Alvinn, Chess and Google

Plus The Matrix and Terminator, all from the Decade of the 90s

Related: History of AI | AI Research

The 1990s were a transitional decade for AI. After the Cold War ended, funding shifted, and the field pivoted from earlier hype to practical wins and new ideas. The 1990s followed the AI Winter of the late 1970s and early1980s, where overhype led to funding cuts. By 1990, AI was rebounding with fewer grand promises--such as general intelligence--and more about solving real problems with advanced technology. The ending of the Cold War in 1989 shifted priorities, but the internet's rise thanks to DARPA and faster computers powered a new AI wave.

This decade essentially set the stage for the dramatic AI advances of the 2000s and 2010s. This was accomplished with better hardware and algorithms, and by establishing foundational techniques, proving concepts, and keeping AI research alive. We highlight three AI success stories from the decade of the 90s; Alvinn, Chess, and Google.

 

 

The Story of ALVINN

ALVINN, not the sassy chipmunk, but rather the acronym for Autonomous Land Vehicle In a Neural Network

ALVINN is an early example of an autonomous vehicle system. It was funded by DARPA and developed by researchers at Carnegie Mellon University in the late 1980s and early 1990s. It was one of the first attempts to use neural networks for autonomous driving. It laid the groundwork for many of the advancements in self-driving car technology we see today.

ALVINN is considered a milestone in the history of autonomous vehicles and AI. It showcased the feasibility of using machine learning for real-world robotics tasks and paved the way for the development of more sophisticated autonomous systems. Today, its principles are reflected in the deep learning and computer vision technologies used by companies like Tesla, Waymo, and others in the autonomous vehicle industry.

Key Features

• Neural Network-Based Control: ALVINN used a neural network to process input from a camera mounted on the vehicle and generate steering commands. The network was trained to map visual input (road images) to appropriate steering angles.
• Training Process: The system was trained using supervised learning. A human driver would operate the vehicle while the system recorded camera images and corresponding steering angles. The neural network learned to mimic the human driver's behavior by associating visual patterns with steering actions.
• Real-Time Operation: Once trained, ALVINN could operate in real time, processing video input and making steering decisions autonomously. It demonstrated the ability to navigate on roads, avoid obstacles, and stay within lanes.

 

Limitations

ALVINN was limited by the technology of its time. The neural network was relatively simple compared to modern deep learning models. It struggled with complex or unfamiliar environments, as it relied heavily on the training data it was exposed to.

Impact

ALVINN was a pioneering project that demonstrated the potential of neural networks for autonomous driving. It inspired further research and development in the field, leading to more advanced systems like those used in modern self-driving cars.

 

Chess

A watershed moment for AI occurred in 1997 when IBM's Deep Blue defeated chess grandmaster Garry Kasparov

AI has a rich history in chess, marked by significant milestones that showcase the evolution of AI in the game. AI's journey in chess has not only transformed how the game is played, but has also influenced broader developments in AI and machine learning. The interplay between human players and AI continues to evolve, with AI serving as both an opponent and a tool for enhancing player skills. Learn more.

 

Google

Google changed search, and over time AI changed Google

Google's history in the 1990s is a fascinating story of innovation, ambition, and rapid growth. The 1990s laid the foundation for Google's dominance in the tech industry. By the end of the decade, Google was well on its way to becoming the world's most popular search engine, setting the stage for its expansion into other areas in the 2000s.

Some History

1995: The Meeting of Larry Page and Sergey Brin
Larry Page and Sergey Brin, both graduate students at Stanford University, met in 1995. Larry was considering Stanford for his PhD, and Sergey was assigned to show him around. They bonded over their shared interest in organizing and retrieving information from large datasets.

1996: The Birth of "Backrub"
Larry Page began working on a research project called Backrub, which aimed to analyze the relationships between websites by examining backlinks (links from one site to another). Larry and Sergey developed the PageRank algorithm, which ranked web pages based on the number and quality of links pointing to them. This was a revolutionary approach to search, as it prioritized relevance and quality over simple keyword matching.

1997: The Name "Google"
Backrub was renamed Google, a play on the word "googol" (a mathematical term for the number 1 followed by 100 zeros). The name reflected their mission to organize the vast amount of information on the web. The domain google.com was registered on September 15, 1997.

1998: The Founding of Google
Google was officially founded on September 4, 1998, in a garage in Menlo Park, California. The garage was rented from Susan Wojcicki, formerly the CEO of YouTube. Larry and Sergey secured an initial investment of $100,000 from Andy Bechtolsheim, co-founder of Sun Microsystems. This allowed them to set up their first office and expand their operations.

1999: Rapid Growth and Relocation
In early 1999, Google moved to an office in Palo Alto, California. Google's homepage was famously simple, with a clean design and a focus on delivering fast, accurate search results. In June 1999, Google raised $25 million in venture capital funding from firms like Sequoia Capital and Kleiner Perkins. This funding helped Google scale its infrastructure and improve its search technology.

PageRank Algorithm

PageRank was the foundation of Google's search engine, which set it apart from competitors. The PageRank algorithm, developed by Page and Brin in the late 1990s, is often considered an early form of AI machine learning, or at least a precursor to modern machine learning techniques. While it wasn't explicitly called "machine learning" at the time, it shared several key characteristics with machine learning approaches.

PageRank used the structure of the web (specifically, the links between web pages) as its input data. Each link from one page to another was treated as a 'vote' or endorsement of the linked page's importance. The algorithm produced a ranking of web pages based on their perceived importance or relevance. This process of taking raw data (links) and transforming it into meaningful outputs (rankings) is a fundamental aspect of machine learning.

In case your wondering, PageRank was named after Larry Page. The name is a double entendre, for it refers not only to Larry Page, who developed the algorithm along with Sergey Brin, but it also refers to the algorithm's purpose of ranking web pages. The name reflects both the personal connection to Larry Page and the algorithm's function. It's a clever and fitting tribute to one of the key figures behind Google's founding technology!

By the end of the 1990s, Google had already begun to change how people accessed and thought about information on the internet. Its emphasis on relevance and simplicity made it a standout in the crowded search engine market.

 

Links

g2.com/articles/history-of-artificial-intelligence

techbullion.com/artificial-intelligence-in-the-90s-a-decade-of-innovation-and-challenges

en.wikipedia.org/wiki/Applications_of_artificial_intelligence

ai-bees.io/post/artificial-intelligences-early-history-and-future

klondike.ai/en/ai-history-the-innovations-of-the-90s-and-deep-blue

tableau.com/data-insights/ai/history

calmu.edu/news/future-of-artificial-intelligence

linkedin.com/pulse/evolution-artificial-intelligence-ai-joseph-n-martinez