Cloaked in a halo of mythification, the product of literary and fanciful interpretations, artificial intelligence is a combination of computing technologies, data storage and algorithm programming. It is naïve to expect from an AI system imaginative responses, improvisation in unexpected situations, or decisions based on feelings, emotions and desires, as one might expect from a human intelligence. However, it can be expected to respond to sensations: if it receives a low temperature input, compared to the target temperature, it will react by issuing a command to turn on the heating, if this is foreseen in its catalogue of responses.
This is the immense field of AI: analysing the data it receives and comparing it with the data it already has, in order to establish the predefined response that best fits the situation. An activity without spatial barriers, because it can interact with the user via a smartphone.
From this simplistic explanation, capabilities progress geometrically as computational power, the depth of the data catalogues AI has access to, the complexity of the algorithms that interpret the information to infer its answers, and the variety, quality and sophistication of the channels and devices for capturing data.
The availability of massive amounts of data and the development of algorithms capable of detecting in them certain patterns of behaviour to 'predict' future behaviour, associated to a specific block of information (let's say, the information of a subject, compared to the patterns of a large number of subjects), attribute to AI a learning dimension, called Machine Learning, which allows it to infer probable outcomes regarding events that have not yet occurred.
The ability of computer vision, analysing images to distinguish objects from each other, adds another dimension to the AI's inference capability. It can, for example, 'read' text from an image, identifying letters, just as it can recognise a person.
The complexity of this type of data requires greater computational capacity to identify objects that are similar to others or seen from different perspectives. This is achieved by applying a strategy of neural networks, which are structured in a series of hierarchical levels, for a technique known as Deep Learning.
This layered structure serves as a sieve for the system to filter the information from a basic level to a succession of combinations of what each level detects, creating a more complex data pattern.
Artificial intelligence interprets sound, recognising spoken words and learns to understand, with knowledge from the written word, natural language, the structure of semantic patterns, sentences and meanings.
It does not understand, but interprets the logical meaning of sentences. It is unlikely, for the time being, that an AI will detect subtleties of human speech, such as irony and double meanings. But it can interpret commands, queries or comments, giving coherent answers, also in natural language (in writing, or with a speech synthesiser), from its knowledge base.
The sum of these capabilities, plus the ability to activate mechanisms and actuators to generate action and movement in the real world, paint a sufficiently realistic picture of the limitless scope of applications for artificial intelligence, even in the bare definition of the first paragraph. Complex systems, such as the autonomous vehicle, must be based on the integration of multiple inputs of environmental data, vision, communications with automatic information systems and other vehicles, the ability to identify static and moving objects, and real-time processing, with no margin for error in making decisions such as moving forward, braking, turning and obstacle avoidance.
The accuracy and rigour with which AI can operate gives it a place in healthcare to control high-precision systems, such as those needed for microsurgery, radiological examinations and treatments, symptom-based diagnosis and detailed observation of X-rays, CT scans and other physical tests.
Spain's role
Political paralysis is keeping the draft National Strategy on Artificial Intelligence in limbo, but there are endless possibilities for the development of applications for the use of artificial intelligence. Starting with the creation of trust and security environments, with standards and ethical criteria for its application. The application of intelligence in all kinds of educational, research, industrial and production processes opens the field to developments for local activities, or with global ambitions. In the short term, i.e. now, it will be a decisive factor in the competitiveness of companies.
With regard to the use of natural language, in applications such as personal assistants (such as Siri, Cortana, Alexa... and the Spanish Sherpa, among others) and chatbots (conversational algorithms that proliferate in online customer services), it is obvious that Spain must play a relevant role in developments in a language that is spoken by more than 400 million people. This is also applicable to the many types of assistance and customer service bots.
