The Evolution of Artificial Intelligence 

Professor Pietro Perona, a leading AI expert at Caltech, discussed the evolution of artificial intelligence, focusing on visual recognition. His work in machine learning has led to algorithms that help machines recognize objects like animals and cars with minimal human supervision. He explained how AI has advanced from "AI by Design," where systems were manually programmed, to modern generative AI, which can analyze and generate content based on data.  Perona highlighted key milestones, including the shift to supervised learning in the 1990s, where AI learned from labeled data. By 2012, this led to breakthroughs in fields like dermatology, where AI systems began outperforming human experts. Today, unsupervised learning allows machines to analyze vast amounts of data and identify patterns, even learning from global user contributions.  He gave the example of the "iNaturalist" app, which grew from recognizing 5,000 species to over 100,000, thanks to crowdsourced data. Despite these advancements, AI still struggles with understanding causality and physical-world learning, which requires real-time interaction and is slower than data-driven learning.  Finally, Perona emphasized the importance of creating responsible AI—systems that are ethical, unbiased, and transparent. He concluded by noting that while AI holds great promise, it requires ongoing collaboration between engineers and society to ensure it benefits everyone.