Hemathri Balakrishnan

Now that we know there is only so much we can do with #PromptEngineering with these #Large and #Larger #LLMs here comes the phase of rationalization. Going by the literature some of these are much simpler and would turn out to be much cheaper than the other lazy alternatives that we are getting comfortable with. I don't want to undermine the effort for #DataPreparation and #Evaluation required to ensure the results are worth the effort. Sure, it means a lot of hard work! - Improve #Retrieval: Use contrastive loss to train a dual encoder embedding model for both questions and answers - Improve #Generation: Finetuning with use-case specific instructions and responses #weekend #learnings - https://lnkd.in/g93svug3 - https://lnkd.in/g3wYxJGR

24

As we move into a realistic adoption scenario, deterministic models, with probabilistic learnings fed into them, will pave the way for a meaningful evolution of LLMs.

8

Exploring the AI Jungle: For our June episode of the Data Book Club, we debated insight in AI strategy and breakthroughs from AI/ML research from six bestsellers on the topic of #AI. 1. Architects of Intelligence - The truth about AI by the people building it 2. The Book of Why - The new science of Cause and Effect 3. The Coming Wave - AI, Power & the 21st Century 4. Generative Deep learning - Teaching machines to Write, Paint, Compose & Play 5. Genius Makers - The Mavericks who brought AI to the world 6. The Master Algorithm - The quest for the Ultimate Learning Machine We discussed AI strategic themes that included: ▪️AI DNA – How do we build prior knowledge into AI? Avoid training from scratch. ▪️AI Learning Strategies – how do we generalise and remember better? ▪️Emotional AI – approaches for managing 'what' and 'how' we goal seek with machine intelligence. ▪️ Causal AI – how do we move from correlation to understand cause and effect relationships in the world around us? ▪️Imaginative AI – how do we create new possibilities, form new ideas and concepts through to testing them? ▪️Unified AI – how do we combine a unified view across the contemporary machine-learning paradigms? AI/ML research is progressing on many fronts for these questions - its a big jungle out there. 🌴 Examples of ideas we discussed: ▫️Learning by Algorithm competition with the evolution of GAN architecture. ▫️Model-free vs. Model-based reinforcement learning using simulation ▫️Causal AI for understanding cause and effect for true explain-ability and counter factual reasoning. ▫️xLM The many flavours of Language Models - from large to small to visual language models. ▫️Human aware AI - from Artificial Social Intelligence, Agentic AI to Large Action models. ▫️and so much more.   Takeaways ● Prepare for an uncertain and fast moving roller-coaster so 'Built for Change' principal will be important for agile, adaptable, open & flexible architecture. 🤸‍♀️ ● Invest in your AI/ML skills. We are still in the early stages of AI adoption; lot of options out there and still lots to be done. 💪 ● #GIGO remain as true today as ever. Without Good data there is no Good AI. ℹ️ ● Learning could become our most important asset & resource (so how are you managing & securing)? 🎓 A fascinating set of books that recommend as AI will be a central force in our future. So much insight we had to break up over two episodes. What AI strategic themes and research ideas are you most interested in? Our #westpac data book club, is our member-driven community for the curious, data inspired and passionate continuous learners. 📚 #databookclub #ArchitectsofIntelligence #GeniusMakers #BookofWhy #ComingWave #MasterAlgorithm #GenerativeDeepLearning #ArtificialIntelligence #DataScience #AdvancedAnalytics #MachineLearning #DeepLearning

48

7 Comments

"𝗜𝘁'𝘀 𝗰𝗿𝗶𝘁𝗶𝗰𝗮𝗹 𝗮𝘁 𝘁𝗵𝗶𝘀 𝗽𝗼𝗶𝗻𝘁 𝘁𝗼 𝗶𝗻𝘃𝗲𝘀𝘁 𝗶𝗻 𝗯𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗮 𝗾𝘂𝗮𝗻𝘁𝘂𝗺 𝘁𝗲𝗮𝗺. Quantum computing is still in the R&D phase, as quantum computers are not yet powerful enough to be usable in production. However, it's important to become quantum-ready now." Wise words from Marco Pistoia and great to see HSBC listed as a "leader" rather than a "lagger" in the latest American Banker piece. See the full article 👉 https://lnkd.in/eqhhbvAS #Quantum  #QuantumTechnology  #Innovation

193

5 Comments

Building an end to end Generative AI stack on Oracle Cloud Infrastructure (OCI) requires a multi-layered approach for integrating artificial intelligence within enterprise systems. This reference architecture describes a four layer AI stack and all the different components that are needed to implement an enterprise grade Generative AI solution within an enterprise setting. 1. Application Layer 2. Access Layer 3. Logging and Monitoring across the solution 4. AI layer consisting of the following five modules: - AI integration - LLM - AI Development - Data Integration - Context and Data Catalog For Strategy and Finance you can consider the following use cases: a. Synthesis of unstructured data from earnings calls, analyst reports and other sources. b. Automation of complex, context heavy processes like expensing. c. At scale monitoring of competitors and customers, across public or private sources More information on this link: https://lnkd.in/gH3qHUch David Ricardo Orellana Moyao Ernesto Osorno Rodriguez Monica Salaris Luis Valdez Cassius Di Cianni Thiago G.O. do Amaral Vitor Alves Junior Diego Chaves Carlos Chacon Tosso Marcos E Palma M Acknowledgments Author: Badr Tharwat

36

🎓 Exciting Update! 🎓 Amidst the joys of parental leave, I seized the opportunity to delve into the world of AI by completing an AI Application for Growth course from Northwestern University - Kellogg School of Management - Executive Education! But what sets this course apart? Here's a glimpse: - Gain insights into how both established companies and start-ups are leveraging AI to disrupt industries and gain a competitive edge. - Understand the tangible business applications and outcomes achievable through AI. - Navigate the complexities of AI implementation, from strategy development to ethical considerations. This journey not only allowed me to invest in my professional growth but also sparked a newfound passion for innovation in my role. I extend my heartfelt thanks to Mohanbir Sawhney for his invaluable contributions to my learning experience. Although our interactions were limited, the insights and industry knowledge you shared during those moments were truly invaluable and greatly appreciated. I would also like to express my gratitude to the program leaders, Dr Tayo Olabumuyi, P.E., Danny Malter, Alonso Hurtado, Edmund Sefton for their unwavering support and guidance throughout the course. Your insights and encouragement played a significant role in helping me navigate this learning journey with confidence and determination. Here's to embracing new challenges and staying at the forefront of technological advancements! #AI #ProfessionalDevelopment #ParentalLeaveSuccess #ContinuousLearning #CareerGrowth

76

9 Comments

Generative AI is revolutionizing the Investment Industry. Watch this 1 min video to see how MSCI Inc. is partnering with Google on our GeoSpatial solution for managing climate hazards and physical risks at Google Cloud Next 24 conference.

231

7 Comments

I delved into various use cases in trading, investing, risk management, and collateral management while exploring a fascinating study on data. The study highlighted how financial organizations can leverage AI/ML to gain deeper insights and predictions through data-driven approaches. Large datasets necessitate extensivecalculations, prompting firms to strategize effectively. Key elements of the approach involve the use of quantitative strategies, relying on statistical and mathematical models intertwined with human intuition. Unlike traditional methods, AI/ML models aim to extract unique insights that human analysis might overlook. The realm of quantitative finance is evolving with generative AI and LLMs, offering robust tools for automated decision-making, predictive modeling, and data analysis. Despite the potential benefits, hurdles remain in the adoption of Gen AI and LLMs in the Financial Services industry, including the need for domain-specific models and regulatory adherence. By addressing these challenges head-on, technology can pave the way for the seamless integration of Gen AI and LLMs within the sector, revolutionising operational efficiency and decision-making processes. #aiapplications #ai #aiadoption #machinelearning #quantfinance #genAI TransformHub Nischal Tanna Mayur Tanna Nilesh Kanawade Sonal Tanna

19

#GenAI and #LLMs models hallucination is real and a still a big challenge to resolve if it is ever meant to be widely adopted by enterprises. But what’s more interesting is that no one seems to even care! #AI #BigTech https://lnkd.in/demhkNvB

10

3 Comments

JPMorgan Chase recently announced the deployment of AI for its 140,000 employees as part of a strategy to optimize every operational process (Wilkinson, 2024). To achieve this, CEO D. Pinto decided to reorganize his leadership team with profiles focused on technology and operations. This strategy promises significant gains, but will it not encounter limitations? Indeed, JPMorgan appears to favor a strategy centered on technological optimization and a top-down approach. The company imposes solutions designed by management without sufficiently involving stakeholders in the design of this change. This directive approach, focused on data-driven optimization, may face internal resistance if employees do not feel included in the decision-making process (Bushe & Lewis, 2023). Furthermore, this "Top Down" strategy can also lead to "silent killers," that is, internal barriers that will prevent the effective implementation of AI, such as poor coordination or vertical communication issues (Beer, 2021, p.17). An effective organizational change strategy emphasizes employee engagement from the outset to create genuine alignment and authentic buy-in (Bushe & Lewis, 2023). #HECDOOD #AI #OD Wilkinson, L. (2024, September 11). JPMorgan Chase launches generative AI tool for employees. CIO Dive. https://lnkd.in/e_Gx_Xqp Beer M. (2021) . «Reflections: Towards a Normative and Actionable Theory of Planned Organizational Change and Development» ,Journal of Change Management , vol.21 , no 1 , p. 14-29 Bushe G. R. & Lewis S. (2023) . «Three change strategies in organization development: data-based, high engagement and generative» ,Leadership & Organization Development Journal , vol.44 , no 2 , p. 173-188

6

What a great event and even greater discussions & sessions... Thanks 3AI for sharing this reel. It illuminates the essence of #innovation zone perfectly. I invite you all to check out my post for more details. https://lnkd.in/gXdCjBvb #youinaction #artificialintelligence #AI #ML #genai #technology #womenintech #leadership #event #financialcrime #antimoneylaundering 3AI Sameer Dhanrajani Nandakumar Ramaiah Kapil Gandhi

15

3 Comments

🌟 New from #NVIDIAResearch: Weight-Decomposed Low-Rank Adaptation (DoRA). A groundbreaking advancement in fine-tuning technology that's set to revolutionize how we optimize pretrained models without increasing inference costs. 👀 https://dy.si/qdKovS

11

1 Comment

Interesting positioning of SageMaker as infrastructure when it comes to GenAI, where it’s another tool when it comes to AI/ML or home built Foundation Models #SageMaker #Bedrock #GenAI #AWSSummit

22

2 Comments

CNI choice in Kubernetes always dragged me in serious debates and comparisions. To add a humour laden flavour , here is a bit of gig :) A humorous comparison of Calico and Cilium: Note: I am a big fan of both and cant choose one over the other.. Calico: - Like a calico cat - colorful, playful, and a bit quirky. - "Hey, I'll just randomly assign some IP addresses and hope for the best!" - "Networking? Meh, I'll just wing it!" Cilium: - Like a ninja - stealthy, precise, and deadly efficient. - "I'll sneak up on those packets and route them with ease!" - "Networking? Child's play. I'm a ninja, hear me roar!" Of course, this comparison is all in good fun, and both Calico and Cilium are powerful tools in their own right. Calico is known for its flexibility and ease of use, while Cilium is renowned for its performance and security features. Choose the one that fits your needs, but remember - with great power comes great responsibility :). Have a great weekend !!

18

4 Comments

Pundits' predictions notwithstanding, financial institutions are increasingly building customized large language models (LLMs) like JPMC's "LLM Suite" and Morgan Stanley's "Debrief." This demonstrates the strategic importance of generative AI, leveraging advanced techniques like Agentic RAG and Mixture of Agents for innovation and competitive advantage. Oracle empowers this trend by providing foundational tools that abstract out model optimization complexities, enabling a focus on core business needs and responsible AI adoption. This marks a significant shift in the industry, showcasing the power of generative AI to drive growth and innovation.

12

This article is short; however, the implications are profound. After reading a McKinsey article on Gen AI's impact on the software sector, I'm pondering potential gains in the data space. Here are a few disciplines to consider. * Self-Service Analytics  Resource consumers: Extracting and combining data, data quality, shadow IT, sunsetting obsolete solutions, complying with regulations, data sharing * Reporting and visualisation  Resource consumers: Combining data, building data storytelling, KPIs, sunsetting old KPIs and reports, managing access * Combining data Resource consumers: Extracting and combining data, data quality * Data Management  Resource consumers: Putting data products, AI solutions and KPIs in a standard inventory/catalogue for discovery and reuse and the classification of these assets to better manage IAM and regulatory requirements My interpretation of the article's graphics and underlying material is that I cluster Data Aggregation, Analytics virtualisation, and Data and AI development with a total of 26% when using market size as a proxy for disruption. The critical point is that we anticipate a new wave of solutions that will likely significantly reduce the number of full-time employees (FTEs) required to perform the same work within these disciplines. These solutions will release resources to enhance business processes and cut costs. The reduced costs will free up budgets for investing in these solutions. Please note that I do not expect a lower cost of ownership that will go up; however, the big value for the business is a significantly shorter time to market. * We, data professionals/leaders, need to prepare for this change that has already begun.  * We should consider how these tools improve productivity and how to drive the adoption of new ways of working to harvest the impact. * We need to understand how the vendor landscape will change, forcing us to change the preferred vendors and take a more active approach to avoid lock-in. * What competencies will be needed going forward? (more business/less technical) My key takeaway is that we should implement this quickly, as our stakeholders will be looking to us to drive this change and enable productivity gains. Have a lovely Friday.

12

1 Comment

Given the rapid advancements in Generative AI over the past year, I wanted to revisit a session I had the privilege of speaking at during AWS re:Invent last year. Together with experts from AWS, Emil Lerch and Adnan Bilwani, we discussed the transformative potential of this technology in the software development lifecycle (SDLC) and I shared FINRA's R&D project on the role of code assistants and how they are reshaping the way we build and deliver software. During the session, we delved into practical use cases, discussed the benefits and challenges of adopting these technologies, and shared insights on integrating them into existing workflows. Catch the session here: AWS re:Invent 2023 - Generative AI tools for Developers https://lnkd.in/eZVHvhyt As generative AI continues to evolve and mature, it’s important for us as technology leaders to understand its capabilities, limitations and consider how we can leverage these advancements to drive efficiency, innovation, and competitive advantage in a responsible manner. #GenerativeAI #SoftwareDevelopment #AWSreInvent #Innovation #Leadership #FinraTech

52