InfoQ Homepage Generative AI Content on InfoQ
-
Alibaba Releases Two Open-Weight Language Models for Math and Voice Chat
Alibaba released two open-weight language model families: Qwen2-Math, a series of LLMs tuned for solving mathematical problems; and Qwen2-Audio, a family of multi-modal LLMs that can accept voice or text input. Both families are based on Alibaba's Qwen2 LLM series, and all but the largest version of Qwen2-Math are available under the Apache 2.0 license.
-
Apple Unveils Apple Foundation Models Powering Apple Intelligence
Apple published the details of their new Apple Foundation Models (AFM), a family of large language models (LLM) that power several features in their Apple Intelligence suite. AFM comes in two sizes: a 3B parameter on-device version and a larger cloud-based version.
-
LLMs and Agents as Team Enablers
Eric Naiburg and Birgitta Böckeler published articles on the benefits and challenges of using AI as a multiplier in dev teams. We report on their insights for scenarios such as simplifying the germane cognitive load of a domain, automating code migrations, and coaching scrum masters on team facilitation. We also cover Böckeler's experiments with using LLMs to onboard onto a complex project.
-
MariaDB Introduces Open-Source Vector Preview, Aiming to Become Default MySQL Option
With the release of MariaDB 11.6, the MariaDB Foundation has announced the public preview of Vector search for the open-source fork of the MySQL engine. Database experts and open-source advocates see vector support as an opportunity for MariaDB to lead the MySQL ecosystem, especially since Oracle reserves most new features for its enterprise editions only.
-
Amazon MemoryDB Provides Fastest Vector Search on AWS
AWS recently announced the general availability of vector search for Amazon MemoryDB, the managed in-memory database with Multi-AZ availability. The new capability provides ultra-low latency and the fastest vector search performance at the highest recall rates among vector databases on AWS.
-
Mistral AI Releases Three Open-Weight Language Models
Mistral AI released three open-weight language models: Mistral NeMo, a 12B parameter general-purpose LLM; Codestral Mamba, a 7B parameter code-generation model; and Mathstral, a 7B parameter model fine-tuned for math and reasoning. All three models are available under the Apache 2.0 license.
-
Increasing Productivity by Becoming a Dual-Purpose Stream Aligned and Platform Software Team
To manage their increased workload effectively and maintain quality and efficiency, a software team decided to become dual-purpose: stream-aligned and platform. They rewrote their main application to be API-first and implemented micro releases with their customer-facing products, to provide value to their end users quickly and maintain a steady flow of accomplishments for the team.
-
AWS Announces a Generative Artificial Intelligence-Powered Service AWS App Studio in Preview
AWS App Studio, a new generative artificial intelligence (AI)-powered service designed to enable technical professionals without software development skills to create enterprise-grade applications using natural language, has been launched in preview by AWS in the US West (Oregon) AWS region.
-
AWS Introduces Amazon Q Developer in SageMaker Studio to Streamline ML Workflows
AWS announced that Amazon SageMaker Studio now includes Amazon Q Developer as a new capability. This generative AI-powered assistant is built natively into SageMaker’s JupyterLab experience and provides recommendations for the best tools for each task, step-by-step guidance, code generation, and troubleshooting assistance.
-
Redis Improves Performance of Vector Semantic Search with Multi-Threaded Query Engine
Redis, the in-memory data structure store, has recently released its enhanced Redis Query Engine. This comes at a time when vector databases are gaining prominence due to their importance in retrieval-augmented generation (RAG) for GenAI applications. Redis announced significant improvements to its Query Engine, using multi-threading to enhance query throughput while maintaining low latency.
-
Google Open Sources 27B Parameter Gemma 2 Language Model
Google DeepMind recently open-sourced Gemma 2, the next generation of their family of small language models. Google made several improvements to the Gemma architecture and used knowledge distillation to give the models state-of-the-art performance: Gemma 2 outperforms other models of comparable size and is competitive with models 2x larger.
-
Amazon Brings AI Assistant to Software Development as Part of Amazon Q Suite
Amazon has recently released Amazon Q Developer Agent, an AI-powered assistant that uses natural language input from developers to generate features, bug fixes, and unit tests within an integrated development environment (IDE). It employs large language models and generative AI to understand a developer's natural language request, and then generate the necessary code changes.
-
Meta's Chameleon AI Model Outperforms GPT-4 on Mixed Image-Text Tasks
The Fundamental AI Research (FAIR) team at Meta recently released Chameleon, a mixed-modal AI model that can understand and generate mixed text and image content. In experiments rated by human judges, Chameleon's generated output was preferred over GPT-4 in 51.6% of trials, and over Gemini Pro in 60.4%.
-
Generative AI Capabilities for Logic Apps Standard with Azure OpenAI and AI Search Connectors
Microsoft has announced that the Azure OpenAI and Azure AI Search connectors for Logic Apps Standard are now generally available, following an earlier public preview. These connectors are fully integrated into Azure Integration Services, providing developers with powerful tools to enhance application functionality with advanced AI capabilities.
-
Meta Open-Sources MEGALODON LLM for Efficient Long Sequence Modeling
Researchers from Meta, University of Southern California, Carnegie Mellon University, and University of California San Diego recently open-sourced MEGALODON, a large language model (LLM) with an unlimited context length. MEGALODON has linear computational complexity and outperforms a similarly-sized Llama 2 model on a range of benchmarks.