Bowling with Muppets & My TLDR on LA Summit Keynote

Dr. Ryan Ries here, back in action today after yesterday’s AWS LA Summit.

I figured this would be a great place to recap Dr. Matt Wood’s keynote with a little summary + my takeaways.

So, let’s begin.

AI momentum is still building

First off, let’s just take a moment to appreciate the sheer momentum behind generative AI right now. It’s mind-boggling to think about where we are today with AI versus this time last year.

Matt mentioned he hasn’t seen this level of excitement from customers in a long time, and I couldn’t agree more.

He also hit the nail on the head about something else — we're not even close to the inflection point with AI on the innovation S-curve.

The possibilities are virtually limitless, and Matt dropped a few stats about AWS AI/ML that I found particularly interesting.

• More machine learning workloads run on AWS versus any other cloud
• AWS launched 326 new AI/ML capabilities since early 2023
• Over 96% of AI/ML unicorns run on AWS

The main focus of Matt’s keynote was 5 Laws for AI success.

5 immutable laws of Gen AI success

1. The Foundation for AI is in the Cloud

No surprise here — training beefy foundation models takes some serious compute muscle. That's why you see companies like Anthropic, Stability AI, and Mistral AI all leveraging the AWS cloud.

2. Pervasive & Efficient Experimentation is Key

If you want to stay ahead of your competition, experimentation with AI is key.

Security with AI is a big concern for a lot of companies, but secure experimentation IS available.

Tools like Amazon Bedrock Studio make it easy and secure to prototype Gen AI apps without being an expert.

Folks at Mission have been playing around with Bedrock Studio quite a bit.

• First impressions from one of our Gen AI engineers, Ben H.: It’s pretty easy to get up and running / prototype ideas. It makes doing single-file and knowledge-based RAG easy to implement.
• And from our Technical Enablement Lead for Cloud & AI, Kelby E.: It's going to be good. It's not 100% there yet, but they're building the right thing. This gets people using Bedrock without having to do everything from the console, and it puts teams in a spot where they can focus on developing reusable components and functions that can be reused for other use cases. It’s like PartyRock but with more integration.

3. Mix & Match Models to Your Mission

There is no one model to rule them all. AWS offers the broadest selection of top-tier models so you can lean into each one's superpowers — whether its summarization, reasoning, code interpretation, speed, etc.

A question my team often gets is how to choose the right model.

Matt spoke about this a bit and how Bedrock has a model evaluation tool. You can even set your own parameters with the tool to measure against so you can truly find the best model for your use case.

I’m a big fan of Bedrock.

4. Consistency, Coherency, & Control of FMs Comes from Data

Hallucinations are a frustrating point with these models.

As Matt said, these foundation models are powerful and capable but can get distracted really quickly.

By grounding your models with your own high-quality data using tools like Bedrock Retrieval Augmentation, you'll keep those responses coherent and compliant.

The cool thing about AWS, too, is that you don’t need to trade security and confidentiality for innovation. They’ve built enterprise-grade security and governance controls directly into Bedrock.

Bedrock also has robust guardrail capabilities to put up big “DO NOT ENTER” signs for any topics or data that are off-limits. But guardrails aren’t just about blocking the bad stuff. They’re also about giving you and your teams the freedom to innovate with confidence. They help prevent your AI from going rogue and making up nonsense.

5. Remove the Muck with Amazon Q

70% of dev work is pure undifferentiated heavy lifting.

But with the AI sidekick Amazon Q, you can increase your productivity by 10-50X and focus on the important work that pushes your organization forward (AKA the fun stuff).

Our CTO Jonathan LaCour will be hosting an Amazon Q webinar pretty soon. I’ll be sure to add the details in my next newsletters but you can always find all of our upcoming events on Mission Cloud’s events page.

Lonely Planet’s AI use case

I was blown away by Lonely Planet's AI-powered travel buddy use case.

An app that doesn't just regurgitate stale wiki facts but truly understands what you like to do when you travel and makes recommendations based on your preferences.

By tapping into Lonely Planet's massive archive of travel guides, reviews, and insider gems, they're creating an AI concierge that can go off the beaten path.

It's not just about the top 10 tourist traps — it's unearthing those hidden hole-in-the-wall spots that make a trip unforgettable.

This is the promise of Retrieval-Augmented Generation (RAG) in action. By combining the vast knowledge of foundation models with the depth and nuance of proprietary data, we can build AI experiences that feel magically intuitive and insightful.

PGA is more than just golfing

PGA is harnessing AWS Gen AI to pique the interest of everyone, not just golf fans.

By ingesting a wealth of sensor data and content, they’re building interactive apps that go beyond simple stats and highlights. They’re generating real-time insights into your favorite player’s strategy.

And they’re predicting where your favorite players’ ball will end up before it lands (hello, new gambling paths).

Looking to the near-term future

As Matt wrapped up his keynote, he left us with a glimpse into the near future of AI.

From the rise of agentic systems that can tackle complex, multi-step tasks to groundbreaking research in hallucination detection and novel architecture paradigms, it's clear that the pace of innovation is only accelerating.

But perhaps most importantly, Dr. Wood emphasized that true success hinges on equipping organizations with the skills and resources to harness these powerful tools effectively.

Through initiatives like the Amazon AI Ready program, Bedrock Playground, and the AWS Generative AI Innovation Center, Amazon is empowering customers to not just keep pace with the AI revolution, but to drive it forward.

As we left the keynote hall, one thing was certain: the AI industry is quite an exciting place to be.

Until next time,
Ryan

Now, here’s our weekly AI-generated image & the prompt I used.

"Generate an image of a 40-year-old white male with brown hair bowling at a cool bowling alley in Los Angeles. The bowling alley has neon light signs that say Lucky Strike. The man is wearing a gray t-shirt and jeans, and he is actively bowling. He is surrounded by muppets who are waiting for their turn to bowl."