Artificial Intelligence: From Symbolic Rules to Deep Learning
Computing and artificial intelligence. Wikimedia Commons.
Free levelled English reading practice. Read for meaning first, then use the full Speed Reading tool to time yourself and track your progress.
Artificial intelligence has repeatedly shifted between ambitious expectations and periods of disappointment. Early researchers often pursued symbolic systems that represented knowledge explicitly through rules and logical structures. Other approaches developed alongside them, including statistical pattern recognition and neural networks. The recent success of deep learning reflects several converging conditions: larger datasets, powerful specialised hardware, improved algorithms and the ability to train neural networks with many layers. Generative models can now produce fluent text, images, audio and code, but impressive output does not eliminate problems of factual error, bias, opacity, copyright, labour disruption or misuse. The central policy challenge is therefore not merely whether machines can perform particular tasks, but how institutions should govern systems whose capabilities diffuse rapidly through society.
