Eduardo Kassner

Wondering how to deploy #responsible #GenAI applications or mitigate harmful content generation or user manipulations of existing implementations build around a #LLM? Then have a look at this iterative, layered approach shared by Microsoft: 🤖 Model Layer: Safety fine-tuning built into the model 🚨 Safety System Layer: Processing both prompts and completions through an ensemble of classification models, detecting and preventing harmful content 👩‍💻 Metaprompt and Grounding Layer: Applying effective metaprompts and advanced prompt engineering techniques 👩‍ User Experience Layer: Educating end users about the system’s capabilities and limitations ensures responsible use and informed interaction. While the model and safety system layers typically apply across various applications, the latter two layers need customization based on each application’s purpose and design. If you need further information about how to do that, we are happy to provide free consultancy and share our existing experience 💚 #ResponsibleAI #SecureAI #AzureAI

3

Microsoft has unveiled a novel, compact #AI language model called Phi-3-mini that offers comparable capabilities to large language models (#LLMs) like GPT-4, but with a significantly smaller size and reduced operational costs. Phi-3-mini has only 3.8 billion parameters, compared to the hundreds of billions or even trillions of parameters in the largest LLMs. This makes it suitable for running locally on consumer hardware like smartphones and laptops, without requiring an internet connection or powerful data center GPUs. According to Microsoft, Phi-3-mini's performance "rivals that of models such as Mistral 8x7B and GPT-3.5," which power the free version of ChatGPT. The innovation lies in the carefully curated, high-quality training data used to develop the model, which includes textbooks and synthetic data. The introduction of Phi-3-mini represents a step towards reducing the environmental impact of AI, as smaller models require less energy and computational resources than their larger counterparts. This could lead to more accessible and sustainable AI solutions for everyday tasks. https://lnkd.in/egHGu_MA

16

AWS G5 instances, powered by NVIDIA A10G GPUs, are transforming GPU-optimized workloads for AI, ML, and HPC. Learn how they outperform previous generations with better performance, scalability, and cost-efficiency. Explore real-world use cases, benchmarks, and best practices in our latest blog Read more: https://lnkd.in/gxawSJT9 #AWSG5 #GPUInstances #MachineLearning #ML #AI #CloudOptimization #AWS #EC2 #CloudOptimo

7

Newly published in Google Cloud's Architecture Framework 🏗️ this new AI/ML perspective provides a roadmap for designing, building, and managing workloads on Google Cloud. 🤖 It covers best practices for achieving operational excellence, security, reliability, cost optimization, and performance optimization. 🎯 Who should read this? 🤔 Decision-makers 👔 Architects 📐 Administrators 💻 Developers 🧑‍💻 Operators 🛠️ Key takeaways: Learn how to design and build AI/ML workloads that meet your business goals. 💼 Understand the principles of operational excellence, security, reliability, cost optimization, and performance optimization in the context of AI/ML. 💡 Get practical recommendations for implementing these principles on Google Cloud. ☁️ #AI #ML #GoogleCloud #CloudArchitecture #OperationalExcellence #Security #Reliability #CostOptimization #PerformanceOptimization

19

Incredibly proud of the latest Kinesis Client Library version (3.0) which, when compared to previous KCL versions, reduces compute costs by up to 33%. KCL 3.0 introduces an enhanced load balancing algorithm that continuously monitors resource utilization of the stream processing workers and automatically redistributes the load from over-utilized workers to other underutilized workers. This ensures even CPU utilization across workers and removes the need to over-provision the stream processing compute workers which reduces cost. Additionally, KCL 3.0 is built with the AWS SDK for Java 2.x for improved performance and security features, fully removing the dependency on the AWS SDK for Java 1.x. #Kinesis #DataStreaming #KCL

94

Big news from Microsoft Ignite! MongoDB and Microsoft Azure are teaming up to make building Generative AI apps and real-time analytics easier. Here’s what’s new: ✅ MongoDB Atlas integrates with Azure OpenAI services. ✅ Enterprise Advanced tools now on the Azure Marketplace. Simpler, faster, and more flexible data innovation starts here. Learn more in this CRN article. 👇 https://lnkd.in/gSafxumj #MSIgnite

15

4 Comments

Unifying Data Sets: The Future of Integrated Architecture. Hear Jay Gopalakrishnan, Knowi's CEO and Founder in conversation with Yves Mulkers, talk about how an unified data stack is the future of data engineering. Listen to the full interview here: https://lnkd.in/gh_ita-r #dataengineering #datastack #businessintelligence

10

To determine the potential of AI technologies across the Business Capability Model (BCM), Chief Enterprise Architects will need to develop an understanding of the kinds of problems AI can solve. Build a model that defines and classifies the types of AI technologies: https://gtnr.it/4di57o4 Gartner for IT | #GartnerIT #EnterpriseArchitecture #ArtificialIntelligence #AI

1

David Linthicum, former #Chief_Cloud_Strategy_Officer at #Deloitte_Consulting and founder of Linthicum Research, told Fierce that in many cases "#on-premises resources are going to be #more_cost-efficient, and considering that AI systems cost 2 to 3 times the cost to build and run, enterprises are not automatically sold on public cloud providers for AI." "#On-premises_systems are generally #cheaper, can be #more_secure, and feel like they provide #more_control; enterprises are opting for them more than the cloud providers would like," he added. "I suspect that the cloud providers will see good growth around AI, but enterprises will be repatriating the platform options, considering the #high_cost_of_the_cloud."

1

David Linthicum, former #Chief_Cloud_Strategy_Officer at #Deloitte_Consulting and founder of Linthicum Research, told Fierce that in many cases "#on-premises resources are going to be #more_cost-efficient, and considering that AI systems cost 2 to 3 times the cost to build and run, enterprises are not automatically sold on public cloud providers for AI." "#On-premises_systems are generally #cheaper, can be #more_secure, and feel like they provide #more_control; enterprises are opting for them more than the cloud providers would like," he added. "I suspect that the cloud providers will see good growth around AI, but enterprises will be repatriating the platform options, considering the #high_cost_of_the_cloud."

2

🚀 2024 brought major Oracle Database upgrades! From AI-driven Vector Search to Exadata Exascale, see how Oracle is making cloud databases smarter and more efficient: https://lnkd.in/gu_QUeB6

8

Results that matter for enterprises!

3

IBM #granite-guardian-hap-38m is built for speed, small enough to run on a CPU, and quick enough to filter #data at each phase of the #LLM lifecycle, from pre-processing to inferencing. IBM just #opensourced the 38-million parameter #encodermodel on Hugging Face https://lnkd.in/gb6JVgcs

15

If you’re a Board Chair and don’t have an AI Risk Assessment and Discussion Paper how will you lead the discussion this year? Christopher Foster-McBride from Digital Human Assistants is an expert that you need on your board, but in his absence you can use the tools he uses. https://lnkd.in/gE7xC47F www.ceoadvantage.com.au #airisk #ai #boardgovernance #boardchair #ceo

1