Page 7 - AI Vol 1: Foundations of AI

marking the start of AI as a legitimate area of effective in tasks like image and speech
scientific inquiry. recognition, leading to practical applications such

as facial recognition software and voice-activated
In the 1970s, AI researchers achieved some assistants. Hardware continued to improve,
notable successes, particularly in the realm of giving researchers increasingly easy access to
rule-based systems, known as expert systems, the computing power needed to train and deploy
designed to mimic the decision-making process these more advanced models. The Internet also
of human experts in specific domains, such as played a critical role in allowing collaboration
medical diagnosis or geological exploration. and the creation and sharing of open-source tools

One of the most famous examples is MYCIN, that proved essential for further development.
developed at Stanford, which could diagnose
bacterial infections and recommend antibiotics. Through the 2010s, the pace of advancement
These systems demonstrated that AI could provide and key breakthroughs in AI continued to
practical solutions in narrow, well-defined areas, accelerate. One of the more notable milestones
boosting confidence in the field. However, was the success of AlphaGo, an AI developed by
these systems were expensive to develop and DeepMind, which in 2016 defeated one of the
maintain, were difficult to scale, and could not world’s top players of the ancient board game
handle complex reasoning beyond their narrow “Go” which is renowned for its deep strategic

domains of expertise. These issues, along with complexity. Unlike chess, where brute-force
the hardware limitations of the time, tampered calculation is often effective, Go requires a form
interest in AI. Although there were brief periods of intuitive understanding due to the vast number
of renewed interest in AI over the next several of possible positions. This complexity made Go
decades, AI research largely languished.

The late 1990s and 2000s marked a turning point

in the history of AI, driven by the explosion of
data and advancements in computing power. This
era witnessed a shift from rule-based systems
to machine learning, where instead of being
programmed with explicit rules, algorithms
could “learn” patterns from data, allowing the AI
systems to tackle more complex, less structured
problems, from image recognition to natural
language processing. The advent of the Internet

led to unprecedented availability of data, allowing
AI to be trained on large and diverse data sets.
During this period, deep learning, which uses
neural networks many layers “deep” began to
gain prominence, which proved particularly

FOUNDATIONS OF AI | LOZANOSMITH.COM VOLUME 1 | 7