Everything that we call AI is NOT real AI
by Pradeepta Mishra, Head of AI at Fosfor
Recently we have seen many products claiming to be AI/ML-based. Even software claim to be artificially intelligent. AI-based systems are software systems that include one or more than one AI component. Building, operating, and maintaining AI-based systems is different from developing and maintaining traditional software systems. AI as per the Wikipedia definition is the intelligence decision-making or intelligent behavior displayed by a software system which is called a machine. The machine is predominantly a computer or any other associated device that contains a computing system for inference and decision making.
So the answer to the basic question of what is AI, or what are the features or attributes that we see in a computer system or machine in order to call it an AI-based system are the following.
Goal: there has to be a goal, target, we can call it a task, the task can be simple classifications, separating one group from rest of the data points, the task can be finding out patterns and similarities in a group of data points that can be clubbed together to understand the group behavior, or it can be simple recommendations based on features, attributes, and behavior
Performance measurement: there has to be a process to measure the progress of the computing system by defining various metrics, baselining the metrics, and closely monitoring the metrics in order to assess whether the computing device or machine is learning from the new data points or not.
Training process: a thorough training process on data with respect to the defined task, data can be of good quality, data can be of bad quality, the need for developing a system that learns well requires good, quality data. As data increases, every second so does the training and retraining process has to be. As a refresh happens some hidden patterns may emerge, hence it is necessary to keep the system always in the training mode for up-to-date learning.
Improves over time: there are software’s and there are systems that contain computations, if the machine is not improving with the additional amount of data and not becoming intelligent then the entire exercise is fruitless.
ABOUT THE AUTHOR
Pradeepta Mishra has 16 + years of experience and currently is the Head of AI Fosfor, leading a group of Data Scientists, computational linguistics experts, Machine Learning, and Deep Learning experts in creating value. He has expertise across core branches of Artificial Intelligence including Image Processing, Audio Processing, NLP, NLG and NLI, and design and implementation of expert systems and personal digital assistants.