From SohajPedia, the daily world briefing
Artificial intelligence
Artificial intelligence (AI) is the ability of computer systems to do things that normally need human intelligence — learning, reasoning, understanding language, recognising images and making decisions. It is also the name of the research field that builds those systems, a field that began, formally, in a summer workshop in 1956 and spent decades being written off before coming back to reshape almost every industry.
How AI got here
What AI actually is
Strip away the marketing and AI is software that improves at a task by finding patterns, instead of following only fixed instructions a programmer wrote. A spam filter is AI. So is the system that recognises your face to unlock your phone, the one that finishes your sentences in a messaging app, and the one that beat the world chess champion in 1997. Researchers disagree about how close any of this is to real intelligence — but the practical definition is simple: if a machine does something that would require intelligence in a person, it counts as AI.
A short, bumpy history
The field was founded as an academic discipline in 1956, at a workshop at Dartmouth College in the United States. Early progress was real but narrow, and it was repeatedly oversold; when results failed to match the promises, funding dried up in periods researchers still call "AI winters". The comeback began in 2012, when a deep learning model trained on graphics processors (GPUs) beat rival image-recognition systems by a wide margin. In 2017, researchers introduced the transformer — the architecture behind today's large language models. Then, on 30 November 2022, ChatGPT was released and became the fastest-growing consumer software application in history, taking AI from research labs into ordinary homes within weeks.
How modern AI learns
Most modern systems are built with machine learning: rather than being told rules, the model is shown enormous numbers of examples and adjusts millions — sometimes billions — of internal settings until its answers get better. Deep learning does this with layered artificial neural networks, loosely inspired by the brain. Large language models are trained on vast amounts of text, which is why they can write, summarise and translate; the same basic approach, fed with pictures, sound or video, produces the image and voice tools now everywhere. The catch is built in: a model learns patterns, not truth, which is why AI can produce confident-sounding mistakes — "hallucinations" — that a human has to catch.
Where you already meet it
Search results, video recommendations, maps that predict traffic, photo search, voice assistants, fraud checks at your bank, translation, and the autocorrect that just fixed your typing — all of it is AI of one kind or another. In industry it reads medical scans, inspects factory output, forecasts demand and writes first drafts of code and copy. Companies such as OpenAI, Google DeepMind and Meta have stated their aim openly: artificial general intelligence, a system that could handle nearly any mental task at least as well as a person. Whether that is near or far away is the field's biggest argument.
The argument about rules
Success has brought scrutiny. AI systems consume large amounts of electricity, can reproduce bias present in their training data, and are getting good enough to worry regulators about safety and misuse. Governments are responding at different speeds — Japan, for example, is moving to make AI safety measures legally binding, while airlines and banks are setting their own limits on what AI agents may do; SohajPedia's briefing on the worldwide tightening of AI rules covers the latest moves. The technology's direction is no longer just a technical question; it is a political one too.
Frequently asked questions
See also
References
Photograph: the Kismet robot, via Wikimedia Commons.