International Journal of Science, Engineering and TechnologyAuthors: Anisha Singh, Anubha Khanna, Ansh Goyal, Abhijeet Sharma
Abstract: Selective forgetting is a natural part of the human thinking that enables the effective thought preparation, plasticity and decision making. By doing so, humans are unable to store the information indefinitely but this way they prioritize the knowledge that is related and eliminates irrelevant and outdated information with the course of time. The outcome of this process is that this avoids overloading of memory, allows adaptive thinking as well as flexible generalization functions, which are not easily replicated by the existing artificial intelligence (AI) systems. In comparison, the existing AI systems are primarily designed with the paradigm in which continuous data storage is useful in that the more a data is stored, the more effective the performance is developed. However, this approach has drawbacks such as being slowed by computer, overfitting, reinforcing bias, maintaining stale knowledge, and privacy. Being a concept of machine learning, forgetting may be used as pruning to improve generalization; it may be used as a concept in cognitive science, it may help to eliminate irrelevant or damaging information; and as an ethical concept, it can be used with the issue of the right to be forgotten in legislation, such as the GDPR. The article inspecks the notion of whether intentional/ algorithmic forgetting is necessary to develop Artificial General Intelligence (AGI). Theoretical memory management model is also adopted through imports of the significance, frequency of use and time based wear to evaluate and optimize data relevance. This model is utilized with the assistance of a system called ForgetAI that demonstrates the optimization of memory in real-time with the aid of scoring and filtering systems. Through experimentation, it has demonstrated that selective forgetting relieves memory load without any noticeable effect on system performance and plasticity. This paper thesis is that the concept of an ideal memory retention may not be most intelligent system-wise and that controlled forgetting can enhance efficiency, equality and scalability. The research questions which the studies address are the following: does forgetting increase the adaptability of AGI and what can be done to optimize memory methods to transform AI systems to be more fair and human-like?
DOI:
