Last Week in AI: Microsoft's GPT-4o Update, Google's DataGemma, Cloudera's AMPs, More -- Pure AI

Last Week in AI: Microsoft's GPT-4o Update, Google's DataGemma, Cloudera's AMPs, More

By Pure AI Editors
09/16/2024

This edition of our roundup of AI products and services announced last week that didn't make headlines, but should be on your radar includes Microsoft's GPT-4o update, Google's DataGemma series, Cloudera's AMPs, DorothyAI's AskDot, and more.

Microsoft announced the release of the GPT-4o-2024-08-06 API, available through the Azure OpenAI Service. This new AI model introduces significant advancements, the company said, including the highly anticipated JSON Structured Outputs feature, aimed at boosting developer productivity and application precision. This model is designed for real-time applications, the company said, offering faster text generation with minimal latency. With 100% reliability in generating outputs that match the defined JSON Schema, it is ideal for diverse use cases, from customer support chatbots to advanced analytics tools. Developers can define the desired output format using a JSON Schema, ensuring consistent, structured data. This feature reduces the need for post-processing and offers up to 50% savings on input costs and 33% on output costs compared to previous models. The GPT-4o-2024-08-06 API is now accessible in all US regions and

Google introduced the DataGemma series, a lineup of language models designed to deliver more accurate responses to numerical queries compared with previous algorithms. The series is based on the Gemma 2 27B model, a 27-billion-parameter LLM using the Transformer architecture, and leverages Google’s Data Commons, a repository with more than 240 billion data points from trusted global sources such as the UN, WHO, and CDC. DataGemma will expand the capabilities of Gemma models, the company said, by harnessing the knowledge of Data Commons to enhance LLM factuality and reasoning using two distinct approaches: RIG (Retrieval-Interleaved Generation) enhances the capabilities of the Gemma 2 language model by proactively querying trusted sources and fact-checking against information in Data Commons. RAG (Retrieval-Augmented Generation) enables language models to incorporate relevant information beyond their training data, absorb more context, and enable more comprehensive and informative outputs. The source code is available on Hugging Face.

Cloudera, provider of a hybrid platform for data, analytics, and AI, unveiled a set of new Accelerators for ML Projects (AMPs), designed to expedite enterprise AI adoption. Cloudera's AMPs are end-to-end machine learning (ML) based projects that can be deployed with a single-click directly from the Cloudera platform. Each AMP encapsulates industry-leading practices for tackling complex ML challenges with workflows that facilitate seamless transitions, no matter where enterprises are running examples or deploying data, the company said. The latest AMPs and updates include: Fine-Tuning Studio, a comprehensive tool for managing, fine-tuning, and evaluating large language models (LLMs); RAG with Knowledge Graph, which demonstrates how to enhance retrieval-augmented generation (RAG) applications using a knowledge graph for deeper context and relationship mapping; PromptBrew, an AI-powered assistance for creating high-performing prompts via a simple interface; Chat with Your Documents, an enhanced accelerator that improves LLM chatbot responses by incorporating context from user-uploaded internal knowledge bases. Cloudera’s AMPs are open source, enabling deployment in any environment, and providing data scientists with ready-to-use AI solutions.

WitnessAI, the creator of a secure AI enablement platform, has announced integration support for Microsoft 365 Copilot and GitHub Copilot within Visual Studio Code applications. This allows businesses to monitor and manage Copilot usage without the need for an agent installed on employee devices, the company said. WitnessAI provides organizations with visibility into prompts and responses within Copilot, applying policy-based blocking, redaction, or routing as needed. Unlike typical DLP products that require browser plug-ins, WitnessAI operates at the network level, detecting and managing shadow AI activity across browsers, BYOD, and native applications like Copilot. This platform ensures compliance with corporate AI usage policies by tracking and safeguarding employee interactions with AI tools, such as Microsoft Word and GitHub Copilot, preventing the potential exposure of confidential data or intellectual property. WitnessAI is currently in beta testing and available for demonstrations.

DorothyAI, a provider of AI-driven intellectual property (IP) management solutions, announced the launch of AskDot, a GenAI tool for patent research and analysis. Billed by the company as the ultimate digital assistant for optimizing IP and R&D workflows, AskDot's global patent database is coupled with multilingual document support, bringing a new level of inclusivity to patent management and expanding its utility to international users, the company said. Users can search for specific patents, compare multiple patents, or summarize complex patent data in different languages with a few clicks. AskDot is available via subscription and integrates with the company's flagship product suite including Novelty, Freedom, CrossCite, People+AI, and InsightView. Together, these tools streamline and enhance the efficiency of IP management processes.

GrubMarket, a leading food technology company, introduced GrubAssist Charts, a new AI-powered data visualization tool designed to offer real-time insights across the food industry. This latest feature in the GrubAssist platform allows users to create dynamic charts and graphs using natural language commands. The new GrubAssist Charts features include: AI-Powered Visual Reports for transforming complex datasets into actionable charts with real-time insights on KPIs, such as inventory, pricing trends, and customer demand; User-Friendly Interface that allow users to generate customizable charts with natural language inputs, selecting attributes such as chart type, time interval, and data formatting without any setup; Multi-Platform Access, for seamlessly creating, viewing, and sharing charts on both desktop and mobile devices, with options to download or share visual reports across platforms; Real-Time Data Updates, which refreshes saved charts instantly with the latest data to keep visual reports current and accurate. The new GrubAssist Charts feature is available to all GrubAssist customers.

AtScale, a leading provider of semantic layer technology, announced the private preview of its Natural Language Query (NQL) feature, which allows business users to ask questions in plain English. This new capability eliminates the need for SQL knowledge or complex data engineering, the company said, providing users with direct insights from BI reports and custom queries using AtScale’s Semantic Layer and Generative AI. Building on a 92.5% accuracy rate in Text-to-SQL tasks, AtScale’s NQL feature aims to help organizations process large data sets more efficiently by translating natural language queries into complex SQL commands. The tool was designed to reduce query costs, accelerate data retrieval, and simplify decision-making by bypassing traditional data engineering methods. John Langton, VP of Engineering at AtScale, said the preview marks a major advancement in business data interaction, combining large language models (LLMs) with AtScale’s Semantic Layer for unparalleled accuracy. Enterprises interested in the private preview can contact their AtScale representative or request a 14-day free trial.