In this blog, I plan to write about the current landscape of AI in medicine and future exciting developments. But before I start to discuss that topic, I first want to briefly describe the history of AI.
Artificial Intelligence has been around for a very long time. The best infographic I found online was from Dr. Paul Marsden.
The Birth of an Idea
The academic discipline of AI began in the 1950s. Alan Turing published a landmark paper in 1950, speculating about creating machines that “think.” The premise of the Turing Test was that a computer was truly intelligent if a normal human could not design questions that could differentiate between a computer response and a human one.
The Dartmouth Conference in 1956 asserted the name, goal, and eventual mission of artificial intelligence – “every aspect of learning or any other feature of intelligence can be so precisely described that a machine can be made to simulate it.”
The First Hype Cycle (1950s-1970s)
Money and excitement poured into AI in the 1950s to 1970s. At the same time, there were significant advancements in algorithm design, natural language processing, robotics, and machine vision. Some cool exciting developments include UNIMATE (the first robot that replaced humans in the assembly line), ELIZA (the first realistic chatbot), and STUDENT (a program that solved high school algebra word problems).
First AI Winter (1970s)
In the first hype cycle, there were bold predictions that computers could soon defeat human chess champions, solve unsolved mathematical problems, and eventually have general intelligence comparable to a human. Critiques and financial setbacks dampened this initial set of optimism due to limited computational power, intractability of solving problems in exponential time, and the requirement of massive amounts of data required. This era was labeled the first AI winter, lasting from the 1970s to 1980s.
Brief Resurgence and Second AI Winter (1980s)
A brief resurgence in “expert systems” in the 1980s created additional excitement about AI. Expert systems are computers that mimic the decision making abilities of humans through a series of if-then statements. Picture the ATM machines at your local Bank of America. This created industrial hardware companies like Symbolics and Lisp Machines, who were designed to support these systems. During this time, key developments in Hopfield neural networks and backpropagation were formulated. Money poured in, starting with Japan, United Kingdom, and DARPA. However, the perception of AI by investors soon fell. Expert systems were too difficult to update and maintain.
Rebirth of AI (1990s- Now)
Why is AI regaining popularity now? In short, the field is finally achieving some of its old goals. New formulations in Bayesian probability, Markov models, information theory, stochastic modeling, and optimization brought a wealth of new tricks to the trade. Advancements in microprocessors, computers, data storage, and data transfer have addressed earlier concerns about speed, computational cost, and information requirements.
In 1997, Deep Blue defeated Garry Kasparov, the reigning world champion. In 2002, Roomba can autonomously navigate and clean homes. In 2005, a Stanford Robot was able to drive autonomously for 131 miles along an unrehearsed desert trail. In 2011, IBM Watson won the $1M prize for winning in Jeopardy. In 2014, Eugene Gootsman defeated the Turing test by mimicking a teenage Ukranian boy (I’m not sure if this is what Alan Turing really had in mind). In 2017, Google’s Alphago defeated Lee Sedol and Ke Jie in Go, a game that has more possibilities than the number of atoms in the universe.
AI is now the buzz in every industry. Andrew Ng describes AI as the New Electricity of our lifetime.
I agree 100% with Dr. Ng. In my upcoming posts, I will describe the potential of AI in medicine. I believe that AI combined with advancements in genetics will completely transform the field of medicine. The best place to start is a detailed review of Eric Topol’s seminal paper: