April 21, 2023

208: Azure AI Lost in Space

57 minutes

Welcome to the newest episode of The Cloud Pod podcast! Justin, Ryan and Matthew are your hosts this week as we discuss all the latest news and announcements in the world of the cloud and AI. Do people really love Matt’s Azure know-how? Can Google make Bard fit into literally everything they make? What’s the latest with Azure AI and their space collaborations? Let’s find out!

Titles we almost went with this week:

Clouds in Space, Fictional Realms of Oracles, Oh My.

The cloudpod streams lambda to the cloud

A big thanks to this week’s sponsor:

Foghorn Consulting, provides top-notch cloud and DevOps engineers to the world’s most innovative companies. Initiatives stalled because you have trouble hiring? Foghorn can be burning down your DevOps and Cloud backlogs as soon as next week.

News this Week:

General News

@00:57 – Interesting article – What is Open AI doing that Google Isn’t

(Besides making a usable product, obviously.)

-Google AI lab is separate, meaning researchers are separate from the engineers, versus Open AI where they are one combined team, which – go figure – works out better. -The article goes on to question whether Google is “losing their edge” which, as the number 3 player in the AI industry, is pretty evident. The guys discuss the two services, as well as how Bard can be crammed into every product Google makes.

02:49 Ryan: “I find it kind of fascinating that Open AI, because they were first to market, gets to dictate what AI is.”

@07:01 Are you an AI developer? Are you looking to build out your own models? -Good luck. Finding the hardware to do that continues to be an issue.

The Information put out an article about a shortage of servers at all the major cloud companies, including AWS, Azure, GPC, and OCI. The biggest issue is a shortage of GPUs and GPU processors, which was one of the first and main resources to have supply chain issues.

Desktop computer GPUs are having less issues with supply. Some of that is thanks to the bottom falling out of the Bitcoin market (no need for mining anymore.)

07:57 Ryan – “It’s a run on a limited resource, and GPU’s – they were the first to hit supply chain issue… it’s always been sort of a scarce resource. When I first heard of GPU’s being used for machine learning and those types of workloads, there weren’t enough of them, and it wasn’t really embedded in the type of hardware you need to run in a data center.

09:07Justin – “A lot of GPU returns and GPU availability in the desktop market, which those GPU’s are better suited for doing high computational work of 3D and things that are required for getting to bitcoin… so you could use desktop GPUs but your experience won’t go as far.”

Unfortunately the smart British guy isn’t here to tell us all the ins and outs of the differences between types of GPUs, so do tune in for that next week!

@10:37 FinOps slack channels had some chatter in regards to the Amazon spot market pricing increases.

For the past couple weeks prices have continued to grow in US East 1, US AP Southeast 1A, and European servers (which are always more expensive anyway) among others. Justin discusses his ideas for why this is the case. Surprisingly (or not surprisingly at all) most of his theoretical reasons for these prices increases are pretty cynical – but they include capacity constraints in the supply chain, Amazon limiting additional buying because they’re going into earnings, and (most

...more

View all episodes

By Justin Brodley, Jonathan Baker, Ryan Lucas and Matthew Kohn

4.9

3434 ratings