Abhijat Saraswat’s Post

Today's BuzzBelow post talks about Glean. Glean is an AI startup transforming corporate data accessibility. Their platform uses advanced technologies like RAG and integrates large language models for tailored knowledge graphs and chat assistants. Co-founded by Arvind Jain, T.R. Vishwanath, Piyush Prahladka, and Tony Gentilcore, Glean has raised over $200 million, offering easy implementation with over 100 connectors for businesses seeking efficient data analysis and decision-making. Read the post here: https://lnkd.in/gFqUA2YQ #GleanAI #Insights #StartupSuccess

🚀 AI and especially generative AI has been a hot topic for quite some time and use cases have started to emerge across industries and corporate functions, reducing efforts on administrative tasks and increasing efficiency. In Ørsted we recently launched our own #ChatGPT #Energenie 👩💻 through our Technology Incubator team in #ArchitectureandTechnology and we see how internal capabilities in prompting these LLM’s are improving at a high pace. It is however as essential as ever that the data and information (structured and unstructured) these LLM’s tap into are prepared, accessible, secure, of right quality and rightfully contextualized for LLM’s to navigate and provide us meaningful answers. So why are we not using AI in the first place to improve our input sources? 🤔 This topic has been on my mind for a while and was also discussed at the most recent #DataLeadershipForum (hosted by Implement) where one of the topics was how AI can be used to support #datagovernance. When aiming for a #DataMesh maturity where data is a product in its own right managed accordingly, how can we intelligently metatag our data, detect and correct anomalies and automate the tasks around maintaining the data products and keeping them clean, current and easy to consume. There are obviously many tools around for this and #datafabric offers a structured starting point but all this still appear less dynamic 📑. We are aware of the shortcomings of generative AI when answers are tough to provide accurately and we risk these models will hallucinate. I’d be happy to hear any thoughts on this and will continue to seek for ways we can make the roles around data governance and data ownership more about the data product users’ needs and less about maintaining them.   I want to add a great thanks to Victor Emil Funch and Tobias Søndergaard 🙏 who hosted our first prompt engineering session in Ørsted – and to Implement Consulting Group for hosting us in the #DataLeadershipForum 👊 that I can only recommend more data leaders to join. Reach out to Victor or Tobias 😉

Rejecting OpenAI's Merger Offer, Anthropic levelled up the game with Claude 2.1! 🚀 1️⃣ 200K Token Context Window: -Claude can now handle -much bigger -and complicated documents. 2️⃣ Hallucination Cautions: -We claim there's a 2x decrease in the AI's tendency to make things up. It's essential to check if the balance is just right between -reducing mistakes -and the AI's enthusiasm to respond. 3️⃣ Accuracy Across Spans: -With a larger context window -it's crucial to check how accurate Claude is at different points in a text. 4️⃣ Improved Homework and Comprehension Skills: -Claude 2.1 is like the homework wizard! -It claims to better understand -and summarize tricky documents. Remember, one person's loss is another person's gain! Keep your eyes peeled for all the cool things Claude can do!! #ai #aiinnovation #productivity

🚀Ai Acquisitions is paving the way for companies to positively adapt #AI and #MachineLearning into their operations!🌐 📰Here are some exciting headlines from the industry: - UK & US sign pact to develop AI safety tests 🌟 - Databricks claims DBRX sets ‘a new standard’ for open-source LLMs 💡 - Large language models could revolutionise the finance sector within two years 💼 - Stanhope raises £2.3m for AI that teaches machines to ‘make human-like decisions’ 🧠 - EU approves controversial AI Act to mixed reactions 🇪🇺 - OpenAI announces new board lineup and governance structure 📝 - International Women’s Day: What it takes to innovate in the age of Gen AI 👩💻🚺 - Pace of innovation in AI is fierce – but is ethics able to keep up? 🤔 - Anthropic’s latest AI model beats rivals and achieves industry first 🥇 Progress in AI and Machine Learning is unstoppable! Join us as we help businesses harness the full potential of cutting-edge technology to thrive in the age of AI. 💪 Visit our website for more information: ➡️ https://lnkd.in/gaG8hdp4 ⬅️ #AIacquisitions #ArtificialIntelligence #ML #Innovation #Technology #BusinessGrowth #EmergingTech

This week, we made a change to our RAG pipeline AskJunior that might seem small, but in the startup world, small changes can lead to seismic shifts. We swapped in GPT-4 Mini for routing and planning. It's like upgrading from a solid state drive to quantum storage - suddenly, everything moves faster. Our RAG pipeline is the backbone of our system, crunching through thousands of pages per cases in legal tech. It's the kind of task you'd expect to need industrial-strength AI for. But here's where it gets interesting: even with this high-powered GPT-4 mini handling the heavy lifting, our smaller models are still holding their own. The metrics are off the charts. We're not just sitting on this breakthrough. We're building an API and ingest pipeline that will let users customize their workflow. It's like handing them a programmable AI swiss army knife. We've released this to some beta testers, and the results are eye-opening. Even though we only changed the entry point to GPT-4 Mini, the entire system is humming like a well-oiled machine. And here's the kicker: I'm genuinely surprised by the planning ability we're seeing. It's as if we've unlocked a new level of AI capability. This is one of those moments that remind you why you got into startups. We're not just iterating; we're redefining what's possible in AI. And we're just getting started. In the fast-paced world of AI startups, being surprised by your own product is a good thing. It means you're pushing boundaries, challenging assumptions. This unexpected performance isn't just impressive - it's a glimpse into a future where AI can do things we haven't even imagined yet. And that's the kind of future worth building. The fact that we can leverage GPT-4 Mini power while still seeing outstanding performance from our smaller models suggests we're onto something big. It's not just about raw power anymore; it's about smart integration and optimization. We're building a system that's greater than the sum of its parts. https://lnkd.in/gcVcMtmf

At the DLA Piper European Tech Summit, Victoria reflected on how Springbok AI works with our clients to rapidly implement fine-tuned document generation modules within their private, secure and fine-tuned GPT platform. Our clients' key to success? Moving quickly 🚀 Partner with a legal-specialised Generative AI firm like Springbok AI ⬇️ Get Springbok-GPT -->> Customise and fine-tune to your firm, and brand it however you like ⬇️ Turn your backlog of Generative AI ideas into scoped modules and assistants ⬇️ Roll-out the modules and assistants quickly within your Springbok-GPT platform ⬇️ Refine and repeat. This is effective for 3 reasons: 🏆 Familiarity with Generative AI drives better use case generation. The difference in idea quality between firms who have vs don't have Springbok-GPT is remarkable. 🏆 Your firm will be known for *shipping* solutions. Success loves speed. Conversely, the most costly mistake firms can make right now is inaction. Leveraging the expertise of a #generativeAI partner can help you to get to where you'd like to be, and get these tools into users' hands. Interested in a custom GPT, fine-tuned for your firm? Get in touch with us today: https://lnkd.in/eVMG5xcx 🚀 #AIInnovation #TechnologySummit #LegalTech #DLAPiper #EmpowerYourTeam #InnovateWithAI #GenerativeAIinlegal #GenAInlegal Hannah Smith

Generative AI has created new opportunities, and you want to grab them. But, what you see is only the tip of the iceberg. It takes a lot of work to bring great ideas to production, like building secure data APIs, monitoring and deploying your GenAI application, and all the boring stuff you see. 😅 But, you know what 𝘸𝘦 think?  We think you should be focusing on that next big opportunity and let Hasura and Portkey do the heavy lifting for you. 💪 Want to understand how to become 10X efficient in building production-ready GenAI RAG applications? 🤓 Read our blog! ➡ https://buff.ly/3PTTCsY #GenerativeAI #Innovation #Productivity

Seizing the Opportunity of GPT-4o Cost Reductions   OpenAI's recent decision last week to slash token costs for its GPT-4o model by 75% to 87.5% is a game-changer for businesses looking to harness the power of AI. This significant cost reduction (see chart below from Peter Gostev) has made previously unviable AI applications financially feasible, opening up a world of possibilities across industries.   A couple of hypothetical use cases illustrate the impact: a.) A management consulting firm analyzing 1 million slide decks to create a knowledgebase for its consultants could save $550,000 to $775,000 using the batch processing. b.) An oil and gas company could save approximately $1,965,600 annually using GPT-4o vs. GPT 4 to analyze dynamometer cards for predictive maintenance c.) A law firm with 100 attorneys summarizing 5, 50 page documents per day could save $1,054,800 annually on content summarization.   To capitalize on these opportunities, leaders must reassess their AI strategies and business cases. By identifying high-potential use cases, evaluating feasibility and impact, and developing detailed plans, organizations can position themselves to reap the benefits of AI in this new cost landscape.   At Alvarez & Marsal, we have a proven track record helping clients identify, quantify, and prioritize AI use cases while designing for cost efficiency. If you're ready to unlock the potential of AI for your organization, and have hesitated because of costs, let's connect and explore how we can help you navigate this transformative landscape.   #AIRevolution #GPT4o #CostSavings #BusinessTransformation #InnovationStrateg

For AI Giants, Smaller Is Sometimes Better - Companies are turning their attention to less powerful models, hoping lower costs and solid performance will win more customers. The start of the artificial-intelligence arms race was all about going big: Giant models trained on mountains of data, attempting to mimic human-level intelligence. Now, tech giants and startups are thinking smaller as they slim down AI software to make it cheaper, faster and more specialized. This category of AI software—called small or medium language models—is trained on less data and often designed for specific tasks.  The largest models, like OpenAI’s GPT-4, cost more than $100 million to develop and use more than one trillion parameters, a measurement of their size. Smaller models are often trained on narrower data sets—just on legal issues, for example—and can cost less than $10 million to train, using fewer than 10 billion parameters. The smaller models also use less computing power, and thus cost less, to respond to each query. Microsoft has played up its family of small models named Phi, which Chief Executive Satya Nadella said are 1/100th the size of the free model behind OpenAI’s ChatGPT and perform many tasks nearly as well. “I think we increasingly believe it’s going to be a world of different models,” said Yusuf Mehdi, Microsoft’s chief commercial officer. Link to Free Full Story here: -https://lnkd.in/gCERWin4 Written By - Tom Dotan and Deepa Seetharaman https://lnkd.in/gCERWin4

Now is the Time for SWIRL AI Connect: Empowering Enterprises with Next-Level AI Infrastructure Software In today’s rapidly evolving technological landscape, businesses are aggressively turning to Artificial Intelligence (AI) to gain a competitive edge, optimize operations, and innovate at unprecedented speeds. What leading enterprises are telling us is that SWIRL AI Connect is: “The Right Product, at the Right Time, and in the Right Place” This blog post explores why now is the perfect moment for enterprises to integrate SWIRL AI Connect into their operational frameworks, thereby transforming their potential to leverage AI for sustainable growth and innovation. Last month, we celebrated Pi Day on 3/14. This is also the first anniversary of GPT-4’s historic launch. Unlike any other, it attracted more than 100 million users by demonstrating truly inspiring capabilities with clear potential to change how business works. Of course, the large enterprise promptly blocked it. Major publishers changed their terms of service to deny it access to licensed content, and a few filed suit against OpenAI. Today, large enterprises have done what they always do: engage with disruptive technologies in their arena. Many have successfully obtained and installed LLMs in their private clouds. Trained with annual reports and employee handbooks, these cheerful demos offer tantalizing hints at the future. But there is one problem: they can’t answer business questions.  More: https://lnkd.in/edc4gK-W