The latest on AI. What does AI do? Its main applications
AI’s current state is comparable to the early phases of human evolution. Despite the incredible technological advances that artificial intelligence has undergone in recent years, in reality, it is still at a somewhat preliminary stage, characterised by its rudimental and inaccurate emulation of the human mind.
The ambitious aim of creating autonomous thinking machines still remains a Utopia then, although there has been plenty of progress over the last 40 years, from the building of ever more complex neural networks to the development of sentient nanotechnologies.
Through the implementation of innovative algorithms, it has been possible to give life to technologies such as cognitive computing, with machines able to emulate the thought of a brilliant student. What is still lacking however is that spark of genius: that inspiration typical of certain human beings, that is the source of ideas and intuition.
Weak AI and strong AI: the differences
This distinction between different forms of AI makes it easier to understand the concept mentioned in the previous paragraph. Weak AI is the kind of artificial intelligence (also known as augmented) which can be achieved using current technologies.
In reality, it involves complex programs, which are able to emulate human thought, based on the elaboration of a huge amount of data without any actual intelligence. It essentially consists of the emulation of the human cognitive processes used in problem solving.
In order to develop what is known as strong AI, a significant additional step is required: training AI how to understand the meaning of language. Based on this, it will be possible to teach AI deductive or inductive reasoning, thereby providing it with the necessary instruments to achieve creative thought.
The challenge of achieving this target remains complex and limited by technology, but it is not impossible. One day, forms of real artificial intelligence like those described in science fiction will exist and could give rise to a series of ethical issues (not discussed here).
Machine Learning and Deep Learning: How AI is trained
The challenge of teaching a machine how to think has fascinated researchers since the early twentieth century. Machine learning is that combination of algorithms which allows AI to learn information via the analysis of data and pre-determined models, leading to increasingly accurate solutions when solving problems.
There are many different methodologies which, as they have gradually evolved, have enabled the calculation of increasingly complex and articulated results (how can we forget the example of DeepBlue, the first computer capable of beating a legendary chess champion).
Deep learning may be considered the most recent evolution in AI ‘thought’ processes. Based on artificial neural network technology, it allows for the simulation of human brain neurones using binary language.
This kind of AI, which is considerably more complex compared to that based on machine learning, is able to provide even more precise results, giving very accurate predictions. In this case, it is necessary to train AI using specific simulated examples, so that it can ‘learn’ to exploit them in order to generate accurate solutions to problems of a similar type.
What is AI used for today?
Although AI is still not yet intelligent enough to be able to replace humans, it is widely used in many sectors, providing valuable assistance. It is to be found in all environments requiring the analysis of huge quantities of data:
- It is fundamental, for example, in marketing, for the generation of forecasts and the analysis of different target groups, providing firms with data which is extremely useful for the development of their commercial strategies;
- It is used in the medical field to recognise and catalogue new pathologies, comparing data which has not yet been correlated;
- It is used to present products and services interactively (also using virtual and augmented reality).
Translated by Joanne Beckwith
