Understanding Artificial Intelligence: Of Minds and Machines

Friendly assistant or cold, rational killer? Science fiction reflects our ambivalence about artificial intelligence. Is it a good thing or bad? In this lecture, discover how AI already pervades daily life-from navigation apps to investing to weather forecasting. Trace its origins to Ada Lovelace, who worked with Charles Babbage in the early 1800s and foresaw the power of machine computation.

What exactly is AI? That depends on what we mean by intelligence. Compare rigid instinctual behavior with flexible, goal-directed problem solving in humans and animals. Consider theories of human intelligence-from multiple intelligences to the g-factor and the role of consciousness. See what genuine cognitive flexibility would require, and whether machines could ever achieve it.

From Aristotle's logic to the 1956 Dartmouth Conference, the conceptual foundations of AI grew for centuries before the field finally took off. Explore the two rival traditions that emerged at Dartmouth: "Good Old-Fashioned AI," seeking to replicate reasoning through logic and rules; and neural networks, inspired by the brain's own architecture. Both approaches hit an impasse by the mid-1970s.

Descending in the 1970s, the long winter of artificial intelligence was eventually followed by the AI spring now blooming around us. Discover how neural networks-once considered dead ends-learned how to learn thanks to big data, limitless computational power, and new training methods. The outcome is today's deep-learning systems, which brought both astonishing promise and disquieting opacity.

Can machines truly think or only simulate thought? Explore the limits of current AI: black-box opacity, where even designers can't see how results arise, and overfitting, when systems learn their training data too well. Then, confront the philosophical puzzles of creativity, understanding, and mind through the "Chinese Room" thought experiment and the "hard problem" of consciousness.

AI has amazed even its creators with two breakthroughs. Reinforcement learning has mastered games of strategy far beyond human skill, and ChatGPT-style systems have transformed the human-machine interaction via language. At heart, ChatGPT is a predictive program that associates a written input with a next word or phrase, drawing on a database of billions of web pages-and sounding eerily human.

Explore the mixed blessing of relying on the internet's aggregate wisdom for information, advice, and the seeming intelligence of AI. Surprising accuracy is possible when drawing conclusions from many independent sources, as in the classic "guess the ox's weight" experiment. But when independence breaks down, the same approach can drive misinformation cascades and echo-chamber polarization.

Follow Darwin's idea of evolution by natural selection from biology into code. See how what works in nature also drives optimization in AI, where evolution-inspired algorithms automatically craft solutions to complex problems through variation and selection. Then, trace the evolution of the eye and its striking parallel in computer vision, where layered neural nets learn to see much like we do.

Look at Isaac Asimov's Three Laws of Robotics from his classic sci-fi stories. Then, see how his fictional scenarios anticipated our own ethical dilemmas with AI, where algorithms make decisions that affect real lives. From biased parole software to self-driving cars forced into life-or-death trade-offs, today's systems already test our moral frameworks. Should AI be regulated? If so, how?

As powerful as AI is, we've dealt with transformative technologies before. Think of Gutenberg's press, nuclear energy, or the genomic revolution-each bringing progress and peril. What do they teach us about managing AI? Explore its promise in medicine, science, and education, and its risks, including the phenomenon of AI increasingly feeding on its own output.

Arnold Schwarzenegger's cyborg assassin in The Terminator embodied the fear behind the "Singularity"-the moment when machines outthink and outmaneuver us. That idea now shapes serious debate over AI's accelerating intelligence. Explore visions of a future where self-improving systems could render humanity irrelevant or even extinct, and the counterarguments that such fears are exaggerated.

Look beyond the doomsday scenarios to consider what we might realistically expect from AI. On the positive side, advances in medicine, scientific discovery, and assistive technologies are already taking shape. On the negative side, we face rising waves of AI-generated junk content and escalating attacks on our privacy and the security of our digital infrastructure.