What is artificial intelligence – background information
Artificial intelligence (AI) is a fast-growing group of technologies with the potential to deliver a wide variety of socio-economic benefits across all industries and areas of social activity. It is the ability of computer programs to exhibit human skills such as reasoning, learning, planning and creating new content.
How is AI different from a traditional computer program?
A traditional computer program is a set of instructions written in a specific programming language that describes the tasks the program is supposed to perform. The result of its work is therefore predictable and planned from the beginning. Artificial intelligence is based on self-learning algorithms (so-called machine learning), but interestingly enough, the learning process requires some training, which – for the time being – has to be carried out by a human.
How is it possible that an algorithm can mimic the human brain operation?
This all occurs through the so-called neural networks. Interestingly, the theoretical model of an artificial neuron built similarly to a human neuron was created in the 1940s (1943), when American scientists Warren McCulloch and Walter Pitts used their knowledge of human neuron structure, mathematics and computer tools available at the time. The model was simple: the algorithm had one output and one output, and in between these values different weights were given that transformed the output into the final result. However, in order to realise the potential of an artificial neuron, it was necessary to connect multiple neurons and create a network of them. 1956 is considered to be the year when the term artificial intelligence was first used, and this took place at a conference at Dartmouth University.
In 1957 Frank Rosenblatt created the Perceptron network of interconnected neurons and demonstrated that, with this structure, a machine could learn in the likeness of humans and animals. However, the available computing power of machines was too slow for the technology to develop, so rapid development did not occur until the 21st century.
What are neural networks?
Currently, neural networks are a complex structure of thousands of neurons and there are many types of them. What is common is that they consist of individual neurons connected by synapses, with which weights, that is numerical values, are associated, the interpretation of which depends on the model adopted. Neural networks analyze the input data and then the result of this analysis (output) – thanks to the weights (values) applied – is passed on to the next neuron. The calculation and correction processes continue until the final result is determined. The final stage is the comparison of the result with the predefined target. This stage involves trainers who verify the result data and provide feedback to the algorithm, so that the algorithm acquires further data for improvement.
What can we expect in the near future?
The processes that take place in artificial neural networks are not fully understood (the so-called black box effect). As Professor Włodzisław Duch said in an interview, it is not untypical in sciences that it is possible to artificially reproduce processes known in nature, even in situations where this natural process has not been fully explored, e.g., when the first airplane was created it was not known how it happens that birds have the ability to fly.
We cannot predict where the development of artificial intelligence will take us in the coming years. It will certainly change reality in very many areas. Therefore, one cannot remain indifferent towards artificial intelligence. It can be expected that it and their ability to process information quickly will make it particularly difficult to catch up with. However, it still seems to be only a tool in the hand of man, which can bring him many benefits.