AI, Springer, 2016](https://doi.org/10.1007/978-3-319-26485-1%5F33) * **Metaculus community (question 5121, first general AI system)** (2026): median April 2030; interquartile range about 2026 to 2038\. [Metaculus community forecast, as shown in search results on 16 October 2026 (check the live question)](https://www.metaculus.com/questions/5121/) * **Dario Amodei (Anthropic CEO)** (2024): ‘powerful AI’ as early as 2026, or much later. [‘Machines of Loving Grace’, Oct 2024](https://darioamodei.com/machines-of-loving-grace) * **Demis Hassabis (Google DeepMind CEO)** (2025): capabilities emerging in 5–10 years (2030–2035). [CNBC interview, 17 March 2025](https://www.cnbc.com/2025/03/17/human-level-ai-will-be-here-in-5-to-10-years-deepmind-ceo-says.html) * **Ray Kurzweil** (2005): computers pass the Turing test / human-level intelligence by 2029\. [The Singularity Is Near (Viking, 2005; ISBN 978–0670033843), reaffirmed in The Singularity Is Nearer (Viking, June 2024)](https://aiimpacts.org/kurzweil-the-singularity-is-near/) * **Hans Moravec** (1998): hardware matching the brain in home computers before 2030\. [‘When will computer hardware match the human brain?', Journal of Evolution and Technology vol. 1, 1998](https://jetpress.org/volume1/moravec.htm) Address: 36. **Reversible computing** (2027–2040 · Forecast) Reversible logic avoids the Landauer minimum energy cost of erasing bits. It’s proposed as the route to far more computing per watt. A startup projects a first inference chip in 2027 and a 4,000-fold efficiency gain 10 to 15 years later. A 2014 expert interview didn’t expect wide use by 2030. Built on: [Moore’s law](#node-moores-law) Published forecasts: * **Hannah Earley (Vaire CTO); Michael Frank** (2025): chip in 2027; 4,000x in 10–15 years (2035–2040). [Dina Genkina, ‘Reversible Computing Escapes the Lab in 2025’, IEEE Spectrum, 2 Jan 2025 (update 22 May 2025)](https://spectrum.ieee.org/reversible-computing) * **Erik DeBenedictis (Sandia)** (2014): not widely used by 2030; some applications by 2030\. [MIRI interview, 3 April 2014](https://intelligence.org/2014/04/03/erik-debenedictis/) Address: 37. **Superintelligence** (2027–2045+ · Forecast) Surveyed experts expect superintelligence within 30 years of human-level machine intelligence. Scenario authors and Kurzweil give dates between 2027 and 2045\. Researchers surveyed in 2023 put substantial but not majority odds on a fast ‘intelligence explosion’. Built on: [Human-level AI](#node-agi) Published forecasts: * **Müller & Bostrom** (2016): within less than 30 years after HLMI. [Future progress in AI: a survey of expert opinion, 2016](https://doi.org/10.1007/978-3-319-26485-1%5F33) * **Kokotajlo, Alexander, Larsen, Lifland, Dean** (2025): ASI Dec 2027 in the scenario; medians later (e.g. one author’s median 2030). [AI 2027 scenario, 3 April 2025 (updated Nov-Dec 2025)](https://ai-2027.com/) * **Ray Kurzweil** (2005): 2045\. [The Singularity Is Near, 2005 (via AI Impacts summary)](https://aiimpacts.org/kurzweil-the-singularity-is-near/) Address: