Atria AI

🚀 Exciting News from Google Research! 🚀 Google has just released a groundbreaking research paper that's a must-read for anyone interested in the future of language models. The paper introduces SPECULATIVE RAG, a novel approach that addresses some of the key challenges faced by large language models (LLMs). SPECULATIVE RAG employs a smaller, specialized language model to create multiple drafts of potential answers from different subsets of retrieved documents. These drafts are then evaluated by a larger, generalist language model, which selects the best draft based on its reasoning. This innovative framework showcases the power of collaborative architectures in improving retrieval-augmented generation (RAG) performance through task decomposition. By harnessing the strengths of both specialist and generalist models, SPECULATIVE RAG achieves remarkable accuracy and efficiency in generating knowledge-intensive responses. If you're keen on exploring cutting-edge advancements in AI, this paper is definitely worth your attention! 📚✨ #GoogleResearch #AI #LanguageModels #Innovation #SPECULATIVERAG https://lnkd.in/e3c8HJu4

Atria AI is growing! We are thrilled to welcome Goutham Boddu as our second intern at Atria! Goutham is currently pursuing his LLM in Company Law and Corporate Governance at Bristol University and is passionate about the intersection of law and advanced technology. During his time at the Dhan Foundation, he led several groundbreaking legal research initiatives, displaying a unique talent for merging traditional legal frameworks with cutting-edge technological solutions. We’re excited to have Goutham on board and look forward to seeing the impact he will make. Welcome to the team, Goutham! #AtriaAI #LegalTech #InternSpotlight #internship #InnovationInLaw #startup #genAI

We are delighted to introduce the first of our two talented interns from Bristol University, Livia Iriawan, who is joining us this month. Livia Iriawan has just completed a placement with KPMG and has a strong background in Computer Science and Business Analytics. She's now studying Law and is eager to dive into the dynamic world of legal tech. At Atria AI, Livia is excited to explore the intersection of law and technology, gaining hands-on experience in a professional setting.   Join us in welcoming Livia to the team! We're looking forward to supporting her and all the amazing contributions she'll make. #AtriaAI #LegalTech #InternSpotlight #internship #InnovationInLaw #startup #genAI

I'm fascinated by how AI has evolved and by how we can combine various techniques, stochastic and deterministic, to build better solutions. In a 2-part series, I explore knowledge graphs and try to understand how we can combine them with LLMs to improve performance. https://lnkd.in/eUZjsjuT

How can we do better prompt enginering? By automating it! New research introduces OPRO, a method using large language models (LLMs) as optimizers. By framing optimization tasks as natural language prompts, OPRO allows LLMs to iteratively refine solutions. The real game-changer? Automated prompt engineering. OPRO-optimized prompts significantly outperform human-designed ones on reasoning benchmarks, with accuracy gains up to 50%. This showcases LLMs' potential as versatile optimizers adaptable to various tasks through natural language. While challenges remain, OPRO opens new possibilities for AI-driven optimization across industries. Check out the paper below or try Anthropic's prompt generator https://lnkd.in/eM-qx_Ct #anthropic

How can we improve accuracy and precision from LLMs? Graph Retrieval Augmented Generation (Graph RAG) is emerging as a powerful approach that combines traditional vector search with structured graph databases: ↗ Enhanced depth and context in information retrieval ↗ Superior handling of complex, interconnected data ↗ Seamless integration of structured and unstructured information Graph RAG's ability to capture diverse relationships and attributes across data types potentially makes it a game-changer for RAG applications. By marrying graph data with vector search, we're unlocking new possibilities in AI-driven information processing. Checkout Googles and Neo4j's approach with Langchain #neo4j #langchain https://lnkd.in/gxYqien8

Our Founder and CTO, Faisal Sultan, had the honor of speaking at the Cambridge Judge Business School live stream event, #TheBalanceSheet. Where he delved into the fascinating world of Enterprise AI implementation and ethics. Alongside Imogen Palmer, he navigated the ever-evolving landscape of enterprise AI, discussing both the technical intricacies and the ethical considerations crucial to its successful adoption. Faisal highlighted how AI is being used to automate high-volume, low-margin processes at scale, the paramount importance of data privacy and security for clients, and the vital role of cloud providers in the AI ecosystem. Huge thanks to Conrad Kheng Hwa Chua for organising such a thought provoking discussion. Listen to why we think the benefits of AI are worth the ethical and regulatory challenges. Catch the full episode here: https://lnkd.in/e842-WQd #EnterpriseAI #AIethics #TechTalk #Innovation #AI #LegalTech #Networking #Collaboration #FutureOfLaw #AtriaAI #Startupsuccess #FutureForward #startup #GenAI #earlystagestartups

Exciting News! Atria AI has partnered with leading UK toy retailer, The Entertainer. We're piloting our state-of-the-art Pathfinder platform to automate and revolutionise their business processes. Generative AI is evolving at lightning speed, and we're committed to leveraging its transformative power to provide businesses with cutting-edge solutions. Stay tuned for more updates on how our partnerships drive innovation and efficiency! #AtriaAI #TheEntertainer #GenerativeAI #Innovation #AIRevolution #startup #Legaltech #GenAI #earlystagestartups

Thank you to the Slaughter and May Collaborate team for delivering a brilliant Cohort Day! Our special appreciation goes to Richard Batstone, Alex Baker, Emma Walton, Rhys Hodkinson, Will Seaton, Patrick Skinner, Emily Lew, and Annabel Allen for their great presentations and for providing so many “ah ha” moments. Our Co-founders had this to say: Reuben Greet-Smith shares, "I especially enjoyed hearing from cohort alumni about things they wished they had known—such invaluable insight!" Vicky Jacobs shares, "The day was exceptionally beneficial for Atria AI, providing precisely the insights our business needs. The range of topics covered was perfect, and the speakers were outstanding in their delivery and expertise." Faisal Sultan shares “It was great to hear the perspectives of lawyers and innovators, helping us understand their challenges to craft a successful product. A big thanks to Billie Kaur Heer for coordinating us all and we’d like to applaud our fantastic mentors Kiah Bradley-Arnold, and Jack Eames for all their incredible support, positivity and energy throughout the Collaborate programme so far. #Innovation #LegalTech #Networking #Collaboration #PitchPerfect #FutureOfLaw #Gratitude #AtriaAI #Startupsuccess #FutureForward #startup #GenAI #earlystagestartups #investmentopportunity

Hallucinations in LLMs....problem solved? Not yet but getting there. Scientists have developed a promising method for detecting hallucinations using a multi-layered LLM approach to identify instances where AI systems confidently generate inaccurate or fabricated information.  This is a 4-eye approach to checking the response to a prompt implemented using additional LLMs. By focusing on semantic meaning and paraphrase detection, the system shows potential for improving the reliability and trustworthiness of the initial LLM. However it could also potentially amplify the bias and hallucination, instead of 'yes, but' the system could respond with 'yes, and'. Learn more about this development in the latest issue of Nature. #AI #LLM #Innovation #TechAdvancement https://lnkd.in/gWudvaV6