AI and AR get lumped together in headlines because they're converging right now, but they're genuinely separate technologies with separate histories — one is about machines doing something resembling thinking, the other is about layering digital information onto the physical world you can see.
AI: from rules to learning
The term "artificial intelligence" dates to a 1956 academic conference at Dartmouth. Early AI ("symbolic AI," roughly 1950s–1980s) worked by hand-coding logical rules — useful for narrow, well-defined problems, but brittle outside them. Interest and funding rose and fell more than once in cycles researchers call "AI winters," as the hand-coded approach kept hitting a ceiling.
The shift to "machine learning" — a system learning patterns from data instead of following rules a person wrote — gradually took over. A widely cited turning point was 2012, when a deep learning approach dramatically outperformed competitors at the ImageNet image-recognition competition, convincing much of the field that this approach genuinely worked better. That set off a decade of rapid progress that led to today's large language models and generative AI — systems that generate text, images, and more from a prompt.
AR: overlaying the real world
Ivan Sutherland built what's usually credited as the first head-mounted AR display in 1968 — nicknamed "the Sword of Damocles" because it was so heavy it had to hang from the ceiling. For decades after, AR mostly stayed in military and industrial research (heads-up displays for pilots, for example), far from consumer use.
Smartphones changed that by putting a camera, a screen, and enough processing power in everyone's pocket at once. Apple's ARKit and Google's ARCore (both 2017) gave developers standard tools to build AR experiences on ordinary phones, and headset makers have continued pushing toward standalone AR and mixed-reality hardware since.
Where they're converging
The current wave of AR hardware increasingly leans on AI — for recognizing what a camera is looking at, translating text in real time, or generating a digital object that convincingly fits a real space. The two technologies developed on almost entirely separate tracks for fifty years and are only now genuinely merging.