So let’s review AWS re:invent, which occurred in Las Vegas this week. Here are some highlights. There are a lot.
Amazon Web Services (AWS) has repurposed its Fire TV Cube streaming device into a desktop productivity device called the Amazon WorkSpaces Thin Client. Priced at $195, the device is designed for enterprise customers to offer to their employees. By reusing the hardware design of the Fire TV Cube, AWS aims to provide a low-cost solution for virtual desktop services. The WorkSpaces Thin Client uses cloud-based storage and processing, and feedback from beta testers has been positive. The device is being marketed to industries such as healthcare, financial services, and contact centers.
Amazon announced three new serverless offerings: Aurora Serverless, Elastic Cache Serverless, and Redshift Serverless. These offerings aim to simplify database management, improve response times, reduce costs, and optimize data warehouses using AI.
Amazon has released its Titan text-to-image AI model, which can generate realistic images with built-in guardrails against toxicity and bias. The model is aimed at an enterprise audience and requires access to Amazon Bedrock. All images generated by Titan will include invisible watermarks to indicate they were created with AI. Amazon has also created an API to detect the watermark, allowing developers to choose how to provide that information to users. Additionally, Amazon announced the availability of other Titan models and extended copyright indemnity to customers using its foundation models.
AWS has enhanced its Amazon Transcribe platform with generative AI, allowing transcription in 100 languages and introducing new AI capabilities. The system was trained on millions of hours of audio data and uses self-supervised algorithms to recognize patterns in human speech. The improved language recognition also benefits the Call Analytics platform, providing better accuracy and summarizing interactions between agents and customers. AWS also announced additional features for Amazon Personalization, including Content Generation for thematic connections in recommendation lists.
Amazon has introduced Clean Rooms ML, a privacy-preserving service that allows AWS customers to collaborate on AI models without sharing proprietary data. Clean Rooms ML enables the creation of “lookalike” AI models by training them on a small sample of customer records, which can be expanded with a partner’s data. The service also offers controls to customize model outputs and plans to add settings for healthcare applications in the future. Additionally, Amazon announced Clean Rooms Differential Privacy, a service that provides aggregate insights while protecting customer data.
AWS has introduced Amazon One Enterprise, a palm-based identity service that offers accurate and secure enterprise access control. The service combines palm and vein imagery for biometric matching, achieving an accuracy rate of 99.9999%. It eliminates traditional authentication methods and can control physical and digital access. Customers and partners like Boon Edam, IHG Hotels and Resorts, Paznic, and KONE already use Amazon One Enterprise.
The headline is Amazon Q is an AI-powered assistant to help users use their cloud services best. It provides expertise specifically for AWS and can assist developers and analysts in various tasks such as building projects, debugging, and suggesting resources. It is also available for business users, offering help with data analysis and creating narratives. Pricing starts at $20/month for the business package and $25/month for the developer edition.
Amazon plans to offer human benchmarking teams to evaluate AI models for bias and toxicity. This initiative aims to improve the accuracy and fairness of AI systems. The teams will provide feedback and recommendations to enhance the models’ performance.
Why do we care?
There’s a lot here, and different announcements will be useful for different listeners. I want to highlight two specifically.
First, Amazon is great at finding efficiency and repurposing, and there’s brilliance in using their Fire TV cube as a thin client. Being able to be produced at volume means stock issues are already managed, and they get scaled. Clever.
Second, Amazon has thrown its ring into the market with an AI assistant – and end users can’t purchase this one. Amazon’s approach is about a marketplace of LLMs, so there’s a unique angle to what they’re up to. I’ll tease that I’m talking to someone from Amazon for an upcoming episode.