Pitcrew AI’s Post

The next wave, or the next great chapter if you will, of AI will be all about the development and deployment of AI agents. These agents will enable users to interact with complex systems comprising multiple multimodal models and integrated tools, capable of executing tasks autonomously on their behalf, rather than relying on a single AI model. I am honored to be part of a team driving this significant advancement in AI technology. If you find today’s AI systems impressive, and I do, prepare for the transformative innovations on the horizon. It is going to be mind-boggling. I personally believe Ida, the world's first AI Drilling Engineer, will play a significant role in democratising groundbreaking technologies in the energy sector, as well as bridging the generation gap, and much more. #AI #AIAgents #greentech #ReinforcementLearning #CCUS #generationgap #LLM

I gave a talk about implementing DeepMind's groundbreaking AlphaZero algorithm from scratch to play perfect Connect Four! In this talk I cover: - The revolutionary approach DeepMind took to achieve the world's first superhuman performance at Go - The elegance of Monte Carlo Tree Search (MCTS) - How the Multi-Armed Bandit Problem underpins modern reinforcement learning - Implementing ResNet neural networks for board game mastery - Leveraging Rust's power to simulate millions of games in parallel I'm blown away by the idea that genuine intelligence can emerge purely through experimentation and iteration, without human guidance! Hope you find it as interesting as I do!

For Gemma users, can you explain the difference from RIG and RAG? https://lnkd.in/dm7qU4Sm #kaggle #gemma #AI #GenAI

🛰 Recent research by Amazon Web Services (AWS) could bring a new wave of advancements in geospatial analytics, significantly improving commercial applications like land use classification and segmentation (ref - https://lnkd.in/gq6qJuva) The paper proposes a foundation model that leverages three remote sensing data types - optical (RGB and multispectral imagery), microwave (SAR) and elevation data (digital surface models). This results in a significant improvement over current single sensor-trained models in capturing a more holistic and comprehensive representation of the Earth. I believe this work can greatly benefit the development of vision-language models for geospatial applications in the future. This is because contextually relevant descriptions of multi-sensor observations can enable complex question-answering which requires reasoning about information on different characteristics of an area. Risk assessment is an exciting application - for eg. identify areas with high vegetation density that are also prone to flooding. Prediction, however, will require datasets spread out across time, as noted in the paper. Platforms like Picterra can bridge that gap with temporal information and change detection models. Find below a demo of its platform, showing detection of deforestation over years. #computervision #geospatialanalytics #remotesensing #satelliteimagery #visionlanguagemodels #artificialintelligence #generativeai #ai

🚀 **Introducing Nsight StrataVision!** 🚀 We’re excited to unveil **Nsight StrataVision**, our cutting-edge application that leverages advanced machine learning to optimize drilling performance and operational efficiency. We’re excited to unveil Nsight StrataVision, our cutting-edge application that leverages advanced machine learning to optimize drilling performance and operational efficiency. 🔍 With Nsight StrataVision, you can: Predict Drilling Performance: Get highly tailored predictions on key metrics like Weight on Bit (WOB), Rate of Penetration (ROP), torque, and RPM, allowing you to make smarter, data-driven decisions. Real-Time Adjustments: Monitor live data to make real-time parameter adjustments, ensuring maximum efficiency and minimal downtime. Comprehensive Insights: Gain a granular view of critical drilling parameters and adjust your strategy based on in-depth analysis. Proactive Planning: Forecast drilling timeframes and outcomes for future wells to streamline operations and optimize resource usage. Whether you're aiming to improve well completion, reduce risk, or benchmark performance, Nsight StrataVision gives you the tools to succeed. Check out the video to see how Nsight StrataVision can transform your drilling operations! 📊⚙️ #machinelearning #drilling #oilgas #dataanalytics #NsightStrataVision #AI #drillingoptimization

"How much is the lice?" is a song not yet created by my favourite artist, Scooter, but it is a question up there with "to be, or not to be". In Scooter's absence, fortunately there are researchers putting their expertise to work in order to find what methods of lice counting provide the highest quality estimates. Our friends at Institute of Marine Research (IMR), Norway have developed a scientific method of counting lice, ensuring that as few lice as possible fall of during counting, and that the counted fish are as representable for the whole as possible. They have also used OptoScale's cameras to do automated lice counts in parallel, and found the following: 🐟 "Normal" manual lice counts are not good estimates of the real lice count. These are the kinds of lice counts all fish farmers are mandated to do every week. 🐟 Scientific lice counts of at least 30 fish or more give a good estimate of the real lice count: the industry should rather adopt this method than manual counts, if one wants to do manual counts at all 🐟 OptoScale's automated lice counts correspond well with the scientific lice counts, meaning it is safe to transition the industry to a reality where lice counts are performed using cameras, avoiding any handling of fish in the process. We are very happy to have participated in several projects where we are trying to discover the best way to count lice together with strong scientific partners, providing valuable insights to both us and the industry. We are ready to contribute to many more, in our journey to create the best possible solutions. And for all thinking that lice counting using AI sounds easy, I can assure you it is not! But when you have people like Ragnhild Skjetne Hollup and Ingar Stian Nerbø leading the efforts, incredible things can be achieved. Article in Norwegian, but it tells the whole story well: https://lnkd.in/d_pYdQXg

Reason #262847988 why I love Clay..... Claygent 🌈 🤠 The most advance AI web scraper, summarizer, navigator and more! I had to classify a list of 1000 companies as Investment Managers or Fund Managers - a task that would take a human about 10 hours. Naturally, I put it in Clay and asked Claygent to classify them for me while I went for lunch 🥐 🥙 Not only did it classify them, but gave me a confidence score and reasoning! Drop your Claygent use cases below ↘ #Claygent #WebScraping HANGAR49

Done & Dusted 🚀 Red Bull gives you wing ❤️ I just completed Red Bull's On-Premise Sales on Forage. In the simulation I: Completed a job simulation focused on On-Premise Sales activities at Red Bull Used Excel to visualize which on-premise clients are performing best and worst. Communicated observations to Red Bull team members about why certain clients are performing better than others. Utilized active listening and social proof to overcome a client’s objections resulting in them more prominently featuring Red Bull in their bar. Check out the simulation here: https://lnkd.in/dg2mJzUV #AI #Deepfake #Innovation #ComputerVision #EthicalAI #DataScience #MachineLearning #ArtificialIntelligence #BigData #DataAnalytics #DeepLearning #DataVisualization #DataScientist #DataEngineering #PredictiveAnalytics

#firemdpi #highlyviewed 🔥Fire and #Smoke Detection Using Fine-Tuned YOLOv8 and YOLOv7 Deep Models 👉https://lnkd.in/dsCRrRSu #firedetection; #wildfires; #deeplearning; #objectdetection