Google Brain introduced Transformer in 2017, a flexible architecture that outperformed existing deep learning approaches, and is now used in models like BERT and GPT. GPT, a decoder model, uses a language modeling task to generate new sequences, and follows a two-stage framework of pre-training and fine-tuning.
The article discusses the importance of understanding context windows in Transformer training and usage, particularly with the rise of proprietary LLMs and techniques like RAG. It explores how different factors affect the maximum context length a transformer model can process and questions whether bigger is always better.
OpenAI has released an easy-to-use web tool to create custom AI assistants without coding, requiring only a Google or Microsoft account and a $20/month OpenAI Plus subscription. Users can personalize their AI assistant's name, picture, tone, and interaction style, and enhance its knowledge by uploading specific documents.
OpenAI introduces updates to ChatGPT AI models, addressing the "laziness" issue in GPT-4 Turbo and launching the new GPT-3.5 Turbo model with lower pricing. Users have reported a decline in task completion depth with ChatGPT-4, prompting OpenAI's response.
This article explores the limitations of using Large Language Models (LLMs) for conversational data analysis and proposes a 'Data Recipes' methodology as an alternative. The methodology allows for the creation of a reusable Data Recipes Library, improving response times and enabling community contribution.
Generative AI applications using large language models (LLMs) offer economic value, but managing security, privacy, and compliance is crucial. This article provides guidance on addressing vulnerabilities, implementing security best practices, and architecting risk management strategies for generative AI applications.
Developing LLM applications can be both exciting and challenging, with the need to consider safety, performance, and cost. Starting with low-risk applications and adopting a "Cheap LLM First" policy can help mitigate risks and reduce the amount of work required for launch.
The article discusses the singular value decomposition (SVD) algorithm and the author's process of refactoring the Jacobi algorithm from the GNU Scientific Library to Python/NumPy. The author validates their from-scratch SVD function using the np.linalg.svd() function and highlights the usefulness of SVD in classical statistics and machine learning.
GeForce NOW levels up PC gaming on mobile with higher-resolution support on Android, offering immersive gameplay on the go. New games added to the library include Stargate: Timekeepers, Enshrouded, and Metal: Hellsinger.
The UK's top intelligence agency warns that malicious cyberactivity will increase with the incorporation of AI, with ransomware being the biggest threat. AI will lower barriers to entry, allowing both novices and experienced threat actors to exploit vulnerabilities and bypass security defenses more efficiently.
This article explores monocular depth estimation (MDE) and its importance in computer vision applications. It provides a walkthrough on loading and visualizing depth map data, running inference with Marigold and DPT, and evaluating depth predictions using the SUN RGB-D dataset.
Google has announced Lumiere, an advanced text-to-animal AI video generator capable of creating realistic videos of cute animals in various scenarios. Lumiere's unique architecture allows it to generate the entire temporal duration of a video at once, achieving global temporal consistency.
Data science enthusiast builds successful AI web app serving thousands of users, sharing technical tips and lessons learned along the way, including using Next.js, FastAPI, and AWS Lambda for deployment, and Firebase + Stripe for user authentication and payments.
NVIDIA introduces RTX Video HDR, using AI to transform standard dynamic range video into stunning HDR on HDR10 displays. The new GeForce RTX 4070 Ti SUPER offers enhanced performance for video editing and rendering, running up to 2.5x faster with DLSS 3.
Generative AI solutions are transforming businesses, but they can produce factual inaccuracies. A Retrieval Augmented Generation (RAG) pipeline, using an AI-native technology stack, offers accurate, transparent, and secure generative AI applications by providing additional information from an external knowledge source.