Ask ten researchers when machines will think like people and you are likely to hear ten different answers, ranging from "within a decade" to "probably never". The disagreement says as much about how hard it is to define intelligence as it does about the pace of technology.
What counts as "true" AI?
Artificial intelligence is an umbrella term. Most systems in use today are narrow: they perform specific tasks such as recognising speech, recommending products, translating text or spotting patterns in images, often extremely well. "True" AI, sometimes called artificial general intelligence, usually refers to a system that can learn and reason flexibly across many domains, the way a person can. Some definitions add self-awareness or understanding, which raises philosophical questions that are difficult to test.
A short look back
The term artificial intelligence was coined in the mid-1950s, when researchers were optimistic that human-level reasoning was within reach. Progress turned out to be slower and more uneven. Funding cycles rose and fell, with periods of reduced interest often called AI winters. The recent wave of progress grew from machine learning: systems trained on large amounts of data rather than programmed rule by rule. Voice assistants, recommendation engines and large language models are products of that shift.
Where AI already shapes daily life
- Search engines and content feeds that rank information.
- Navigation apps that predict traffic.
- Fraud detection in banking and payments.
- Medical imaging tools that help clinicians spot findings for review.
- Chat-based assistants that draft text and answer questions.
Healthcare is a particularly active area, and commentary such as this piece on youngupstarts explores how AI might influence that field in the years ahead.
The video below offers another perspective on where artificial intelligence may be heading:
Why timelines vary so much
Forecasts depend on assumptions about several open questions. Will scaling up current methods keep producing gains, or will new ideas be needed? How much computing power and energy will be available, and at what cost? Can systems learn reliably from less data, reason about cause and effect, and handle unfamiliar situations safely? And how will regulators, businesses and the public respond? Each answer shifts the estimate by years or decades.
Barriers beyond the technical
Even capable systems need to be trustworthy before they become common in sensitive settings. Self-driving vehicles illustrate the point: impressive demonstrations have not yet translated into universal deployment because edge cases, liability and safety validation are hard problems. Similar questions about accountability, bias, privacy and explainability apply in finance, healthcare and public services.
A realistic expectation
The most likely near-term path is a steady spread of increasingly capable specialised tools rather than a sudden arrival of human-like machines. Whether general intelligence follows in a few decades or much later remains uncertain. What is clear is that AI is already woven into society, and the decisions made now about how it is built and governed will shape whatever comes next.
Tell us what you think.
Corrections are always welcome.