In recent years, computer science, artificial intelligence and data science have become popular topics. Many claims are made about these areas, which are not always correct. This even holds for academia. The purpose of this discussion page is to provide some clarity.
How are the different areas defined?
- Computer science is the study of the data and algorithms.
- Artificial Intelligence is the study of making representations (definition by Luc Steels), of course for creating smart solutions.
- Data science is the study of extracting usuful knowledge and insights from data.
Computer Science
Unlike the word “computer science” suggests, computer science is not about computers. The focus of computer science is on algorithms that can process data. These algorithms are implemented in programs using some programming language. A fundamental principle of computer science is that the choice of the programming language is irrelevant except of pragmatic reasons.
- Every programming language with sufficient expressiveness is equivalent to a Turing-machine.
- This principle implies two other principles:
- There are problems that cannot be solved with any computer program.
- The computational complexity of an algorithm can be addressed independent of the actual implementation in a program.
Another fundamental principle of computer science worth mentioning is:
- We can describe the world is a digital way.
Artificial intelligence
Artificial intelligence (AI) aims at making smart systems. In other words, AI aims at developing algorithms that gather knowledge about the world and use this knowledge to act or advise. A fundamental question is always: how can we represent the knowledge about the world (the study of making representations). This problem is addressed using tools from computer science, mathematics, statistics and logic. Although AI has also contributed to these fields, it is not an aim of AI research itself.
A deep neural network is currently a popular representations studied in AI. It owes its popularity to the availability of AI tools that can be used by non-experts, and to some impressive success stories. A deep neural network is not the only successful representation developed within AI. Other types of representations have successfully been developed for: machine learning, (route) planning, scheduling, diagnosis, legal advise systems, the semantic web, etc.
A major disadvantage of representations, such as a deep neural network, is their explainability. They are essentially black-boxes that make decisions and recommendation without motivation. Black-box machine-learning techniques are unacceptable when decisions may have a profound impact on peoples’ lives. Explainable AI aims at addressing this issue.
Data science
Data science aims at extracting knowledge and insights from data. The knowledge and insights should enable us to improve performance and make better decisions. For instance, will medicine X be effective in a patient with profile Y. Data science is multi-disciplinary. It uses tools from different fields, such as: computer science, statistics, artificial intelligence (machine learning), semantic web technology and data visualization, to extract the knowledge and insights that enable us to improve performance and make better decisions.
What is the difference between data science and artificial intelligence?
Artificial intelligence focusses on making representations that can be used to address specific types of problems. This has resulted in tools that can be used address practical problems. Data science focusses extracting knowledge and insights from data. Data science is an engineering discipline that uses tools from developed in AI and other fields for extracting knowledge and insights from data.
Are you doing AI research when you apply AI tools?
Regularly, researchers not working in AI, claim that they also do AI research because they apply AI tools in their research. These claims are unjustified if the aim is not to study representations. When applying statistical tools, you do not claim that you do research in statistics.