Powered by Data - July 31, 2023

Data Points

Last Week in Conversational AI

• Amazon has expanded its Bedrock platform, enabling generative AI-powered apps to develop using pre-trained models. The latest addition, Agents for Bedrock, allows customers to create conversational agents that deliver personalized answers and execute actions based on proprietary data. For example, this could include customer service chatbots that use internal information to customize orders. The Agents utilize prompts to determine steps and knowledge to use, and they can access third-party models to enhance their functionality. Amazon's expansion in this field indicates its seriousness about generative AI, despite a rocky start for Bedrock.
  Read more
• OpenAI has launched the ChatGPT mobile app for Android following its release for iOS in May. The Android version offers similar features, including voice input through OpenAI's Whisper speech recognition model, allowing users to transcribe spoken prompts into text. However, voice commands to submit prompts and a text-to-speech tool for reading the AI's response are unavailable.
  Read more
• Microsoft has introduced TypeChat, a structured conversational AI library for developers. Based on TypeScript types, the library allows developers to define types representing different intents or questions that the AI application can handle. TypeChat then constructs prompts to the Large Language Model (LLM) AI, such as OpenAI's ChatGPT, asking for data in a manner that conforms to the defined types. If responses fail validation, it will attempt to repair them through further interaction and summarize the interaction before taking action.
  Read more
• AI startup Stability AI has announced the release of Stable Diffusion XL 1.0, a text-to-image model that promises improved color accuracy, contrast, and lighting compared to its predecessor. The model is now available on various platforms, including Stability AI's API, GitHub page, and consumer applications, Clipdrop and DreamStudio. SDXL 1.0 focuses on image generation, bringing photorealism to AI-powered image generation, particularly face generation. It supports "image-to-image" prompts, allowing users to input an image and add text prompts to create detailed variations.
  Read more