Bartley Richardson

LongRoPE is making it possible to extend language model context windows, including for the Microsoft Phi-3 family of SLMs, while maintaining performance. Learn about the work, featured at #ICML2024, with podcast guest and Senior Researcher Li Lyna Zhang. https://msft.it/6043lbpeH

55

Applied Affective Computing https://bit.ly/3naKZQw provides State-of-the-art and emerging themes in affective computing, including a comprehensive review of the existing approaches to affective computing systems and social signal processing. Authors: Leimin Tian, Sharon Oviatt, Michal Muszynski, Brent Chamberlain, & Jennifer Healey. #AI #AffectiveComputing #MachineLearning #SocialRobots #BuiltEnvironments #Multimodal #DataCollection #ReinforcementLearning #Synthesizing #Natural and #Believable #EmotionalExpressions ACM, Association for Computing Machinery

1

How does the founding director of the #AITHYRA Institute Michael Bronstein work with #AI? His research focuses on geometric deep learning and graph neural networks. Learn more about his work ➡ https://lnkd.in/gjZrcu5z #boehringeringelheimstiftung #künstlicheintelligenz #artificialintelligence #ai #ki #biomedizin #biomedicine

57

AI is confusing, full of buzzwords and is only going to get more important. Solomon Grey and I are hosting our AI 101 Webinar which answers questions like, "what is a large language model, anyway?" https://lnkd.in/eQqq9Rti

10

I'm pretty frustrated with the current review process in ML (both from an author, reviewer and meta-reviewer perspective). There's possible solutions or at least experiments and changes, but I feel like business as usual is no longer feasible. There's a great overview of challenges and proposals here: https://lnkd.in/gsDqDpZe (check slides and doc). If you agree that something needs to change, I'd suggest talking to a PC member or conference organizer.

80

5 Comments

In this paper, George Mason University researchers Sanket Shukla and Sai Manoj P D propose an ML framework that efficiently harvests kinetic energy and performs energy-aware early exits with memory-optimized runtime checkpointing on edge devices like the Nano 33 BLE Sense.

44

1 Comment

Prompt engineering has emerged as a critical skill for effectively interacting with large language models. This comprehensive study examines the diverse landscape of prompting techniques, security vulnerabilities, and key considerations for prompt design. The authors have done a great job creating a taxonomical structure of prompt techniques, making it easier for AI practitioners to systematically implement and understand the relationships among them. A highlight for me was the detailed case study on using prompt engineering to detect signals of suicidal crisis in text, showcasing a step-by-step process to improve prompts and enhance results. The study also emphasizes how sensitive LLM responses are to different prompts and the black-box nature of these models, underscoring the need for automated prompting techniques to minimize result variability. Paper - https://lnkd.in/gjHtXmDD #LLM #AI #PromptEngineering

29

💫 Still thinking about the great convo I had with Joshua Patterson (CEO, Voltron Data) for Outerbounds last week. Here's a clip about the next big frontiers in data processing: ✅ AI accelerating data growth and interactions ✅ Systems asking smarter questions, faster ✅ 50-70x faster processing with lower energy use ✅ Rethinking data centers for sustainability Link to full chat in comment! 🍿

41

4 Comments

What an amazing episode of the Super Data Science Podcast with Jon Krohn! Sol Rashidi, a data executive and author which I admire a lot, shared her approach to executing successful enterprise AI projects. She emphasized the importance of balancing offensive and defensive strategies, building a cohesive data ecosystem aligned with business needs, and overcoming challenges like role confusion and data governance. Here are some highlights: 1. Build a data ecosystem that aligns with business needs ⚙️ 2. Prioritize AI projects based on impact and feasibility 🚀 3. Focus on high criticality, low complexity AI use cases 📈 4. Maintain clear data engineering and data science roles 🧩 Her book, "Your AI Survival Guide," offers valuable strategies to maximize impact while avoiding common pitfalls. Don't miss this insightful conversation! 🎯📊 #AI #DataScience #SuperDataScience #Leadership

7

3 Comments

Hey AV, Self-driving, and ML Infra (:wink:) communities! I'm proud to announce a new Nuro technical publication called "FTL Model Compiler Framework". The previous time we shared our efforts on scaling the ML Training (link: https://lnkd.in/gXBszEME ). This time, let's talk about the inference. Can we integrate multiple ML sub-compilers to provide multi-GPU serving, execution priority control, and even custom GPU kernel injection? The answer is "Yes" with our Faster Than Light (FTL) framework - https://lnkd.in/g_vryPRn ! Thanks to the authors - Ali Boubezari, Muyang Yu, Nick Korovaiko, and Hongze Z.!

63

This week at #ICLR2024, #Microsoft #Research present #MathVista, an open-source #benchmark for measuring foundation models’ #mathematical #reasoning capabilities in #multimodal scenarios. Learn about the work now from coauthor Michel Galley in the “Abstracts” podcast. https://msft.it/6042YpuIk #AI #GenerativeAI #LLM #Mathematics

9

Underrated potential #ai improvement that should make #cybersecurity folks happy: https://lnkd.in/e8Uaadkd If this scales well, cheaper, easier, and performant local AI will be easier to sell to organizations, and we'll be able to move tons of AI operations used for mundane operations onto local endpoints. Less data in transit for shorter periods of time makes life harder for threat actors. Less need to rely on cloud AI means less chance of data leakage, and less worrying if someone is going to go full Adobe and claim the rights to all your data.

1

Have even more questions about ontologies and Open Knowledge Networks (OKNs)? Read our interview with Dr. Srividya Bansal! How can these frameworks be leveraged to understand complex relationships between data? Learn more in our profile with Dr. Bansal or by browsing the NSF Proto-OKN website: https://www.proto-okn.net/ National Science Foundation (NSF) #science #data #datascience #education #ontology #STEM #computerscience #AI #ML #engineering

6

We're pumped to welcome Peng Wu of Meta to the PyTorch Conference 2024 keynote stage! Explore the list of keynote speakers: https://hubs.la/Q02Nfwc60. Register & join us September 18-19 in San Francisco! https://hubs.la/Q02NfBDM0 #PyTorchConf #PyTorch #PyTorchFoundation #AI #MachineLearning #ML #DeepLearning #OpenSource #OpenSourceSoftware #OpenSourceDevelopment #OpenSourceCommunity #OSS #LinuxFoundation #events #linux #SoftwareSupplyChainNews

36

Join Carnegie Mellon University's Philip Koopman on Sept 11, 10 AM EDT, for insights on leveraging #AI effectively. Learn about: • Classification & generative outputs • Addressing bias & safety • The 90/10 principle • Real-world Machine Learning applications Understand AI's potential and limitations. Register now! https://lnkd.in/gvhVk33y #MachineLearning #TechInnovation #BOSS2024 #ArtificialIntelligence #DeepLearning #Semiconductor #ML

8

🚀 Claude’s Game-Changing Computer Use API: Automation on Steroids! 🤖 Imagine an AI that can not only chat but actually use your computer—click, type, navigate, and gather data—all based on your commands. Anthropic’s Claude 3.5 just made that a reality with their new computer use API, and I couldn’t be more excited about its potential! Here’s how it’s changing the game: 1. Automating Mundane Tasks: No more spending hours on repetitive work! Claude can now open applications, fill forms, extract data from spreadsheets, and even complete web-based tasks, all without needing user intervention. Think of it as an assistant that handles data entry, research, and even complex workflows autonomously. 🧑‍💻 2. Real-Time Application Testing: Claude can simulate user interactions for testing—like clicking through an app and checking if it behaves correctly. It’s like having a QA team in your back pocket. 🛠️ 3. Smarter, Not Just Faster: Unlike other agents or RAG systems that simply retrieve and generate answers, Claude learns general computer skills. This flexibility sets it apart from systems like RAG or LangGraph, which focus more on content retrieval or multi-agent workflows. With Claude, we’re talking about an AI that can act independently on your desktop, improving real-time productivity. 4. Use Case Example: Imagine telling Claude to: Open your CRM, pull relevant client data, and draft emails based on that information—automatically. Or even better, navigate across apps to update a report, fetch data from the web, and send you the finished product, all while you focus on strategy. The future of automation is here, and Claude’s new API is pushing the boundaries of what’s possible. Ready to revolutionize how you work? 💡 #AI #Automation #ClaudeAPI #TaskAutomation #Productivity #FutureOfWork #Anthropic #TechInnovation

2

Explore how to fine-tune BERT using the Hugging Face Transformers library. Discover how you can leverage pre-trained models for your own projects from Shaw Talebi's article. #LLM #MachineLearning

13

New use cases for LLMs are emerging every day, and a recent paper from MIT highlights an exciting application in anomaly detection. The researchers demonstrated how LLM models can effectively detect anomalies in time series data, eliminating the need for traditional deep learning and extensive ML training. This approach can be particularly valuable in the security space, where we can apply similar techniques to analyze time series data to enhanced security monitoring.

17

Here's a snippet demonstrating how you can use Mirascope to make your LLMs take conditional actions based on the classification of an input. Check out the full example at https://lnkd.in/eV8wE5C2

11

2 Comments