🚀 New Native 2023: A Year of Breakthroughs and Innovation in AI 🌟
As we bid farewell to 2023 at New Native and lablab.ai, let's take a moment to reflect on the incredible journey of AI innovations and community growth this year! 🌐💡
🔥 Lablab Community: The lablab community flourished, introducing innovative programs like the Slingshot and GAIA Accelerator, which nurtures AI prototypes into market-ready products. The community's engagement in hackathons, project voting, and feedback forms underlined the collaborative spirit driving AI advancements.
💡 Future of AI: Looking ahead, we anticipate further integration of AI in various sectors, addressing challenges like workforce displacement and ethical considerations. With the ongoing development of platforms like Cohere, AI21 Labs, and Stability.ai, the potential for AI to enhance our lives is limitless.
🎉 Thank You! A big shoutout to everyone who contributed to this incredible year of AI innovation. Your creativity, hard work, and passion shape a brighter, AI-driven future!
🔮 2024 and Beyond As we step into 2024, we're excited to continue this journey with you, exploring new frontiers in AI and creating solutions that make a real difference. Stay tuned for more groundbreaking projects and collaborations!
🌟 Join Us: Be part of this vibrant community and shape the future of AI with New Native. Let's make 2024 even more extraordinary!
#AI#Innovation#TechTrends#CommunityGrowth#NewNative2023
🚀Join the Accelerate & Innovate 2023 Year-End Hackathon! Starting December 27th till January 10th! It's the grand finale of this year's tech journey. Don't miss out – register now: https://lnkd.in/dCEWxTRf and be a part of the ultimate Hackathon of 2023!
💡Got a project from a past hackathon? This is your chance to transform it into an innovative AI prototype. Utilize a variety of technologies that we have partnered with throughout the year to achieve significant market impact. Outstanding projects could qualify for the exclusive Slingshot Program! Innovate and make your mark!
🏆 I would like to announce that our project, Le ChatOn Vision 🐱 , won first place at #France's largest #hackathon in #Paris 🇫🇷 with over 250 participants last weekend, organized by Mistral AI and Cerebral Valley. Huge thanks to them and all the sponsors and judges who make this event possible.
🚀 Collaborating with my old gang Kim Borgen and Maureen Allanic from European Space Agency - ESA, we developed Le ChatOn Vision, an open-source tool enabling users to gain insights from endless video data collected by autonomous drones and other robotic platforms using natural language.
🖼 A self-collected video dataset in STATION F is divided into individual frames and processed by a Mistral-based visual LLM (VLLM) to generate scene summaries. For user queries, a pipeline setup uses the Mistral embedding model to find relevant summaries, and the Mistral LLM to consolidate responses based on these scenes.
🔎 Initial testing shows that while the VLLM generates broad qualitative scene descriptions, it struggles with detailed quantitative insights, such as the exact count of people in an image. To address this, a segmentation model is added to accurately identify and quantify objects, enhancing the VLLM summaries with precise details. This improvement allows Le ChatOn Vision 🐱 to accurately count people in some frames or provide the exact number of sodas in a vending machine, even in low-quality, motion-blurred images. This significantly enhances the performance of standard open-source VLLMs.
🤖 We demonstrated how current open-source LLMs, libraries and tech stacks make it accessible to build systems comparable to closed-source state-of-the-art models like GPT-4o in a very short timespan. While these systems do not outperform OpenAI's model in general, they excel in specific areas such as detailed visual interrogation, particularly valuable in various robotic applications.
💻 Tech Stack used:
- Nebius' cloud instance
- Neon's database solution
- Groq's deployed Mistral model powered by their LPU inference engine
💡 Models used:
- Meta Facebook' Segment Anything: https://lnkd.in/ebez8mD8
- #GroundingDINO: https://lnkd.in/eS-vVNfy
- Mistral Inference: https://lnkd.in/eeXwtpgJ
👨💻 Links:
- Codebase: https://lnkd.in/eQTywpMa
- Devpost: https://lnkd.in/e9HjqJ9e
Our Mistral AI Paris hackathon was a blast! Check out the highlights in this video recap.
Thanks Cerebral Valley for organizing and creating this awesome video.
How's your AI agent coming along? 🧑💻
Time's ticking, but it's not too late to jump into the MindsDB Hackathon and compete for the chance to win up to $5K! 💰
Find out how to get in on the action
→ https://hshno.de/mindsdb
Genai Hackathon provided an exceptional opportunity for me to delve into the intricate realm of artificial intelligence.
It allowed me to develop my skills and gain valuable experience in this remarkable field. 🚀👩🏻💻🤖
Announcing The Intergalactic Swarmathon - the first ever Multi-Agent Hackathon
Multi-Agent is the future of AI as it solve the most critical problems of single Agent including hallucination, context lost, task-focus, collaboration, and mismatched outputs. Welcome to join us for a weekend hackathon where you and top minds in AI gather to craft the very best Multi Agents with Swarms.
Time: Friday May 31, 6pm and Sunday June 1, 6pm PDT.
Location: the Agora Grand Auditorium: https://buff.ly/3ViVJdY
Agenda:
May 31, Friday PDT:
6:00 PM - Welcome and Workshop 1: Getting Started with Swarms
7:00 PM - Workshop 2: Orchestrating Agents with Swarms
8:00 PM - Announcing the Swarmathon: Build anything you want with Swarms. Rules:
Every project must be open-source with a Github link to be validated.
Project must use 2 or more agents
Must solve a real world problem
Can form team sizes of any length
Hangout in the forum to meet teammates
June 2, Sunday PDT:
5:00 PM - Showcase, Discussion, Q&A
See you then!
Senior Specialist in AI/ML | M.D Doctor
1moMistral care to give me the actual hardware to do it? With that... Ezpz