AI in the spotlight (again and again)
The much-anticipated keynote by AWS VP Data and AI, Dr. Swami Sivasubramanian, focused primarily on the key elements for implementing generative AI:
- The right model for the right use case: choosing AI models adapted to specific problems
- A private and secure environment for your business is essential
- Enhance model capabilities with your own data (the famous RAG): customize AI models with company-specific data.
- Clear governance over the use of these tools: establish a clear framework for the use of AI.
- Train, train and train teams 😊
At Microple, we embrace these principles and recommend them to our customers.
AWS provides an incredible range of tools for deploying Gen AI solutions, from infrastructure (with custom-designed processors) to Bedrock and Sagemaker.
Focus on Amazon Q
But the buzz of the moment is the announcement of Amazon Q(https://aws.amazon.com/q/), a versatile Gen AI assistant capable of generating content, summarizing text, answering questions, and more.
Integrated with AWS tools, Q lets you interact with services such as QuickSight, Glue, Redshift or the console itself using natural language, simplifying the design of cloud platforms.
For those looking for a customized solution, AWS Bedrock offers ease of use. Simply choose a model (Claude, Llama2, Stable Diffusion, etc.), drop your data into an S3 bucket, and let Bedrock process it, allowing the model to access the data to answer questions in a more targeted way.
We are witnessing the beginning of a major transformation that will profoundly and durably alter the way we work and interact with the digital world.
This fourth AWS Re:Invent 2023 day focused on cost management and sustainability.
The Keynote by Dr. Werner Vogels (CTO AWS) was eagerly awaited, and did not disappoint. We invite you to watch the video here: https://www.youtube.com/watch?v=UTRBVPvzt9w
Far from making all kinds of announcements, Vogel immediately emphasized the importance of taking cost and sustainability into account when designing a Cloud architecture. In particular, he introduced two key principles:
- Designing a Cloud architecture with cost in mind from the outset. He develops this paradigm in his online site, The Frugal Architect(https://www.thefrugalarchitect.com/).
- Align architecture longevity and related technical decisions with business objectives
In his opinion, an unsupervised architecture results in unpredictable costs, and he rightly pointed out that cost reduction and carbon frugality often go hand in hand when it comes to Cloud architecture. To help AWS customers better control the costs and sustainability of their Cloud infrastructures, Vogels announced the availability of the myApplications function in the AWS Management Console andAmazon CloudWatch Application Signals.
On the subject of AI, he also reminded us that generative AI was not the absolute answer to every problem, and that there were still many cases where traditional AI was perfectly suited to solving numerous use cases. He illustrated this by citing three examples:
- A moving testimonial from Thorn about its Safer solution for combating child pornography by analyzing media stored online.
- The International Rice Research Institute(https://www.irri.org/) selects and saves seeds to guarantee secure access to food.
- Cergenx, a company that has developed an AI-based medical device to identify newborns most at risk of brain damage
A few sessions were also given on the environmental dimension of the cloud, explaining how AWS is improving the carbon footprint of its activities to align with the Paris Climate Agreement, and the principles and tools at our disposal to develop eco-responsible solutions.
Zero-ETL was also on show once again, with the presentation of promising new data sharing features between Salesforce and Redshift (based on Salesforce DataCloud), or the ability to create customized AI models using SageMaker or Bedrock and deploy them in Saleforce (two preview features announced at Dreamforce last August).