AWS re:Invent: Everything Amazon announced, from new AI tools to LLM updates and more

Amazon Web Services was held in Las Vegas for the AWS re:Invent event, which began on November 27th and runs through December 1st. In direct response to increased competition, Amazon has made a rapid-fire series of announcements and disclosures about what it’s been working on recently. Enabling cloud providers to respond to the rise of AI with attractive offers to their customers. AWS CEO Adam Selipsky’s keynote speech set the tone for the event, as he says AWS stands to defend its long-standing lead and deploy AI tools and services to remain the top large-scale cloud provider in the market. I made it clear that.

I understand that you may not have time to participate.So we’ll be covering this over the next few days, bringing you quickly with the biggest news as it comes out. All organized into easy-to-digest and skimable lists featuring local reports. Contributions from editor and cloud expert Frederic Lardinois and his Ron Miller team, as well as his broader TechCrunch team. Let’s go!

Friday, December 1st

The final part of AWS analysis

AWS re:Invent is coming to an end, but the impact of AI announcements and Amazon’s approach to this year’s conference remains. Ron put a pin in this year’s event with his TC+ exclusive revealing that Amazon has his sights set on one big company in particular at this year’s event: Microsoft. read more.

Thursday, November 30th

Q&A with Amazon CTO Werner Vogels

Ahead of Amazon CTO Werner Vogels’ keynote, Frederic sat down with Vogels to discuss the trends he’s seeing (generative AI, of course) and his predictions for next year. read more.

Wednesday, November 29th

neptune analysis

Image credits: AWS

Another new tool is Amazon Neptune Analytics, which combines the best of both graph and vector databases. Ron reports that this new service helps customers leverage vector search to analyze existing Neptune graph data or data lakes on S3 storage to find important insights. read more.

clean room ML

AWS Clean Rooms ML helps you unlock a variety of opportunities for generating insights.

Amazon has launched a privacy-preserving service that allows AWS customers to deploy “lookalike” AI models trained for one-time business-to-business collaborations called AWS Clean Rooms ML. It is a derivative of AWS’s existing Clean Rooms product, which allows AWS customers to share their own data with external partners to build, train, and build AI models. Can be deployed. read more.

SageMaker Hyperpod

Image credits: AWS

Amazon’s AWS Cloud division now offers Amazon SageMaker HyperPod, a new purpose-built service for training and fine-tuning language models at scale. Ahead of today’s announcement, Frederick interviewed his Ankur Mehrotra, his SageMaker general manager at AWS, who said: This provides tools to efficiently distribute models and data across clusters, speeding up the training process. ” read more.

AWS Titan Image Generator

Image credits: AWS

You’re right, Amazon is finally joining the ranks of other big tech companies in releasing its own image generator. Kyle reports that Titan Image Generator is now available in preview for his AWS customers, allowing them to create new images or customize existing images when given a textual description. . read more.

Tuesday, November 28th

Amazon Q: AI-powered chatbots

Amazon Q is Amazon’s new AI-powered chatbot. Image credits: Amazon

The big announcement of the day was Amazon Q, an AI-powered chatbot for AWS customers. During his keynote, AWS CEO Adam Selipsky described AWS as a way to “easily chat, generate content, and take action.” It all depends on your understanding of your systems, data repositories, and operations. ”Kyle reports that Q is trained on his 17 years of AWS knowledge to not only answer questions, but also understand the nuances of app workloads and run them in seconds. It also has features such as suggesting his AWS solutions and products for apps that don’t. read more.

Amazon Bedrock Guardrails

AWS Guardrails for Amazon Bedrock. Image credits: Amazon

The new Guardrails for Amazon Bedrock tool allows enterprises to define and limit the types of languages ​​that can be used in models. For example, Ron writes, defining topics that are outside the scope of the model prevents you from answering unrelated questions. read more.

New AWS Trainium Chip for AI Models

AWS Graviton4 and AWS Trainium (prototype). Image credits: business wire

Amazon announced the latest generation of chips for model training and inference (i.e. running trained models). Kyle writes that Amazon has already discussed his AWS Trainium2, which is designed to deliver up to 4x better performance and he is twice as energy efficient compared to the first generation of his Trainium. Masu. His second chip, announced this morning, is called Graviton4 and is aimed at inference. It is the fourth generation of Amazon’s Graviton family of chips (as indicated by the addition of a “4” to “Graviton”), and is different from Amazon’s other inference chip, his Inferentia. read more.

Amazon S3 Express 1 zone

Amazon has made a major update to its S3 object storage service called Amazon S3 Express One Zone, a new high-performance, low-latency tier for S3. Frederic reports that One Zone significantly improves the performance of his data-intensive applications such as AI/ML training, financial modeling, and high-performance computing. read more.

Three new serverless products

Amazon announced three new serverless products that make it easier to manage Aurora, ElastiCache, and Redshift serverless services. Ron writes: “Each of these options is serverless, so Amazon manages all your hardware in the background and provides you with just the resources you need, when you need them, and IT doesn’t have to deal with everything behind the scenes. -End administrative duties.” Read more.

Amazon One’s palm scanning expands from stores to security

Amazon One Enterprise. Image credits: Amazon

AWS has taken the lid off its new palm scan ID service that allows businesses to authenticate individuals when entering their physical premises. Paul reports that Amazon One Enterprise builds on the company’s existing Amazon One product, which the company debuted in 2020 to enable biometric payments in cashierless stores with Amazon’s own monitoring capabilities. I made it possible. Visitors to Amazon Go stores can associate their payment cards with their palm prints, allowing them to complete transactions by simply entering the store and waving their hand over the scanner. read more.

New thin client virtual desktop environment device

Image credits: Amazon

Amazon has launched a new $195 device that allows enterprise users to access virtual desktop environments such as Amazon WorkSpaces over the Internet. Sarah writes that the device is housed in Fire TV Cube hardware. This is a decision made by Amazon to leverage existing expertise in the retail giant’s division that makes streaming media players. The company explained that the decision to build the new hardware was based on feedback from a customer who wanted to reduce his IT spending by replacing his desktops and laptops with cheaper hardware. read more.