Skip to Content
This is the Beta version of our new Learning Paths approach. Please email feedback.

AI/ML Algorithms

AI/ML algorithms use a combination of logic and mathematics to provide a complex set of instructions that can be programmed for any computing device to process. Their objective is usually to take inputs from the real world and process that data and take decisions and actions as humans would. The more efficient the algorithm the closer it will be to matching human decisions.

While in many cases human understanding of data are consistent (for example, humans will always read or hear words as they are written or spoken), in some cases their interpretation may vary (the same set of words may mean different things to different people depending on their understanding of the language and the context). Their decisions may also vary based on the context and their experiences in similar situations.

With AI/ML algorithms, the interpretations and decisions will be consistent based on how the instructions have been programmed and may only change if more past data and the ML component suggests a change to the interpretation and decision.

For example, if you walk into a coffee shop and ask the person at the counter a simple question: “I would like a coffee please”, the response from the human may be different. If the barista knows you and your preferences he or she will know exactly what to serve you. If you are a new customer, you will be asked for your preferences. If the barista is disinterested or has not understood the task, he or she may just serve any coffee that they believe you want.

A computing device programmed with an AI/ML algorithm will always seek all the information it needs before it can process the request. In some cases this can be less efficient than a human who knows you and doesn’t have to ask every time. Or if it is not programmed properly (similar to a human not being trained properly) it may take incorrect decisions since it has not gathered all the information it should have.

Of course, it is clear that an AI/ML system can be made as sophisticated as required to make it very close to humans. For example, a camera with facial recognition can be placed at the counter to determine whether you are a known customer and if your preferences are already available they need not be asked again.

Systems get more expensive as they get sophisticated and closer to humans, so the decision is based on the criticality of the application.

There are several algorithms, very often in conjunction with each other to provide powerful solutions with AI/ML capabilities. Some of the primary algorithms are described below.

  • Natural Language Processing (NLP): Algorithms that interpret words and sentences and understand their meaning, including their contextual meaning. Solutions built using these algorithms are categorized as Conversational AI solutions which allows humans to interact with them as if they were talking to a person and get answers and information.

  • Speech Recognition: Algorithms that can detect and understand human speech. This is often used in conjunction with Natural Language Processing to take decisions and actions on the spoken words.

  • Text to Speech: Algorithms that can convert written text into a spoken word using human voices. This is the reverse of speech recognition.

  • Text Analysis: Algorithms that can process huge volumes of multi-lingual text and derive required insights and relationships within the texts.

  • Image Recognition: Algorithms that can process images and identify relevant parts or aspects of the image that are relevant to the context.

  • Video Analytics: Algorithms that can process video feed, either pre-recorded or real-time, and identify relevant parts or aspects of the video that are relevant to the context.

  • Data Analytics: Algorithms that can process and analyze vast amounts of data (Big Data) and provides insights to humans. Algorithms are generally domain-specific, for example they will analyze consumer data and provide sales marketing insights, or analyze financial data and provide investment related insights, or analyze machine performance data and provide predictive maintenance alerts, and so on.

  • Personalization: Algorithms that can process individual behaviour patterns and provide recommendations that are relevant to the individual based on the patterns of past behavior. These algorithms are also generally domain specific, for example they will analyze the music you listen to or videos you watch and recommend new music or videos.

  • Generative AI: The newest kid on the block, Generative AI uses models trained using vast amounts of data to produce new text, images, videos, or other forms of data. These models learn the underlying patterns and structures of their training data and use them to produce new data based on the input, which often comes in the form of natural language prompts.

💡

We will be launching a new Learning Path that helps you understand these algorithms in detail and enables you to use these algorithms in your solutions.