Artificial Intelligence (AI)

Artificial intelligence (AI) is the science and engineering of building machines that perform tasks associated with human intelligence — perception, reasoning, language, learning, planning, decision-making under uncertainty. The term was coined at the 1956 Dartmouth Workshop and has covered vastly different technical approaches across its history. In the 1960s-80s, AI predominantly meant symbolic reasoning and expert systems; in the 1990s-2000s, it came to mean statistical pattern recognition; in the 2010s, deep learning; in the 2020s, large language models and agent systems built on top of them.

The modern field is structurally an umbrella. **Machine learning** provides the statistical foundation. **Deep learning** provides the neural-network engineering. **Natural language processing**, **computer vision**, **speech**, and **robotics** are application subfields. **Reinforcement learning** handles sequential decision-making. **AI safety and alignment** address the societal consequences of increasingly capable systems. The boundary between "AI" and "not AI" has always been porous — every technique that works consistently tends to lose the AI label and become just software — so the useful question is rarely whether something is AI, but whether it is the right tool for the problem.

For APAC mid-market enterprises, AI in 2026 is a commodity input to software, not a novel capability. The strategic questions have moved up the stack: which vendor, which deployment model, what governance regime, what training for staff. The common failure mode is treating AI as a destination ("we will become an AI company") rather than as infrastructure ("these workloads get AI, these do not, these are ambiguous and deserve further analysis"). The mature posture is selective — AI is deployed where it produces measurable value, skipped where it does not, and always accompanied by the operational disciplines that any software capability requires.

The non-obvious observation: **the AI field's pattern of hype cycles repeats**. 1970s expert systems, 1980s neural networks, 1990s decision trees, 2010s deep learning, 2020s LLMs — each cycle over-promised near the peak, produced some durable infrastructure, and left a residue of capability that became quietly load-bearing after the hype faded. The 2022+ generative-AI cycle will follow the same arc. The durable technology will still be here in ten years, integrated into everything; the claims at the peak will look overstated in retrospect. Build for the durable part.