Closing the gap: AWS challenges Azure's apparent AI dominance

For the last year, AWS has been on the defensive. While Google and Microsoft have been announcing one AI innovation after another, Adam Selipsky has been in the hot seat, trying to answer why for one of the first times in AWS history the market leader is seen as being behind at a time everyone is “AI or die”.

There’s a raging debate if AWS is actually behind — if you ask Selipsky, these claims are untrue, in large part because of AWS’s gigantic infrastructure and many partnerships which dwarf their competitors. And any delay on their part (not that there is one) is due to a dedication to offering robust security, responsible AI, and a wide offering of models.

While this isn’t the same thing as having strong AI services, I have to agree: once AWS incorporated AI into the robust ecosystem of services they offer, any ground they’d lost would quickly be made up again.

So, did this happen at re:Invent 2023? The answer is a resounding yes. 

Is this a cloud conference, or an AI one?

AWS spent most of re:Invent announcing how they were innovating in the AI space. The biggest news was of course Amazon Q, a generative AI chatbot for workplaces that can be tailored to your business. It has all the strengths of ChatGPT, but you can plug it into data sources like Amazon S3, Microsoft 365, GitHub, Dropbox, Salesforce, Slack, and more. 

Q is a real game changer, and the applications are broad: call centers, programming, cloud architecture, business intelligence, content production — the list goes on. There were back-to-back keynotes just exploring all the potential uses of Q, and I’m sure this was only scratching the surface.

As expected, AWS is taking Amazon Bedrock — its new service for building generative AI applications with foundational models (FMs) — and kicking it up a gear:

• AWS is previewing Guardrails for Amazon Bedrock, a feature that helps put safeguards in place to filter or deny thorny user inputs or FM outputs. 

• There’s Agents for Amazon Bedrocks, which allows GenAI to perform multi-step tasks like answering customer questions.

• You can now privately and securely customize FMs in Bedrock with your own data. 

• You can now incorporate Bedrock with Step Functions, so you can orchestrate tasks to build generative AI applications using Amazon Bedrock, and integrate with over 220 AWS services. Instead of invoking an AWS Lambda function to use Bedrock, you can use Step Functions. 

You can see this wider theme of AWS pivoting to optimize what they already have in things like Amazon CodeWhisperer and DevOpsGuru, both services that use AI to improve programming and DevOps, respectively. In fact, it was no surprise to hear AWS was now offering AI-powered call summarization in Transcribe Call Analytics for customer service calls. 

The resounding message from re:Invent was “Is your data working for you?”, and this aligns with the increasing focus on data services — this makes sense, because like with Q, AI is going to be useful in leveraging that data to drive customer value. AWS wants to encourage people to use their data in more sophisticated ways.

Of course, many of the features AWS has announced are in preview or limited preview, and so the cloud provider has an audacious agenda for getting these things delivered. Only time will tell, but by the time AWS’s New York summit rolls around next year, it will be interesting to see how much of these are still in preview, or still being tinkered with behind the scenes.

AWS: We offer more than just OpenAI

Perhaps the biggest message (a veiled dig at Microsoft?) was diversity of choice. Amazon repeatedly touted how Bedrock offers a wide range of FMs to work with from AI21 Labs, Anthropic, Cohere, Stability AI, and Amazon. The message was even though Microsoft has OpenAI and GPT in their court, AWS has range, and can also leverage Anthropic’s reputation as providing more ethically designed models like Claude. 

Reaching the promised land of cloud computing through AI

A lot of organizations are consuming cloud infrastructure, and the many services that are available. The promise of cloud was that it would allow you to focus on customers instead of spending your time maintaining infrastructure, but the reality is that it is requiring a lot of undifferentiated heavy lifting. AWS is taking the approach that by incorporating AI into its existing services, it will help realize this promise, helping free up resources and people to spend on more interesting things. 

Towards the end of re:Invent, Dr. Werner Vogels, Amazon.com VP and CTO of Amazon.com delivered an honestly refreshing keynote speech on the importance of staying focused on cloud fundamentals. My takeaway was it’s important to remember that at the end of the day, while everyone is focused on all the bright and shiny AI announcements, you shouldn’t lose sight of the cloud services that provide 80% of your needs.