Mark Ryan

🇨🇦 Pumped to announce that Canadian organizations can now both store their data at-rest and perform machine learning (ML) processing for Gemini 1.5 Pro and Gemini 1.5 Flash in Canada. This means that a customer’s data will not leave the region when asking a model to generate a response. We know this is important to many organizations here in Canada, and it's critical to leveraging gen AI to its fullest.…more details here: https://goo.gle/4gbOcWO

532

16 Comments

I learned a LOT at #aisummit2024 Vancouver Here are some of my notes. https://buff.ly/40xISHU #ai_summit_2024

4

1 Comment

Such a key capability for Enterprises to unlock the value in sensitive data. #Azure #AI Confidential Inferencing is now preview. Great to see RBC #Canada mentioned at the forefront of Azure Confidential Computing innovation. https://lnkd.in/gUt3MZS5

6

Hello Vancouver! If anyone is interested in learning more about AlloyDB, Google is hosting an immersive, free hands-on AlloyDB workshop. For those who don't know AlloyDB is a hybrid transactional/analytical Data warehouse based on PostGresSQL which is 4x faster at transactional queries and up to 100x faster at analytical queries. Definitely worth checking out. RSVP before we run out of seats. https://lnkd.in/g_uAnVy5

15

Retrieval-Augmented Generation (RAG) has transformed how we ground LLMs in reliable data sources. But what happens when your knowledge base includes images like charts or photos? 🤔 Enter multimodal RAG—a game changer that lets you integrate image sources into your RAG workflows for more comprehensive answers. 🛠️ In this blog, Pamela Fox walks us through the process of setting up multimodal RAG in your own applications—from integrating image embeddings to handling citations from image sources. https://lnkd.in/gjiVzxWX Are you ready to level up your RAG applications? 🚀 P.S. If you found this post helpful, repost ♻️ and let others in your network know about multimodal RAG! #AI #msftadvocate

46

Transcendence Series: The Founders and Leaders of Deep Learning   Yoshua Bengio, Geoffrey Hinton, and Yann LeCun are three Canadian computer scientists who are widely regarded as the founders and leaders of deep learning, a branch of artificial intelligence that uses neural networks to learn from large amounts of data and perform complex tasks such as image recognition, natural language processing, speech synthesis, and self-driving cars. They have also received numerous awards and recognitions for their work, including the Turing Award, the highest honor in computer science, in 2018. Yoshua Bengio is a professor at the University of Montreal and the founder and scientific director of Mila, the Quebec Artificial Intelligence Institute, the largest academic research center in the world dedicated to deep learning. He is also a co-founder and senior fellow of the Vector Institute, a research institute for artificial intelligence based in Toronto. He has published over 500 papers and several books on machine learning, deep learning, neural networks, and related topics, such as Learning Deep Architectures for AI, Deep Learning, and Neural Networks and Learning Machines. He is also an entrepreneur and a co-founder of several companies that apply deep learning to various domains, such as Element AI, Imagia, and Lyrebird. Geoffrey Hinton is a professor emeritus at the University of Toronto and was vice president and engineering fellow at Google - he departed on 2023 citing concerns on the usage of AI technology. He is also the chief scientific adviser of the Vector Institute and a founding director of the Gatsby Computational Neuroscience Unit at University College London. He has published over 200 papers and several books on machine learning, neural networks, cognitive science, and artificial intelligence, such as Parallel Distributed Processing, Connectionist Models of Cognition and Perception, and The Handbook of Brain Theory and Neural Networks. He is also an inventor and a co-inventor of several algorithms and techniques that are widely used in deep learning, such as backpropagation, Boltzmann machines, contrastive divergence, and dropout. Yann LeCun is a professor at New York University and the chief AI scientist at Facebook. He is also the founding director of the NYU Center for Data Science and the co-director of the NYU Center for Neural Science. He has published over 300 papers and several books on machine learning, computer vision, neural networks, and artificial intelligence, such as Convolutional Networks and Applications in Vision, Deep Learning, and Theoretical Computer Science. He is also an innovator and a pioneer of several breakthroughs in deep learning, such as convolutional neural networks, the LeNet architecture, and the MNIST database.   Note: This is the twentieth post in the series. The goal is to highlight significant figures who have made an impact on technology and its application to human endeavors. 

4

There are lots of awesome works (research papers, tools, libraries,...) right now around Large and Small Language Models (LLMs and SLMs). In you're looking for ways to : * have more control about LLMs output * better ouputs (avoid hallucination) * faster execution  * lower cost I really recommend to look at {guidance} python library which allow you to write guidance program instead of language text prompt (prompt engineering is not a dev thing right :-) The idea with {guidance} is that under the hood, Large Language Models get back the next token and the next token and so on (4 characters each) based on the input prompt (system, user …) based on intensive maths calculation (token probability distribution) which can eventually be change with the temperature parameter. So, what about techniques to avoid generating tokens (because you already know you want it), constrain tokens with possible values and so on... I really like the analogy mentioned by Harsha Nori taking about {guidance} is like a code compiler that transform high level language (python, .net) to specific hardware optimized instructions. here is Microsoft BUILD session talking about {guidance} https://lnkd.in/evpSmbsa #SLM #LLM #AI #MicrosoftAI

13

1 Comment

Most of the generative AI use cases I work with customers on leverage Retrieval Augmented Generation (RAG) in one way or another. Search results are crucial, which is why I always recommend a comprehensive search service like Azure AI Search to capitalize on vector and semantic search results. However, sometimes even that does not yield the accuracy needed. The solution? Read this insightful blog by my colleagues Andrew Ditmer and Liam Cavanagh to discover how to implement #superRAG. [Link to the blog: https://lnkd.in/ebSE_yte] #AI #ArtificialIntelligence #AzureAISearch #RAG #superRAG

29

From my conversations, I'm noticing different trends between how AI engineers deploy evaluation models (model as a judge) in production monitoring vs application testing. 💡 Value - Monitoring: provides real time alerts of performance or used to preemptively block problematic outputs from surfacing to the user -Testing: regression testing applications to establish baseline performance before pushing to deployment ⚡ Speed - Monitoring: near real time, requires extremely fast token processing - Testing: batch, async, often processing large test sets, lower token processing speed required 🎯 Accuracy - Monitoring: overindex on false positives, better to flag something as problematic and be wrong than to let something slip between the cracks - Testing: high correlation with human graders, be as accurate and robust as possible 💰 Cost - Monitoring: prefer cheaper models, otherwise the additional inference cost per generation will add up exponentially - Testing: willing to take on higher costs, calculation is accurate testing leads to better performing outputs 🛑 Risk - Monitoring: if your monitoring judge gets something wrong, there is a high consequence because it directly impacts the end user - Testing: if your testing judge gets something wrong, async review gives some potential to remediate the issue At the moment, it seems using model as a judge for testing is a little more popular due to the high cost constraints of constant monitoring, but that is subject to change as more specialized models come out. As the industry matures, I strongly believe we're gonna see specialization between models for monitoring and models for testing. If you're currently using model as a judge techniques, would love to chat more! Let me know what you think. Twitter Handle: https://lnkd.in/eAsQvtiu

26

Thats the exact explanation I gave to a group of investors last week. (… it’s just Precision/Recall, attributes vectorization as “embedding”, K-Nearest Neighbor, User Search Intent, DAG, ML algorithms, etc all over again …). LOL and they asked what’s new this time around. And I was like not much, ML is still the backbone, and AI is the inference and actions based on the model. It works better when there is a feedback loop in a closed system. But the current state of the art like ChatGPT is based on scraping the Internet (garbage in, garbage out). Kinda like early search engines. What is a better analogy is that NVIDIA/Snowflake etc are selling tools, aka pickaxes to prospectors. The gold is up to you to find. And the opportunity right now is to monetize industry specific (healthcare, travel, financial services) clean data to build models with (aka a treasure map for the prospectors). From an enterprise perspective, what’s out there today aka generative AI isn’t very useful. Too much hype. To build useful engines will require an effort similar to search revolution 2 decades ago. From

16

3 Comments

❄ In case you missed it, Snowflake recently released Arctic, a family of performant and compute-efficient Apache 2.0 licensed Large Language Models (LLMs). Curious about how Arctic might fit your needs? Come explore its capabilities at Clarifai alongside the 70+ other LLMs currently available on our platform. Let's build the next groundbreaking GenAI app together! https://lnkd.in/gNz5Babr

20

Data Sovereignty and privacy are key factors in new AI specific hardware sales and glad to see Google make a concerted effort on this front. Time to get a Pixel 9 pro and start creating some "gems"! #Pixel #Ai #ML #machinelearning #techpolicy #privacy

5

Of all the enhancements we had in store with respect to #DML in #BigQuery in 2024, we saved the best for last (references to the rest are in the first comment) Introducing Fine-grained DML, now in preview! This closes one of the remaining gaps for our DW users who often perform large scale transformations of their data that involve scanning a large corpus, but ending up modifying a small percentage of it. Or, performing pervasive mutations on a non-clustered set of columns in a table (e.g., regular pipelines that perform GDPR-type compliance jobs). For these types of workloads, fine-grained DML offers superior and more cost-effective performance (and this will be the professional, measured language to say "the improvements are totally bananas!!! 😃). Read Pavan Edara's teaser on how this works, and give it a try. You will not be disappointed, I guarantee it! Kudos to the whole team who has been working continuously for a long time on this now: Pavan Edara, BIGANG LI, Zhanqi Liu , Yaling Wu, Zhaozhe Song, David Thivierge-Gaulin, Hua Deng, Yi Yang cc: Tomas Talius

44

4 Comments

Latest innovation that helps organizations fast-track their Generative AI initiatives within their AWS environment. 🔑 Key Benefits: • Dramatically reduced time-to-design for GenAI solutions, Seamless integration with existing AWS infrastructure, Leveraged open-source assets for cost-effective implementation, Enterprise-grade security within customer's own AWS tenant 💡 This launch represents a significant step forward in democratizing AI development, making it more accessible and efficient for Businesses of all sizes. 🤝 Whether you're just starting your AI journey or looking to scale existing solutions, our platform provides the tools and framework to accelerate your success.

20

Curious about combining AI with real-time data retrieval for more accurate results? Discover how to build Retrieval-Augmented Generation (RAG) applications on VS Code's AI Toolkit! https://lnkd.in/gVQPKwZi In this blog, Vinayak Hegde breaks down the process of creating a RAG database using ChromaDB, LangChain, and Python, transforming document retrieval into a seamless experience. From chunking documents to embedding vectors for precise searches, learn how this hybrid approach can enhance content generation, legal references, and more. Dive in and start building your AI today! #RAG #AIToolkit #LangChain #ChromaDB #msftadvocate

18

I had the opportunity to attend the Toronto Machine Learning Society (TMLS) Summit. Sharing my learnings and observations: ➡ RAG RAG RAG ... and did I say RAG? Contextual Semantic Search is how Enterprises are unlocking the value of GenAI ➡ Evaluating the accuracy and reliability of LLMs is becoming increasingly important. Frameworks such as RAGAS.io and TruEra are there to help! 🏁 Hamza Farooq hosted an incredibly interactive session covering the LLM ecosystem and what to take into account for GenAI applications. 🆕 Royal Sequeira Aslesha Pokhrel Christopher Tee delivered a hands-on workshop covering DSPy to optimize your prompts like a software engineer. First interaction with this framework and I really liked how modular it is and felt like Object Oriented Programming for LLMs. 🔗 Yizhi Yin covered Graph RAG and helped us build a personalized product recommendation engine. Pretty cool. 🧠 Ehsan Amjadian talked about lessons learned from building a RAG solution for the financial industry. Very insightful. I need to look into beyond just my chunking strategy 🤓 🙏 Thank you to David Scharbach and the entire organization committee for a great event. The sessions were incredibly useful, relevant and dynamic. I took so many notes and learned directly from AI/ML leaders building in the real world. Looking forward to learning more from this amazing community and contributing to future events! #ML #GenAI #AI #Events #Toronto #summit

39

3 Comments

New post! Probably the most desired enhancement for database privilege management is finally here in 23ai. #database #sql https://lnkd.in/gXkXGWWH

31

Meta's Llama models have set a new standard for open large language models (LLMs) in the industry. With a remarkable 10-fold increase in downloads from last year, Llama models are leading the way in openness, modifiability, and cost efficiency. This advancement empowers customers to create innovative solutions tailored to their specific needs. Excitingly, Oracle Cloud Infrastructure (OCI) Data Science now fully supports Llama 2, 3, and 3.1 models, even on CPUs. #Meta #LlamaModels #OracleCloud #DataScience

10

This is a much easier way to get and load the ONYX model for generating 23AI Vector Embeddings. https://lnkd.in/gMgKPNgC

14