With its vast knowledge and unparalleled ability to understand natural language, ChatGPT has the potential to serve as the world's most knowledgeable assistant to improve research productivity - from topic development to manuscript revision. In this first session of a two-part series, we will explore the power of ChatGPT and how evidence-based prompt engineering can assist you in communicating with ChatGPT to enhance your research. We'll showcase best practices and provide demonstrations on how to write effective prompts to use ChatGPT for study topic formation, literature review, study design, data analysis, etc. Don't miss this opportunity to learn about this powerful tool and how it can revolutionize your research process! This presentation is part of the Brown School’s Artificial Intelligence in the Social Sciences Series.
With its vast knowledge and unparalleled ability to understand natural language, ChatGPT has the potential to serve as the world's most knowledgeable assistant to improve research productivity - from topic development to manuscript revision. More information on this second session of a two-part series is coming soon. This presentation is part of the Brown School’s Artificial Intelligence in the Social Sciences Series.
Moving beyond the contemporary debates centering on data and algorithmic bias, join Professor An as he discusses the philosophical and practical issues concerning AI ethics in the future. Does AI have sentience or awareness, and if so, do we want/need to treat AI as moral agents and offer them rights? How can we learn to coexist with AI (super)intelligence? Do we have a "third way” besides technological singularity (and resulting apocalypse) and trans-humanism (superhuman)?
Obesity is an alarmingly increasing global public health issue. As AI profoundly transforms our world, how can we use state-of-the-art AI technologies to prevent unhealthy weight gain? Join us as Prof. An introduces a diverse array of projects his research team has been working on, from soda tax-related social media sentiment analysis to automatic calorie/nutrition counting based on nut/candy photos, from evaluating chatbots’ responses to body image-related questions to misinformation in obesity-related news articles. This presentation is part of the Brown School’s Artificial Intelligence in the Social Sciences Series.
Algorithmic bias and data ethics long precede artificial intelligence. In this webinar, we will discuss the relationship between AI and the unique opportunities and challenges that it brings to today’s social equity issues. How can we promote responsible, ethical AI and mitigate algorithm-driven discrimination and biases? Why are AI democratization and citizen data scientists critical in realizing AI’s social value? How can you, or anyone, be part of this historic effort? This program is part of the Brown School’s Artificial Intelligence in the Social Sciences series.
Still think AI is just a buzzword or something only to be messed with by computer scientists? In this webinar, Professor An shares his own AI learning experiences as a social scientist with no formal training in computer science, showcasing that anyone can learn and apply AI in their daily work and study. Want to use AI to predict future events? Detect objects in images? Create a chatbot? Write a summary? Recommend a book? Thanks to open-source high-level APIs, state-of-the-art AI models are available at your fingertips for free. Only two questions remain: Can I afford not to learn? If not, how do I start? This program is part of the Brown School’s Artificial Intelligence in the Social Sciences series.
Artificial intelligence (AI), characterized by machine and deep learning, has swept into today’s society like a perfect storm, leaving no stone unturned. From pizza delivery to crime control, from creative arts to chatbot companions, from autonomous driving to auto-trading, AI is already present in many facets of our lives. Of the many views of the future of AI, what version should we believe, if any? Should we embrace it, or should we fear it? Join Brown School Associate Professor Ruopeng An for a discussion of the enduring myths and debates surrounding AI. This program is part of the Artificial Intelligence in the Social Sciences series.
Still think AI is just a buzzword, a black box, a terminator, or a gizmo only to be messed with by computer scientists? Dr. An will dismiss those misperceptions through his own AI learning experiences (as a social scientist with no formal training in CS) and showcase that everyone, indeed, can learn and apply AI in their daily work and study. Want to use AI to predict future events? Detect objects in images? Create a chatbot? Write a summary? Recommend a book? Thanks to open-source high-level APIs, hundreds of state-of-the-art AI models are available at our fingertips. Only two questions remain: Can I afford not to learn? If not, how/where can I start? Join us and have your questions answered by Dr. An.
Join us for a deep dive into the intertwining realms of artificial intelligence and social work. This presentation will cast a spotlight on the historical and philosophical foundations of social work, setting them against the canvas of technological advancements. Attendees will be navigated through the dynamic terrain of AI applications, emphasizing their resonance in social work education, research, and practice, all illuminated by compelling real-world case studies. The potential biases and harms AI could manifest in social work while offering solutions to protect the profession's integrity and values will also be addressed. Finally, we'll consider a future where AI seamlessly integrates with social work, posing pressing questions on the evolution of social workers' roles and safeguarding our unwavering commitment to social justice.
The COVID-19 pandemic in the U.S. led to nationwide stay-at-home orders and school closures. Declines in energy expenditure resulting from canceled physical education classes and reduced physical activity (PA) may elevate childhood obesity risk. Join Professor Ruopeng An for a discussion of his recent study estimating the impact of COVID-19 on childhood obesity.
Artificial intelligence (AI) is fundamentally transforming how the world operates. Social, behavioral, and health scientists with modern data science skills are in surging demand and highly valued in the job market. The Artificial Intelligence and Big Data Analytics for Public Health (AIBDA) Certificate at Brown School trains social, behavioral, and health scientists to master Python programming for data science, learn state-of-the-art machine learning and deep neural network models, and implement responsible, ethical AI applications to solve real-world social and health problems. If you are passionate about data science and its vast applications in health, social work, and beyond, join Prof. Ruopeng An’s talk to understand where to get started and how AIBDA may help you succeed in becoming an interdisciplinary data scientist. This presentation is part of our Brown School Curriculum Showcase.
A common method for adapting a pre-trained large language model (LLM) for a specific downstream task is fine-tuning, where the model's parameters are re-weighted through iterative backpropagations. However, as LLMs continue to grow in size, this method becomes increasingly computationally and time-intensive. In this talk, we will introduce some parameter-efficient fine-tuning (PEFT) alternatives, where techniques are employed that allow the tuning of LLMs by adjusting just a small fraction (~0.01-3%) of their parameters, while still achieving comparable levels of accuracy on downstream tasks. The talk will cover: (1) What is PEFT and why do we need it? (2) Common PEFT methods, such as LoRa, IA3, Prefix-tuning, and P-Tuning, covering their underlying concepts, intuitive understanding, and reported benchmark performances, and (3) How to implement them using the PEFT package with real-life datasets.
In this insightful presentation, we explore the Segment Anything Model (SAM), a state-of-the-art tool in image segmentation, seamlessly integrated with RobotFlow and YOLO to revolutionize image annotation. The talk delves into the fundamentals of SAM, highlighting its key features and diverse applications in image processing, particularly in enhancing medical imaging like lung CT scans. We emphasize the simplicity of fine-tuning SAM, leading to substantial improvements in medical analysis, and explore its significant impact on public health strategies, including the efficient delivery of COVID-19 vaccines. Through practical examples and case studies, the presentation demonstrates SAM's wide-ranging applicability and potential in healthcare and public health sectors.
Join us on an exhilarating journey through the world of computer vision as we delve deep into various tasks, annotation types, and the intricacies of both manual and AI-assisted labeling. Our focus will be on the powerful and versatile Computer Vision Annotation Tool (CVAT), an open-source marvel that's transforming how we handle image data. We'll navigate through the rich features of CVAT, shedding light on its practical applications and how its API can be leveraged for advanced projects.
The increasing demand for efficient data labeling is driven by the need for high-quality training datasets in machine learning projects. As the amount of data generated continues to grow exponentially, data labeling struggles to keep up becoming a bottleneck in many AI development pipelines. Automated data labeling serves as a fast, accurate, and cost-effective alternative. Advancements in AI and machine learning have made it possible to develop algorithms and tools that can automatically label data, reducing the manual effort required and improving accuracy. These innovations have directly contributed to the rise of automatic data labeling, making it a crucial component in modern data processing. In this talk, we will delve into the compelling reasons for the growing need for auto-labeling. We will then explore and demonstrate tools that offer capabilities for labeling both text and images efficiently and accurately. As we navigate the evolving landscape of data labeling, join us to discover how these tools can revolutionize the way we prepare data for machine learning, ensuring optimal model performance while saving precious time and resources.
Are you embarking on your data science journey or seeking to streamline your existing workflow? We invite you to an engaging introductory talk on Metaflow, a user-friendly framework designed to simplify the complex challenges of real-world data science. This session is tailored for newcomers to Metaflow or those curious about integrating it into their projects. Dive into understanding what Metaflow is and its unique position in the data science ecosystem. Learn why Metaflow is rapidly gaining recognition among data science professionals and explore its key features that set it apart as a valuable tool for any project. Experience a basic live demonstration showcasing Metaflow's simplicity and effectiveness.
Causal inference and discovery lie at the intersection of understanding how events influence one another and predicting the outcomes of interventions. Causal inference provides insights uniquely distinctive from traditional statistics and machine learning, which predominantly hinge on associational relationships. While conventional methodologies bring their strengths, the exclusive focus on associations often misses the underlying causal structures. This gap underscores causal inference and discovery’s vital theoretical and practical importance. In the rapidly evolving landscape of artificial intelligence (AI), driven by unprecedented advancements in computational power, there lies a unique opportunity. AI not only propels causal inference and discovery forward but also redefines its boundaries. The implications of these advancements are vast, especially in the domains of public health and policy. With modern AI-driven causal inference and discovery, professionals can craft more effective interventions, drive impactful policies, and shape the social fabric by addressing pressing public health challenges. This talk provides a basic introduction to AI causal inference.
Dive into the cutting-edge world of AI model training with our upcoming presentation on DeepSpeed Chat, a groundbreaking approach that drastically reduces the time and cost associated with developing sophisticated models like ChatGPT. We'll explore the standard workflow of AI language model training, delve into the nuances of Reinforcement Learning from Human Feedback (RLHF), and uncover the key costs of training Large Language Models (LLMs). Discover the innovative DeepSpeed Chat and the advanced ZeRO++ engine, which together enhance the efficiency and accessibility of AI model development. Join us for an illuminating session complete with a live demonstration, and witness firsthand the revolutionary impact of DeepSpeed Chat on the future of AI.
With the rapid rise of Large Language Models (LLMs), many are eager to apply these powerful language models to their personal datasets. However, challenges arise, particularly when dealing with extensive documents, due to token limits in models like GPT-3, while fine-tuning can be resource-intensive. Additionally, user data often comes in diverse forms, ranging from PDFs to videos. LlamaIndex tackles these challenges, offering a robust solution to ingest, vectorize, store, and query large heterogeneous data using the LLM of choice. In this presentation, we will outline the various strategies employed by LlamaIndex to index user data, including the List Index, Vector Store Index, and Tree Index, and detail how it performs queries over this indexed data. Demonstrations will include practical applications like semantic search, summarization, and querying various data types, such as PDF files and videos. These demonstrations aim to highlight how LlamaIndex can be a valuable tool for researchers, particularly in the fields of public health and health policies, by enabling the effective utilization of diverse datasets and documents.
Join us as we embark on an exciting exploration of ChatGPT's latest innovation, GPT-4, and its revolutionary beta feature, the 'Code Interpreter.' Building on the capabilities of its predecessor, GPT-3.5, which provided coding assistance based on user prompts, the Code Interpreter takes this functionality to new heights. It not only generates code but also introduces a novel feature allowing users to upload and download files, expanding the realm of interactive coding experiences. Moreover, it enhances user engagement with additional interface actions, such as the ability to analyze and summarize datasets in response to prompts. But what more can this advanced feature do, and how can we harness its full potential? Let's dive into the world of ChatGPT Code Interpreter 101 to discover the myriad ways it can transform our interaction with programming and data handling.
Discover the power of Pinecone, a state-of-the-art vector database designed for building scalable and efficient similarity search applications. In this talk, we delve into how Pinecone manages and searches high-dimensional vector data, enabling the creation of intelligent applications with fast and accurate nearest-neighbor search. We will explore its core functionality in storing and indexing vector embeddings, making it ideal for applications in recommendation systems, image and video search, natural language processing, and more. Learn how Pinecone can enhance your projects with its ability to efficiently retrieve similar items based on vector representations, revolutionizing content discovery and recommendation systems. Join us to uncover how Pinecone can be leveraged for semantic search using natural language processing, searching through dense vectors, and in question-answering and image-text embedding applications.
Dive into the world of Natural Language Processing (NLP) with our focused talk on HuggingGPT, a cutting-edge language model by Hugging Face that's reshaping NLP research. We're steering clear of the theoretical and diving straight into practical applications, showcasing how HuggingGPT can supercharge research across diverse NLP applications. Our discussion will include a range of useful case studies, from generating high-quality text and graph recognition to enhancing scientific research and sentiment analysis. We'll explore how HuggingGPT boosts efficiency in tasks like data annotation and model evaluation, and how its fine-tuning capabilities tailor it to specific research objectives. This talk goes beyond the basics, highlighting HuggingGPT's synergy with other AI models and tools for even more impressive results. Ideal for researchers, developers, or practitioners in NLP, this session promises to equip you with valuable insights and actionable knowledge, giving you a clearer picture of HuggingGPT's real-world efficacy and potential in tackling various NLP challenges.
Join us for a hands-on exploration of Gradio, the intuitive platform for creating machine-learning interfaces! Our speaker, Anqi Zhao, will guide you through the basics of Gradio, including how to install it using pip, define input and output types, and customize the appearance of your interfaces. We'll also delve into some of the more advanced features of Gradio, such as custom input and output types, live updates, and multiple inputs and outputs. You'll see examples of how these features can be used to create powerful machine-learning prototypes and interfaces that are accessible and user-friendly. You'll come away from this presentation with a solid understanding of Gradio and the skills you need to start building your own machine-learning interfaces. Don't miss out on this exciting opportunity to learn about one of the most powerful machine-learning development toolkits!
Discover the transformative power of DALL-E2 in our insightful talk, where we delve into the art of prompt engineering, a key technique for generating intricate and imaginative images using this advanced technology. We will guide you through the benefits of using prompt templates to enhance your image creation process with DALL-E2, helping you understand its capabilities and how it responds to complex prompts. This session is an opportunity to master the use of DALL-E2 for creating detailed visuals and to gain a glimpse into the future of image generation technology, setting you on a path to elevate your visual creations to the next level.
Explore the remarkable potential of AI language models in our engaging presentation, designed to unveil the power of prompt engineering. If you're intrigued by the capabilities of tools like ChatGPT, our session will take you a step further, demonstrating how prompt engineering can significantly enhance the results generated by these models. In this talk, we will delve into the basics, principles, and advanced techniques of prompt engineering, along with sharing best practices and discussing future trends. Our learning objectives are tailored to provide you with a deep understanding of prompt engineering fundamentals, equip you with practical techniques for your research, and offer insights into the future directions of this evolving technology, empowering you to leverage AI language models to their fullest potential.
Copyright © 2023 Rigor Answers - All Rights Reserved.
We use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, your data will be aggregated with all other user data.