Rajiv Shah’s Post

Mixture of Agents (MoA) combines multiple AI models to improve their overall performance. It uses several layers where each layer refines the output of the previous one, taking advantage of each model's strengths. This method leads to better accuracy and robustness compared to using a single model, achieving a score of 65.1 on Alpaca Eval 2.0, which is higher than GPT-4o. If people are curious, I'll write a more in-depth, technical description. https://lnkd.in/ezP_2ngJ

I just posted on this very subject earlier today! There are many options to build from and utilize for your specific needs. Keeping up is my job. Our whole team at NEXTELLIGENCE led by our expert GenAI creative and genuine human, Darien Davis and our Director of Artificial Intelligence, the seasoned and profoundly authentic, Dr. Uohna Thiessen strives to stay ahead of the wave with a firm finger on the pulse of the heartbeat of the needs of our market and clients. Teams like Hugging Face are helping us do that exceptionally well as a transparent and open company that focuses on value for the people, yielding massive returns in social capital... And if you don't know, they look pretty solid financially, as well. Open Source AI. One way we can use #aiforgood to help us live more, work less, develop sustainable healthy relationships, and #StayHuman. 🧡 #leadership #ethicalai #llm #consulting #nextelligence

The future belongs to small, specialized language models: they're easier, cheaper, and quicker to build, deploy and maintain. They also make more sense in many aspects than exploiting a tiny fraction of huge, expensive models. And Open just widely expands the possibilities. 

🌟 Nexus Raven 13B: The New AI Model Surpassing GPT-4 in function calling 🍺 Less than a month ago at the Amazon Web Services (AWS) x Hugging Face roadshow, Julien SIMON hinted that an open source model will surpass OpenAI GPT-4 in AI capabilities. He mention that hold his beer our models will surpass GPT-4 soon. 🚀🤖 Meet Nexus Raven 13B: This 13 billion parameter model is rewriting the AI playbook. It's not just matching up to GPT-4; it's outperforming it in zero-shot function calling. 👀 Why This Is Huge: Nexus Raven 13B is a testament to the power of open-source innovation. It's a David beating Goliath in the world of AI, all within a month of Simon's tease! 🌐 Impact: This is more than a victory for Nexusflow.ai It's a giant leap for the open-source community, reshaping our expectations from AI models. 🔮 What's Next: Nexusflow.ai Raven 13B is setting new benchmarks, showing us the limitless potential of AI. It's no longer just about coding; it's about partnering with AI 👏 Kudos to Nexusflow.ai! And as for that beer, Julien SIMON mentioned - here's to the future where AI turns our words into action! 🤗 🍺 #AI #NexusRaven13B #OpenSource #AWSRoadshow #huggingface

What excites me the most, other than the fact that multi agents are primed for success, is the fact that using agents are another way to outperform GPT-4 with cheaper models, this time it's GPT-3.5-turbo. There's a factor of 60 difference in cost! (for input token $30 vs. $0.5) https://lnkd.in/dRYnSkKc

Open models now starting to surpass GPT4 for specialized tasks. Let's go! Model by Nexusflow: https://lnkd.in/deAPbUzJ Leaderboard: https://lnkd.in/dx-fQHe5

Absolutely, it's exhilarating to witness the evolution of open models! The progress made by Nexusflow's model exemplifies the dynamic nature of AI development, particularly in specialized domains. These advancements not only push the boundaries of what AI can achieve but also highlight the importance of continuous innovation in the field. It's a compelling reminder of how targeted efforts can lead to breakthroughs that may surpass even well-established models like GPT-4 in specific areas. This progress underscores the exciting potential of AI in solving complex, specialized problems and shaping future technologies.

This is huge and something to keep an eye on... Open models run the potential to be deployed in-house, behind a fire wall, to train on proprietary data, and advance use cases at much more controllable (and potentially lower) costs... But the bar to succeed is still high... - Technical Integration - Fine tuning - Business Integration - User Experience

🔥GPT-4 function calling performance surpassed by this fairly “small” model. 👉 Why is this important? Function calling is important for extending large language models with information from various sources. 👉 Historically, open source models have been quite bad a deciding the what the right function to call is, which is why this release is exciting!

Open source will prevail