Sonali Verma’s Post

An exciting potential use case is that a newsroom could leverage GPT to generate different angles to hook someone into reading the entire article. It also leverages human-in-the-loop, which is a must for now. It makes me think about what the future holds. Could an article be tailored to an individual to get them to read it without changing the meaning? It could not only consider the person but the time of day, device, etc. https://lnkd.in/eaR9_PMG

Run LLMs at home, BitTorrent-style 🤓 Petals 🌸: Decentralized inference and finetuning of large language models Checkout - https://lnkd.in/gdgPvD96 https://lnkd.in/gzcAiWrC #llms #Decentralizedfinetuning #Decentralizedinference #genai #distributedsystems

GPT-4o, new flagship model that can reason across audio, vision, and text in real time. That's increadible. And at this situastion we see that, data science will thriving brisikly. Please look the video you will understand about it. https://lnkd.in/eFR2sKBp

Register here → bit.ly/4ay8yGd What is the right path for enterprises to build Gen AI applications on your own data? Join us on April 25th for a virtual wine-tasting session, where we’ll focus on strategically leveraging large language models(LLMs) using a corpus of proprietary data, uncovering how enterprises can utilize their vast data repositories to train and fine-tune LLMs, enabling a better understanding of domain-specific contexts and providing more accurate insights. This exclusive session will be hosted by Thomas Gibson, Solutions Engineer at SingleStore. Ronita Mullick La-i-matti War Nongbri Senjuti Ghosh Saket Bengani Sandeep Sivaram Mitch Speers Ashutosh Prasad #AIIntegration #LLMInnovation #DatabaseSelection #EthicalAI #AIIntegrationChallenges #EnterpriseAI #LanguageModels #CollaborativeInnovation #DataIntegration #PrivateAI #AIandEthics #InnovativeTechnology #TechIntegrity

Register here -bit.ly/48Uq9aR Tired of generic AI solutions that don't understand your unique business? Join us on March 7th for an exclusive virtual roundtable where we'll explore the exciting world of building Generative AI applications powered by your own data. In this interactive session, you'll discover: · The strategic advantage of leveraging Large Language Models (LLMs) with your proprietary data. · How to unlock the hidden potential of your vast data repositories. · Practical techniques for training and fine-tuning LLMs for domain-specific insights and superior accuracy. Our expert panel from SingleStore will help you chart the right path for building powerful, data-driven AI applications. Don't miss this opportunity to gain a strategic advantage in the age of AI! #artificialintelligence #machinelearning #dataanalytics #database #languagemodels #dataintegration #generativeai #aiintegration

Can the performance of a generalist model like GPT-4 be improved for a specific domain without relying on domain-specific fine-tuning or expert-crafted resources?

Cohere's Command R+ is now available on DevinAI to help accelerate enterprise AI adoption. Command R+ achieves unparalleled accuracy in natural language understanding and generation, ensuring reliable and precise results for your business applications. Tackle your enterprise workloads with Cohere's state-of-the-art RAG-optimized LLM. #Cohere #Commandr+ #RAG #LLM #Enterprise #DevinAI

#ICYMI ▶️ The recording of the seminar “From Text to Map: A System Dynamics Bot for Constructing Causal Loop Diagrams” with Niyousha Hosseinichimeh is now available on our website! In this webinar, Niyousha explored the capabilities and challenges of a novel tool designed to automate the creation of causal loop diagrams (CLDs) from textual data. The System Dynamics Bot aims to automate the process of constructing CLDs from text, leveraging large language models and generative artificial intelligence. 🎥 Watch now: https://ow.ly/iOaZ50SC8m0 🔗 Upcoming Seminars: https://ow.ly/BE9R50SC8m3 #SystemDynamics #systemsthinking #SeminarSeries

Join me in welcoming Gemma, the latest addition to the family of lightweight, state-of-the-art open models from Google 💎 According with the documentation, Gemma 2B and Gemma 7B are built from the same research and technology used to create the Gemini models and are text-to-text, decoder-only large language models available in English, with open weights, pre-trained variants, and instruction-tuned variants. You can explore more about Gemma and access quickstart guides in the link below 👇 #google #ai #gemma #open #models

Different types of #artifticialintelligence are out there with many claims about what they can do for clinical practice. But not all AI is the same. Not all AI is prepared to handle the technical complexities of healthcare environments. This is a cautionary tale in collectively understanding the difference in the midst of a storm of AI hype. I wanted to share this interesting tweet below that has some implications on the use of generative AI in medical diagnostics, with this concern: the tasks most large language models (LLMs) are trained on don't necessarily align with the precision required in clinical AI. For example, consider the difference in complexity between identifying references to a cat in a text and detecting a subtle brain hemorrhage in a CT scan. The latter is a task of immense precision and subtlety, which might require detecting a subtle change in a 15 pixel needle in a 100 million pixels hay stack. Recent advancements in AI, particularly in processing larger context lengths enable these large models to start processing haystacks, and the question we are asking ourselves is whether they are well suited to find needles there, which is a precondition to solving the problems we are dealing with. Researchers have recently led experiments in exactly that domain — accuracy in subtle detection. An example is the 'needle in a haystack' challenge in the tweet below, where the goal was to locate a specific fact within a vast array of 128,000 tokens. Seems like ChatGPT has shown a low accuracy rate ranging from 0% to 50%, when the context length (haystack size) was increased beyond 60,000 tokens. This insight is particularly significant in the realm of clinical AI, where the haystack size is around 100 million (pixels). It suggests that to reach the desired level of diagnostic accuracy, we may need to significantly enhance AI performance or consider integrating other technologies. To clarify, I believe ChatGPT is achieving wonderful things and can play a supportive role in medicine, but we must come to terms with understanding what any kind of AI product can and cannot do as the preparation for wider adoption moves forward. I'm happy to hear what other AI developers and entrepreneurs and healthcare leaders make of this.