Rendra Kurniawan, Aditya Putra, 2026. "Memory-Centric AI Systems: Persistent Knowledge Architectures for Next-Generation Cloud Intelligence", International Journal of Cloud Systems and Applications (IJCSA) 1(1): 17-33.

Over the past 10 years, Artificial Intelligence (AI) has grown quickly and become an industry that converts various fields with the help of intelligent automation, predictive analytics, and advanced decision-making processes. While we have made far fewer significant breakthroughs in storage-based machine learning and particularly deep learning technologies compared to the performance improvements that occurred since 1990 on computation, many AI systems still work as if they can barely remember anything when engaging in longer conversations. Standard AI models have a relatively small context window for information, leading to disconnected bits of knowledge memory and weaker continuity of reasoning. With the growing adoption of intelligent services in the cloud, there is an urgent need for research on AI systems that can leverage constant knowledge and sustained contextual understanding. The Memory-Centric AI Systems, which use memory architectures at the heart of intelligent computing systems arise from this challenge.
Memory-Centric AI System refers to a new development where memory is designed as the core building block in artificial intelligence rather than a supportive element. They are generally used to store, organize, retrieve, and update knowledge over time such that AI applications can learn from past interactions and evolve their behaviour. Memorycentric architectures enable scalable, efficient long-term knowledge management on top of advanced cloud infrastructures, vector databases, retrieval-augmented generation techniques and distributed storage technologies. Unlike conventional AI models that derive their entire performance purely from parameterised knowledge baked into them at training time, memory-centric systems have free access to external storage areas for upfront accurate contextualisation and additional detail required on-demand; together this leads to far more accurately personalised and adaptable AI.
Next-Gen Cloud intelligence propelled wide-spread adoption of persistent knowledge architectures. The cloud today opens up nearly infinite storage and high-performance computing resources as well as advanced data management services that can support the widespread deployment of memory-driven AI applications. Memory-centric systems leverage cloud-native technologies to enable knowledge to be synchronized across distributed environments, provide real-time updates and allow for the same information to be accessed, regardless of location. Based on these capabilities, intelligent systems can offer more context accurate services while not compromising on operational efficiency and scalability.

1. 1. John McCarthy (2007). What Is Artificial Intelligence? Stanford University.
2. 2. Stuart Russell & Peter Norvig (2021). Artificial Intelligence: A Modern Approach (4th ed.). Pearson.
3. 3. Ian Goodfellow, Yoshua Bengio, & Aaron Courville (2016). Deep Learning. MIT Press.
4. 4. Tom M. Mitchell (1997). Machine Learning. McGraw-Hill.
5. 5. Jürgen Schmidhuber (2015). Deep Learning in Neural Networks: An Overview. Neural Networks, 61, 85–117.
6. 6. Sepp Hochreiter & Jürgen Schmidhuber (1997). Long Short-Term Memory. Neural Computation, 9(8), 1735–1780.
7. 7. Ashish Vaswani et al. (2017). Attention Is All You Need. Advances in Neural Information Processing Systems.
8. 8. Jacob Devlin et al. (2019). BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. NAACL-HLT.
9. 9. Tom Brown et al. (2020). Language Models are Few-Shot Learners. NeurIPS, 33.
10. 10. Patrick Lewis et al. (2020). Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. NeurIPS.
11. 11. Jason Wei et al. (2022). Chain-of-Thought Prompting Elicits Reasoning in Large Language Models. NeurIPS.
12. 12. Alec Radford et al. (2019). Language Models are Unsupervised Multitask Learners. OpenAI.
13. 13. Jeff Dean (2021). Challenges and Opportunities in AI Systems. Communications of the ACM.
14. 14. Kai-Fu Lee (2018). AI Superpowers. Houghton Mifflin Harcourt.
15. 15. Pedro Domingos (2015). The Master Algorithm. Basic Books.
16. 16. David Patterson & John Hennessy (2017). Computer Architecture: A Quantitative Approach (6th ed.). Morgan Kaufmann.
17. 17. Martin Kleppmann (2017). Designing Data-Intensive Applications. O’Reilly Media.
18. 18. Geoffrey Hinton, Yann LeCun, & Yoshua Bengio (2015). Deep Learning. Nature, 521(7553), 436–444.
19. 19. Leslie Valiant (2013). Probably Approximately Correct. Basic Books.
20. 20. Richard Sutton & Andrew Barto (2018). Reinforcement Learning: An Introduction (2nd ed.). MIT Press.
21. 21. Joseph Redmon et al. (2016). You Only Look Once: Unified, Real-Time Object Detection. CVPR.
22. 22. Alex Krizhevsky, Ilya Sutskever, & Geoffrey Hinton (2012). ImageNet Classification with Deep Convolutional Neural Networks. NeurIPS.
23. 23. Sergey Brin & Lawrence Page (1998). The Anatomy of a Large-Scale Hypertextual Web Search Engine. Computer Networks.
24. 24. Tim Berners-Lee (2000). Weaving the Web. Harper Business.
25. 25. Andrew Ng (2018). AI Transformation Playbook. Landing AI.
26. 26. Michael Wooldridge (2020). The Road to Conscious Machines. Pelican Books.
27. 27. Erik Brynjolfsson & Andrew McAfee (2017). Machine, Platform, Crowd. W.W. Norton.
28. 28. Melanie Mitchell (2019). Artificial Intelligence: A Guide for Thinking Humans. Farrar, Straus and Giroux.
29. 29. Luciano Floridi (2014). The Fourth Revolution: How the Infosphere Is Reshaping Human Reality. Oxford University Press.
30. 30. Nick Bostrom (2014). Superintelligence: Paths, Dangers, Strategies. Oxford University PressBottom of Form