Savana Hughes - Bokksu | LinkedIn (original) (raw)
New York, New York, United States
250 followers 236 connections
Experience & Education
Bokksu
View Savana’s full experience
See their title, tenure and more.
View Savana’s full profile
Other similar profiles
Explore more posts
- Scale AI LLMs have become more capable with better training and data. But they haven’t figured out how to “think” through problems at test-time. The latest research from Scale finds that simply scaling inference compute–meaning, giving models more time or attempts to solve a problem–is not effective because the attempts are not diverse enough from each other. 👉 Enter PlanSearch, a novel method for code generation that searches over high-level "plans" in natural language to encourage response diversity. PlanSearch enables the model to “think” through various strategies before generating code, making it more likely to solve the problem correctly. The Scale team tested PlanSearch on major coding benchmarks (HumanEval+, MBPP+, and LiveCodeBench) and found it consistently outperforms baselines, particularly in extended search scenarios. Overall performance improves by over 16% on LiveCodeBench from 60.6% to 77%. Here’s how it works: ✅ PlanSearch first generates high-level strategies, or "plans," in natural language before proceeding to code generation. ✅ These plans are then further broken down into structured observations and solution sketches, allowing for a wider exploration of possible solutions. This increases diversity, reducing the chance of the model recycling similar ideas. ✅ These plans are then combined before settling on the final idea and implementing the solution in code. Enabling LLMs to reason more deeply at inference time via search is one of the most exciting directions in AI right now. When PlanSearch is paired with filtering techniques—such as submitting only solutions that pass initial tests—we can get better results overall and achieve the top score of 77% with only 10 submission attempts. Big thanks to all collaborators on this paper including: Evan Wang, Hugh Zhang, Federico Cassano, Catherine Wu, Yunfeng Bai, William Song, Vaskar Nath, Ziwen H., Sean Hendryx, Summer Yue 👉 Read the full paper here: arxiv.org/abs/2409.03733
- Jhonathan P. R. dos Santos One of my core initiatives right now is to scale the processing of multi-modal agricultural data as cost-effectively as possible, without incurring technical debt. The ability to process trillions of data points, with a system that minimizes side effects and leverages the best in programming, is critical for the success of AI in agriculture. It’s essential to understand that AI science differs from data science. In AI, scaling laws must be considered, and we need to leverage knowledge from evolutionary biology to simulate DNA recombination based on sequencing technologies. Additionally, geospatial data must be approached at multiple levels: on a macro scale using satellites, drones, and aerial imagery, and on a micro scale with proximal sensors. This is not a short-term, 6-month sprint project; it requires a long-term, iterative approach. Early releases may fall short of expectations, but through incremental improvements and continuous iteration, we will reach the desired outcomes. I'm fortunate to have leadership within my organization that genuinely believes in our mission. Having a robust and efficient data and modeling infrastructure will be critical to igniting the next green revolution in Agriculture. #AI #Agriculture
- Mehdi Azad I've recently published a blog post that summarizes all these essential concepts. Each of the three sections—covering (1) SQL query commands, (2) CRUD operations, and (3) database design—can be read in less than 10 minutes. It's also a great review for those who are already familiar with these topics. If you're looking to expand your SQL knowledge or refresh your skills, I encourage you to check it out! As ML/Data Scientists, we usually write SQL queries to retrieve data from databases and load it into DataFrames, which we then manipulate using libraries like Pandas. However, SQL is more than just some queries; it's a way of thinking about data. This includes designing schemas, organizing data, understanding relational algebra, knowing how to use data execution engines, working with different file formats, and grasping the basics of data storage and partitioning. Mastering these concepts will make you a more effective user of tools like Snowflake and Databricks and significantly advance your career. #SQL #DataScience #MachineLearning #DatabaseDesign #CRUDOperations #DataManagement #Snowflake #Databricks #Pandas #DataEngineering https://lnkd.in/gZsr8BqU
- Boqor Abdalla 🚀 Unlocking the Power of Python & SQL: A Must-Have Resource for Programmers! 🚀In today’s fast-evolving tech world, mastering Python and SQL can be a game-changer. Pharell Hearst’s book, "Python Programming and SQL", provides an incredibly comprehensive guide for both beginners and seasoned developers. Whether you're diving into data analysis, building robust web applications, or optimizing databases, this book has you covered! 💡 Why this book is essential:Python: Learn how to write clean, efficient code while exploring libraries like NumPy and Pandas for data manipulation.SQL: Dive deep into database management, from basic queries to advanced concepts like joins and triggers. Real-world applications: The book teaches practical techniques for automating tasks, managing data, and building AI models. Combined Power: See how Python and SQL together unlock endless possibilities, from handling complex datasets to building full-fledged applications.This is more than just a programming guide – it’s a resource that helps you think like a problem-solver and create data-driven solutions. 🌟#Python #SQL #DataScience #MachineLearning #WebDevelopment #Programming #TechBooks #LearningJourney #CodeSmart #Automation #recognizeSomaliland.
- Vanshika Bansal Day 11: Meta's Llama 3.1 - A New Era in Open-Source AI 😎 Welcome to Day 11 of our 30-day series on Large Language Models (LLMs)! Today, we’re excited to highlight Meta's groundbreaking release of Llama 3.1, which is setting new standards in the AI landscape. 👉 The 405B Powerhouse **Llama 3.1 405B** is a revolutionary model, the first open-weights AI that rivals closed-source giants like GPT-4 and Claude 3.5 Sonnet. This development significantly narrows the gap between open and closed models, democratizing access to cutting-edge AI capabilities. 👉 Accessibility for All The **Llama 3.1 8B model** is a game-changer for consumer-grade hardware. It outperforms GPT-3.5 on many benchmarks while running locally and at no cost, empowering individual developers and researchers to innovate without expensive infrastructure. 🤘 Key Improvements Llama 3.1 offers several significant enhancements: ✍ 128K context length: Better handling of longer inputs for more complex tasks and extended conversations. ✍ Multilingual support: Usability across eight languages, making it more versatile and inclusive. ✍ Enhanced reasoning and tool use: Improved logical reasoning and effective use of external tools. ✍ Improved instruction-following and chat performance: More accurate and coherent responses in chat applications. 🗨 What This Means for the Future The release of Llama 3.1, especially the 405B model, marks a significant milestone in open-source AI. It promises to accelerate innovation, enable new applications, and push the boundaries of what's possible with locally-run models. check it out here : https://lnkd.in/gx8rvzJB Feel free to share your thoughts or ask questions in the comments. Let's continue the discussion and explore the impact of these innovative advancements together! #MachineLearning #DataScience #ArtificialIntelligence #LLMs #Llama3.1 #OpenSourceAI #Meta #AIResearch https://lnkd.in/gMnssCWY
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.