Artificial intelligence (AI) means developing intelligence in computing systems and robots. Artificial intelligence emulates human brain intelligence and is capable of executing tasks which require cognitive ability. The most common cognitive services which can be implemented in AI are the following:
- Language, including Natural Language Processing (NLP) and Natural Language Understanding (NLU)
- Computer vision, including image recognition in static images and videos
- Speech processing, including the conversion of speech to text (STT) and of text to speech (TTS)
- Decision making, such as Anomaly Detection, Content Moderation and Personalization.
- Content generation, including all AI models falling under the "Generative AI" category. This is implemented by the so-called Large Language Models (LLM).
- Information retrieval and search
AI is strongly based on machine learning (ML) and data science techniques. Machine learning, put simply, is the process of feeding specialized algorithms with large enough sets of data (data sets), so that machines can continuously learn from this data and be able to carry out autonomous actions, such as predicting, forecasting and making decisions, without human intervention. Algorithms are implemented in a computer programming language. One of the most widely used languages for machine learning and AI is Python.
An AI model is an algorithm which has been trained with a concrete data set. AI models fall under the following general categories:
- Supervised learning models, including regression and classification algorithms.
- Unsupervised learning models, including clustering algorithms.
- Semi-supervised models, combining methods from both supervised and unsupervised models.
- Self-supervised models, achieving self training on unlabelled data.
- Reinforcement learning models, using algorithmic agents which are based on a system of reward and punishment.
It must be noted that, besides the above model/algorithm types, Artificial Neural Networks (ANN) are capable of solving all types of AI problems. ANNs are the most complex AI models.
Artificial intelligence engineering processes are broken down to the following phases:
- Collect data
- Clean and prepare data
- Preprocess data
- Train models
- Evaluate (test) and Tune models
- Deploy models. The deployment of models is orchestrated by the utilization of so called MLOps (ML operations), which are procedures to automate the ML pipelines. Security and monitoring considerations are made in this phase.
There are generally three AI generations, in terms of historic evolution and AI model capabilities:
- Artificial Narrow Intelligence (ANI) – “weak”
- Artificial General intelligence (AGI) – “strong”
- Artificial Super Intelligence (ASI) - "super strong" - exceeds human intelligence
While AI technologies advance, there are various ethical, legal, security and privacy considerations which must be taken into account. These considerations are commonly referred to as "Responsible AI".
