Knowing the right terminology is the foundation of effective communication in artificial intelligence. Many of these terms remain in English regardless of the language you speak, so understanding their meaning and their correct pronunciation becomes a critical skill for anyone working in or studying AI.
How are AI systems classified?
AI systems can be grouped into two main categories based on their capabilities. General AI, also known as strong AI [0:38], refers to an AI system capable of performing any task a human being can do. The expression also known as even has its own acronym: AKA. A famous example of general AI is HAL 9000 from the movie 2001: A Space Odyssey, a system capable of following complex instructions, understanding subjective concepts, and making its own decisions [1:25].
On the other side, you find narrow AI, AKA weak AI [1:00]. This type of AI is designed for a specific and limited set of tasks. ChatGPT is a good example: it can handle text-based tasks like editing a letter or a piece of code, but it cannot process images or video on its own [1:08].
What does autonomous mean in AI?
An autonomous system can perform a given set of tasks without human intervention [1:45]. Pay special attention to the pronunciation: it is autonomous, not autonomous with a different stress. A clear example is self-driving cars, vehicles that drive themselves without a human at the wheel [1:58].
Why is bias important to understand?
Bias describes a situation where a system makes a prejudiced decision based on underlying assumptions [2:10]. The correct spelling is B-I-A-S and it is pronounced bias. Consider this scenario: if you feed an AI a collection of pictures of construction workers and most of them show male workers, the system will become biased toward showing only male construction workers — even though women can absolutely be construction workers too [2:25].
What is the difference between dataset and dataframe?
A dataset is a collection of related data points arranged in some organized way [2:48]. You can write it as one word or two separate words: dataset or data set.
A dataframe, on the other hand, is a table structure with rows and columns where data are organized [3:05]. It can also be written as one or two words. The distinction matters: a dataframe always implies a tabular format, while a dataset is a broader concept.
How should you use and pronounce the word data?
Data is technically the plural form of datum [3:22]. However, in everyday situations it is acceptable to use it as either plural or singular.
- Plural: "The data show a clear trend."
- Singular: "The data shows a clear trend."
Both pronunciations — data and data — are correct [3:38].
What are parameters and hyperparameters?
Parameters are variables or numbers set by the algorithm that the model uses to make predictions [3:50]. The word algorithm itself is another essential term in AI vocabulary.
Hyperparameters, in contrast, are set manually by the person training the model. They determine which parameters the model will use during the learning process [4:02]. The correct pronunciation is parameter and hyperparameter — not paramater or hyperparamater.
An important related concept is hyperparameter tuning [4:18]. This involves adjusting the model's hyperparameters until you achieve the desired output. Essentially, you are playing around with the numbers to optimize the model's performance.
Mastering these terms not only helps you read documentation and research papers more effectively, but also ensures that colleagues and collaborators understand you clearly. Share in the comments any other AI terms you think are essential to know.
