IIT Hyderabad develops method to understand inner workings of AI models

IIT Hyderabad develops method to understand inner workings of AI models

New Delhi, September 10: A method for understanding the inner workings of Artificial Intelligence models in terms of causal attributes has been developed by the researchers of IIT Hyderabad.

AI models mimicking working of human brain

So that machines can learn to make decisions in a more human-like manner, ‘Artificial Neural Networks’ (ANN), which are AI models and programs, copy the working of the human brain. Modern ANNs, also known as Deep Learning (DL), have increased greatly in intricacy that can almost match human performance in many tasks.

Dr. Vineeth N. Balasubramanian, Associate Professor, Department of Computer Science and Engineering, IIT Hyderabad, and his students Mr. Aditya Chattopadhyay, Mr. Piyushi Manupriya, and Mr. Anirban Sarkar have performed this work.

The Proceedings of 36th International Conference on Machine Learning has recently published their work. It is one of the highest-rated conferences in the domain of Artificial Intelligence and Machine Learning.

Speaking about this research, Dr. Vineeth Balasubramanian told a national daily, “The simplest applications that we know of Deep Learning (DL) is in machine translation, speech recognition or face detection. It enables voice-based control in consumer devices such as phones, tablets, television sets and hands-free speakers. New algorithms are being used in a variety of disciplines including engineering, finance, artificial perception and control and simulation. Much as the achievements have wowed everyone, there are challenges to be met.”

‘Interpretability problem’- a key blockage

The ‘interpretability problem’ is a major bottleneck in accepting such Deep Learning models in real-life applications, especially risk-sensitive ones. Because of their complexity and multiple layers, the DL models become virtual black boxes that are difficult to decipher. Therefore, troubleshooting becomes difficult, if not impossible when a problem arises in the running of the DL algorithm, explained Dr. Vineeth Balasubramanian.

“The DL algorithms are trained on a limited amount of data that are most often different from real-world data. Furthermore, human error during training and unnecessary correlations in data can result in errors that must be corrected, which becomes hard. If treated as black boxes, there is no way of knowing whether the model actually learned a concept or a high accuracy was just fortuitous,” added Dr. Vineeth Balasubramanian.

The team of IIT Hyderabad is working on this problem with ANN architectures using causal inference with a ‘Structural Causal Model.’