The History of Artificial Intelligence

By nbrand
  • Invention of SNARC

    Invention of SNARC
    The Stochastic Neural Analog Reinforcement Calculator (SNARC) was the first neural net machine, created by Marvin Lee Minsky in the summer of 1951. It used about 40 Hebb synapses that would each hold a probability of a signal. If the signal went through, the synapse would "remember" the signal by engaging a clutch. Then, a large motor would check for engaged clutches. This is a picture of one "neuron" of SNARC.
  • Invention of SHRDLU

    Invention of SHRDLU
    During the late 60s, AI researchers believed that artificial intelligence projects should focus on simple situations known as "micro-worlds". SHRDLU was a robot created by Terry Winograd at MIT. It could communicate in basic English sentences, plan and execute operations. The picture shows the interface between the user and SHRDLU. The user could ask it questions, tell it to do things, and even define concepts for it. It would remember what it did and it could answer questions based on new info.
  • Period: to

    AI Winter

    From 1974 to 1980, AI research faced a number of problems. Computer processing power could not keep up with the demands of AI programs. AI programs also needed exponentially more time to solve problems as they became more complex. AI also needed loads of information that just wasn't possible to store. Eventually, the sources of funding for many of these projects stopped funding because they believed that progress was too slow.
  • Period: to

    AI Boom

    From 1980 to 1987, a new form of AI program began to rise: expert systems. These programs could make decisions for companies; as if they were "experts". One such program, XCON, saved Digital Equipment Corporation $40 million annually. During this time, many countries started funding AI research. Japan set aside $850 million for the Fifth generation computer project. Great Britain and the United States responded with funding programs of their own. AI research picked up steam once more.
  • Deep Blue

    Deep Blue
    Created by IBM, Deep Blue becomes the first AI program to defeat the reigning world champion, Garry Kasparov. It could process 200,000,000 moves per second. This large jump in power was a result of Moore's law, the idea that computer processing speed would double every 18-24 months. Deep Blue was 10 million times more powerful than the Ferranti Mark 1, which was used by Christopher Strachey for chess in 1951.
  • Invention of Duplex

    Invention of Duplex
    In 2018, Google unveiled Duplex, it's new AI assistant. It can carry out conversations with eerie realism, and can adapt to unexpected stimulus. It can make appointments for the user and remember their schedule and preferences. Duplex also managed to pass the Turing test, where judges have to identify if a conversation is with a human or an AI.