Applications

Dr Chris Paton

Learning Objectives

Trace AI's deployment trajectory in clinical medicine, from FDA-cleared imaging tools to large language models for clinical reasoning, and the regulatory frameworks that govern them
Explain how protein structure prediction was transformed by AlphaFold 2 and 3, and how generative tools like RFDiffusion and ESM-2 now drive de novo protein design
Compare the perception–prediction–planning–control architectures of Tesla's vision-only FSD with Waymo's lidar-camera-radar stack, and use the disengagement rate as a yardstick
Describe the rise of vision-language-action models (RT-1, RT-2, OpenVLA, $\pi_0$) and the role of diffusion policy and sim-to-real transfer in modern robotics
Survey AI's reach across weather, materials, code generation, search, finance, education, creative work, scientific discovery, customer service, law and national security
Reason about the labour-market and societal effects of capable AI systems, and develop a clinician's view of when augmentation is sufficient and when replacement is unsafe

The previous sixteen chapters built up the machinery of modern artificial intelligence: linear algebra, probability, neural networks, transformers, generative models, reinforcement learning, alignment. This final chapter asks where that machinery has actually landed. Which industries have absorbed AI into their core operations? Where has it produced measurable wins? Where has it stalled, embarrassed itself or caused harm?

The answer in 2026 is uneven. Some applications (protein structure prediction, weather forecasting, code completion, certain narrow imaging tasks) have moved from academic curiosity to production tooling that practitioners would now find it strange to work without. Others, fully autonomous driving, general-purpose surgical robotics, AI psychotherapy, remain stubbornly limited despite enormous investment. A third category, including image and video generation, has produced both genuine creative tools and a flood of low-quality content that has degraded its host platforms.

This chapter reviews seventeen application domains. The treatment is not exhaustive; each domain has its own thousand-page literature. The aim is to give the reader enough technical and historical context to evaluate claims they encounter, to recognise mature applications from speculative ones, and to understand the regulatory and economic forces that shape deployment. The author is a practising clinician and the medical sections accordingly receive the most detailed treatment, with deliberate attention to the gap between published model performance and what actually changes care.

Textbook of AI

Applications

In this chapter