Over the last 100 years, previously unimaginable inventions came to life, such as computers and the Internet, but the most astonishing and promising is definitely Artificial Intelligence. Today, the world AI market is worth over $10 billion, and quickly growing – but how did it all begin?
Let’s take a trip back to the past and learn about the evolution of AI.
The first neural networks
In the 1940s, developments in the field of neurology, specifically finding out that our brains are electrical networks, gave rise to the idea of constructing an electronic brain – a concept that would later evolve into what we know today as Artificial Intelligence.
By 1943, researchers Walter Pitts and Warren McCulloch described the first idea of a neural network – that is, artificial neurons performing logical functions. One of their students, a young Marvin Minsky, built the first neural net machine in 1951 called the SNARC, or the Stochastic Neural Analog Reinforcement Calculator.
In 1956, the term Artificial Intelligence was officially adopted, and government agencies started pouring a lot of money into the research. Projects were booming in the 1960s, but problems started to emerge in the early 1970s.
Difficulties in evolving AI
Financial setbacks halted AI development in the 1970s, as not many promising results emerged in the past decade. Researchers struggled with limited processing power and difficult access to databases necessary for creating advanced neural networks, and research agencies withheld their funding.
The 1980s, however, saw another boom of AI-related research, and the first AI products that were actually useful began appearing, including the first expert systems, or devices that solved problems related to a specific discipline using a set of rules, derived from expert knowledge.
Modern-day AI: big data and deep learning
Nowadays, companies have easy access to huge amounts of data, also known as big data, as well as cheaper and faster devices, performing many times better than the technology from over 30 years ago. Modern methods of evaluating algorithms have emerged, like the FRVT, or the Face Recognition Vendor Test, employed by the National Institute of Standards and Technology, or the NIST. More information can be found here.
AI is moving towards general intelligence, being able to solve any problem, not just a select few problems related to a single field. Many commercial products already use AI-driven technology, which is evolving faster than ever, with new models releasing frequently like the GPT-3 by OpenAI or Gato by DeepMind.
