IBM Research India 25: Monthly Impact Talk Series

Recording of the Talk

Abstract: Lack of Trust in AI is one of the major impediment towards large scale adoption. IBM Research has been at thefore front of developing algorithms, toolkits and systems to enable diverse stakeholders to infuse Trust in AI. In this talk,we will go over different pillars of Trust, their importance and associated tool kits.

Recording of the Talk

Abstract: Large language models (LLMs)such as GPT-3, have revolutionizedthe field of conversational systemsby providing advance capabilitiesfor natural language understandingand generation. However, thesemodels still pose significantchallenges and limitations such asgeneration of factually incorrect orin-consistent information.Therefore, adoption of LLMs forbuilding chat-bots for enterpriseshas been slow. In this session, wewill talk about different methodsthat we are exploring to providefactually correct, consistent andtrusted information in responsegeneration and how LLMs can beused to build trustworthy chat botsfor enterprises.

Recording of the talk

Abstract: Blockchain and distributed ledger technology have made an enduring impact, especially in the areas of finance and supply chains, far beyond cryptocurrency that was the focus of its initial promoters. As this technology continues to pervade and transform all of our legacy physical and software systems, and as more and more of our data starts to reside on distributed shared ledgers, the need to ensure high performance, privacy safeguards, and especially interoperability among systems of various flavors has become an imperative. In this session, we will discuss the pioneering role played by IBM Research India in advancing blockchain technology and ensuring that it possesses these characteristics by making seminal contributions.

**Abstract: **Hybrid Cloud computing has revolutionized how enterprises build, deploy and manage software across public, private, on-prem, and Edge data centers. While agility and scalability have certainly improved, enterprises struggle with challenges around ensuring reliability, performance, security and compliance. In this talk, we will cover various contributions from IBM Research India towards a reliable Hybrid Cloud. We plan to discuss technologies to migrate workloads to cloud-native platforms, AI-based operational analytics to detect and remediate problems quickly, and techniques to meet regulatory obligations despite the dynamism and ever increasing complexity of Cloud environments.​

Abstract: Benchmarks are designed to mimic a particular type of workload on a component or system of supercomputer to compare platforms, identify performance bottlenecks, evaluate potential solutions, and help users in their decision of buying or using machines most appropriate for their application requirements. One of the most challenging benchmarks on supercomputers has been the rate at which the global memory can be updated - the memory wall. In this talk we will discuss optimization techniques applied for two such benchmarks, the HPCC Random Access and Graph500 benchmarks, that have led to IBM's Blue Gene winning the HPCC class 1 award for 6 years in a row and the Graph500 award for 2 years.