[00:00:00] Speaker A: Foreign.
[00:00:06] Speaker B: Welcome to the Cloud pod, where the forecast is always cloudy. We talk weekly about all things aws, gcp, and Azure.
[00:00:14] Speaker C: We are your hosts, Justin, Jonathan, Ryan and Matthew.
[00:00:18] Speaker A: Episode 288, recorded for the week of January 14, 2025. You might be able to retrain NotebookLM host to be less annoyed, but not your clap pod hosts. Good evening, Jonathan, Ryan. And that's it. Matt's not here.
[00:00:31] Speaker B: Hey, Justin.
[00:00:32] Speaker A: He's not even on the screen. Like I should know. He's not here. Of course not here.
He had sales kickoff, so I'm sure he's. He's struggling. Although Ryan and I were in a lot of Fedram meetings this week, too. That was.
That was mentally taxing. So I think we're all a little wiped out. But Matt just, you know, he has better children than we do.
[00:00:50] Speaker C: And. Yeah, sales kickoff. That's a lot of. He wins.
[00:00:53] Speaker A: Yeah, he wins for sure. There probably was more drinking because we had to move ours virtual because of the LA fires, which, you know, is a tragedy. And for anybody who's been impacted by the LA fires, I hope your family is safe. And if you're lost your home, you know, there's resources through the Fed, you know, through the Red Cross to reach out to, but terrible tragedy, and we'll be interested to see repercussions of that with the new administration starting this week. Yeah, that's gonna get handled.
Well, we don't have a lot of news. Still a little bit in the January druthers. Amazon's going back to the office, which apparently is a disaster with lack of parking and all kinds of problem people stealing stuff from people's desks. All the usual shenanigans that everyone forgot about because we've been working from home or hybrid for so long. I especially enjoyed in Seattle, Como News, which is one of the main TV stations there, had a thing about. Everyone noticed Amazon went back to work because traffic got way worse.
[00:01:45] Speaker B: Wow.
[00:01:47] Speaker A: Apparently, commutes jumped up by almost 40 minutes on average, crossing the bridges into Seattle. So, yeah, people noticed. Amazon went back to work.
All right, well, first up was ces. Was just happening. Of course. CES is the biggest electronic toy show in the world in Vegas, and it always is right after the New Year. So we're checking it out. We didn't see anything that was super exciting, although they did have some nice monitors that I might end up owning later this year. But there was a couple of good announcements from Nvidia I thought we should cover out. So first, if you don't want to hand over all your money to the cloud providers. You'll be able to hand over $3,000 to Nvidia directly for a computer that's probably going to be obsolete in less than 12 months. The new GB10 powered personal AI supercomputer called Project Digits will launch in May this year. According to Nvidia, the heart of Digits is the new GB10 Grace Blackwell superchip. It packs enough processing power to run sophisticated AI models while being compact enough to fit on a desk and run from a standard power outlet. Digits can handle AI models of up to 200 billion parameters and looks very similar to a Mac Mini, although it has this weird black and white speckling that I don't care for. Hopefully that doesn't make it into production. AI will be mainstream in every application for every industry with Project Digits. The Grace Black Hole super chip comes from millions of developers. Placing an AI superpeut on the desk of every data scientist, AI researcher and student empowers them to engage under shape the age of AI, says Nvidia CEO Jensen Huang in the Pressure press release. The Digit System comes with 120 gigs of unified coherent memory, so enough to run Chrome and up to 4 terabytes of NVMe storage. And for even more demanding apps, two digit systems can be linked together to handle more than 405 billion parameter models, which would be something like Llama 3.1 from Facebook.
The GB10 chip delivers up to one petaflop of AI performance, meaning it can perform 1,000,000,000,000, AI calculations per second.
And suppose you plunk down the money for Digits? In that case you also get access to Nvidia's AI software library including development kits or orchestration tools and pre trained models. While available through the Nvidia NGC catalog and the system supports a Linux based Nvidia NGC catalog and Linux variant and supports popular frameworks like Pytorch, Python and Jupyter notebooks.
[00:03:55] Speaker B: It's kind of pricey for what it.
[00:03:57] Speaker A: Is, especially considering it's not going to the GB10 will probably get replaced with a GB11 or 20 or whatever the next model is going to be.
[00:04:04] Speaker B: Well I think the Blackwell is pretty recent but you know it's the one they had a lot of problems with Yield and I I kind of suspect that they're they're sort of packaging this up and selling some of the some of the chips which didn't pass all the tests for the commercial products and so they're, they're enabling whatever cause they can in these things to sell to consumers. But it's not particularly, I mean having all the memory is really great for the big models. Not going to be particularly performant. Now I think the spec I saw was like one teraflop for quite low precision, like FB4 precision, which is quite low.
I think you'd be better off if you really were interested in it buying, buying some like 3090s or 50 90s or something like that. You obviously don't get the memory, but far better performance for the price.
[00:04:54] Speaker C: That's an interesting theory with the chips. I hadn't really thought of that. I thought they're just trying to capture a new part of the market.
But I mean I, I'm, I'm also like trying to figure out like, you know, as someone who's reluctant to, to get the bleeding edge of tech just because I'm too lazy to keep up with it mostly like I always, well, like how fast is this going to go out of date? And like in reality, you know, it's like yeah, it's not going to be the latest and greatest, but is it not going to be able to remodel soon? Does it have functionality there?
[00:05:23] Speaker A: Yeah, I know, yeah, it'll work for a while. My, my M1 MacBook Pro still runs models. Yeah, I mean not always fast, but it does. I can still run them locally if I want to. I guess that's my question is why would I buy this over just using a MacBook or a Qualcomm based Snapdragon windows machine that can run these models as well. Like especially if you're just doing local development, you're not going to be doing training of a model on your, on your desktop. At least I hope you're not because I don't even have that much data in my house to be able to power training of an AI model.
[00:05:55] Speaker B: I don't know. I think there'll be more and more use cases when that will be part of just operation of a consumer product. It could be like training for fine tuning for voice recognition or whatever, you know, faces, face recognition for your doorbell. There could be lots of reason why you'd want to have that kind of thing locally. I think there's a lot of people getting into this now because everyone's scared of their jobs and so everyone's trying to retrain into AI. And I misspoke earlier. It wasn't one teraflop, it was a petaflop which is still a third of one of the 5090s but with four times the memory. So I don't know, I guess it's, it's not a bad compromise, but yeah, still, still $3,000. And that, that's not the full spec on either. I'm pretty sure that's not with the, the 4 terabytes of SSD or anything else. They haven't announced the final pricing yet.
[00:06:41] Speaker A: No.
[00:06:41] Speaker C: I asked Claude earlier if I should.
[00:06:43] Speaker B: Get one and said no, I, I would believe Claude.
[00:06:47] Speaker A: I think Claude's on the right track there.
Also, interesting enough, at the investor day at CES Nvidia Jensen's Huang also hinted at plans for its own desktop cpu. Apparently it's the same CPU that is powering the digits computer, but they did not want to share a lot of data other than it is a co developed chip with MediaTek. MediaTek is well known for making HDMI processing units for TVs but they apparently also have a fab to be able to make ARM based CPUs and we'll see what those details are going to be out later in the year. That sounds like.
Can't wait to tell us more as they quoted to the investors.
[00:07:24] Speaker B: Can't wait to sell you more.
[00:07:25] Speaker A: Yeah, yeah, it'd be interesting to see the dominance of intel fall to the dominance of Nvidia and Nvidia just basically repeat the whole set of stuff all over again.
[00:07:37] Speaker C: Yeah, I was, I was thinking the same thing until I read it was ARM based. Right. So it's sort of like, okay, well I don't know.
I had nothing wrong with ARM based. It's like, but if someone could come in and be intel at their own game, you know, like then, you know, it's not just the limitation of, you know, the technology and getting as small as possible, but maybe it is and that's why ARM is taking off as much as it is.
[00:07:58] Speaker A: Yeah, yeah. It's really about processing speed and performance and being able to get, you know, the nanometer process down small enough. Intel is just screwed on x86 at this point. They can't get it where they need to be. And so, you know, you've reached the end of the X86 era. You can't go faster. So ARM is kind of the new, the new X86. And so eventually we'll reach the end of where we can go with ARM and something else will come out. I imagine at some point I can't, I don't know what that'll be yet, but something will come.
All right, well, in our Cloud Pod AI section tonight, Anthropic's Claude 3.5 sonnet is now in Snowflake Cortex AI. The general availability of this is available to you as the first Anthropic foundation model available in Snowflake's Cortex AI. Customers can now access the most intelligent model in the cloud model family from Anthropic using familiar SQL, Python and REST API interfaces. Snowflake security perimeter fully intact. So that's actually nice. I didn't realize that Snowflake was going to be making Claude available. I didn't. I missed the ea, but glad to see my favorite model is least available there.
[00:08:59] Speaker B: Is that them running the model or Anthropic giving them the model?
[00:09:02] Speaker A: They run the model.
[00:09:03] Speaker B: Cool.
[00:09:04] Speaker A: Okay, so that way it's in the perimeter.
[00:09:06] Speaker B: Yeah. That's nice.
[00:09:07] Speaker C: It is really nice. And you know, like, it's sort of one of those things where I read it, I'm like, oh, of course, duh, you know, like, why can I think of that kind of thing?
[00:09:13] Speaker A: But it's basically what they're doing with Amazon with Bedrock.
[00:09:16] Speaker C: So I mean, absolutely. It's just Snowflake has a ton of data and if you're, you know, like it makes a lot of sense if you're a business to you want to access that data, you don't really want to move it around and you probably have, you know, cloud connectivity set up already. Pretty cool. I like it.
[00:09:32] Speaker A: Me too. All right. AWS has new goodies for us in the AWS Compute Optimizer. They're now expanding to idle and Right size recommendations for your auto scaling groups with scaling policies and multiple instance types. With the new recommendations, you can take actions to optimize cost and performance for these groups without requiring specialized knowledge or engineering resources to analyze the performance and data. So machine Learning came for ASGs. Thank goodness for forever.
[00:09:56] Speaker C: Well, this is long overdue, right? Because you've always had. Or for a long time anyway. You've had optimizations for standalone EC2 instances, but ASGs have always been ignored and a huge amount of waste of people that, you know, set a minimum scale level for these things and they're just sitting there burning through coal but not taking any requests. So I'm glad to see these making the list.
[00:10:20] Speaker A: And yeah, I actually pointed it out. Auto scaling group that we purposely run at three nodes because we want to have AZ redundancy. And it was like, you could probably do a two. I guess I probably could.
[00:10:33] Speaker C: Yeah, you're right.
[00:10:34] Speaker A: I don't have the load needed for three. But we're just going to keep doing three because split Brain problems.
I appreciate that. I recognized something that I already inherently knew. So. Yeah, good.
[00:10:46] Speaker B: Yeah, I actually wish they built it into the auto scaling service itself rather than, rather than being just a report that says, hey, you need to adjust this thing. I wish they'd build it into the service itself and maybe there'll be some rules around it. Like your app has to expose some metrics or something so that it could consume that information, but then use machine learning to scale it appropriately. You know, look at the cpu, look at the memory usage, look at what your app's reporting in terms of latency or key length or whatever else and then actually put some smarts behind the scaling because it's called auto scaling. But it's not really, is it like it's automated scaling but you still have to put all the work in to say when to scale up, when to scale down.
[00:11:24] Speaker C: I think there's a little bit of logic there now, but it's still rudimentary. Like it's not, you know, like what we want now. We're getting it really spoiled by generative AI. Like just do it.
[00:11:33] Speaker B: I mean like CPU and memory have never really been good metrics to scale on. No.
[00:11:38] Speaker C: And I mean, and I think the logic that's there is, is time based as well. Like you can like, you know, scale based off of when, figure out when I'm going to serve, load and scale up to meet it, you know, kind of thing.
[00:11:49] Speaker A: Yeah, I mean I kind of walk themselves into a corner a little bit because they've basically anything that automatically scales up or down and based on machine learning, AI is called serverless. So I don't know how that works for EC2 Auto Scale Group Serverless.
[00:12:01] Speaker C: It's just Amazon taking a whole lot of liability. Right. For the performance of your or uptime of your app. Right. Yeah, it's in this weird thing, but if, but if you click the box and configure it, you kind of want it. So I don't know.
[00:12:13] Speaker B: Yeah, you can have separate sets of pools now, so you could always keep a reserve minimum pool of whatever and then you could use AI to do the smart scaling beyond a certain limit. So I don't know I would like.
[00:12:24] Speaker C: To see much more of that just to get beyond thresholds or simple thresholds.
[00:12:30] Speaker A: Well, Amazon has officially opened their AWS local zone in New York City. It's eating bagels, doing all the things as it provides to you a low wide range of workloads including EC2 instances include the C7i R7i M6i and M6i and EC2 instances EBS volumes are available for you there, as well as ECS, EKS, ALBs and AWS direct Connect, all available in the New York local zone.
[00:12:54] Speaker C: Is it walking here? It might be walking.
[00:12:56] Speaker A: It might be walking here.
[00:12:57] Speaker C: I don't know.
[00:12:59] Speaker B: Yeah, we missed a great opportunity to be like, hey buddy, get out of my local zone.
[00:13:03] Speaker C: Yeah.
[00:13:05] Speaker A: I'm computing here the same way.
[00:13:10] Speaker B: Are you gonna order EC2 instance or not?
Yeah, very friendly place. I love New York.
[00:13:15] Speaker C: Yeah.
[00:13:16] Speaker A: Yeah, I love New York. Like of all the major cities I've been to, it's one of the ones I always kind of finally looking back and be like, other than I almost broke my ankle there. But other than that minor issue, it was a great trip. I enjoyed myself and I need to go back and see much, much more of the city. But what I saw I really enjoyed.
Well, CEO Matt Garmin was on the excellent Decoder podcast with Nele Patel. He recently invited Matt Garman a year after he had Adam Slipsky on about stepping into the AWS CEO role. Matt hits most of the same talking points you've heard in the past. You know, things like that most companies are still barely in the cloud and they have huge market to grow and yada yada yada. Matt does talk about re entertain the computing infrastructure to support the evolving world of generative AI. And it was pretty clear from listening to the interview that Amazon is thinking about AI beyond just the models, but the monetization of service around the model, et cetera. Several other things that I thought was kind of interesting. He talked a little about AGI or artificial general intelligence, talked about Netflix as a customer and particularly around the Jake Paul fight and the NFL streaming that just happened. And was Amazon at fault for that and were they helping? And so there was some commentary that he politely declined to answer those questions, but did say he was texting support in the background like, hey, you're helping Netflix, right?
But overall it's a good listen if you want to hear kind of a new CEO coming into AWS and how he's thinking and what he's thinking about. Nothing really earth shattering, but still interesting to hear it from him. And Nilay was definitely excited to talk to him. You could tell just by his energy and the questions he was asking that he was excited to talk to the guy who was the original project manager for aws. So he has a lot of history at aws, which is interesting as well.
[00:14:56] Speaker B: That sounds good. I'll give that a listen.
[00:14:58] Speaker C: Yeah, I haven't listened to it, but I was excited by the notes just because, you know that I guess maybe it's just like a desire or desperation to hear something besides the normal like AI sort of narrative was going on. So like you know, the how many. How many parameters and how many model. You know, all these things. Like I'd like to hear more about the, you know, the usefulness or sort of like building it into products and that sort of aspect of AI versus more of the technical details of different models. So be kind of cool. I'll just see what his take is and see what, what his thoughts are.
[00:15:35] Speaker B: Yeah. What do you think they mean by reorienting the infrastructure to better suit AI?
[00:15:41] Speaker A: I mean basically building infrastructure services that support the needs of AI driven worlds. And we'll talk about a little bit later in an Azure story that will come up about AI first apps and what that's going to mean and kind of some of those things. But I think that's what he was referring to basically using as catchy as a phrase as Microsoft came up with.
[00:16:00] Speaker B: AI First Apps is a terrible, terrible name.
[00:16:05] Speaker A: Global First Apps, API Driven Apps Security First Security first driven. Everything's first. So nothing's first is how that works typically. But yeah, except for safety.
[00:16:16] Speaker C: Safety's there. Yep.
[00:16:19] Speaker A: Hola Mexico, the new central region for AWS is now open for business 11 months after they announced it in February of 2024. 2024. The region is now fully open with three availability zones and if you want to look it up in the API, it is MX Central 1. In addition to that, they reiterated their intent to invest billions of dollars into the Mexico market to help drive innovation and different things which, you know, might be a good idea with the upcoming presidential tariffs and things that, you know, Mexico might care a lot more about having their data in Mexico very soon. So we'll see.
[00:16:54] Speaker C: Cool.
[00:16:55] Speaker B: I wonder if the pricing is completely different in Mexico.
[00:16:58] Speaker A: You know, I haven't really seen disparate pricing too much in new regions. But yeah, it's not something I've been looking at either.
[00:17:03] Speaker C: So you, you don't see it in new regions. You see it with like when there's specific use cases like data transfer costs in and out of Australia and the APAC region. Right. Like there's but you know, I it used to be a lot more prevalent like you have to build your models for like forecasting, finance and stuff like with knowledge of that. But now every time I try to do it it's the same because I'm not, I'm not building anything.
[00:17:25] Speaker A: It was that really dumb idea they had about putting a data center in California.
[00:17:29] Speaker C: Right.
[00:17:29] Speaker A: That screwed them on pricing for power. Uh, but you know, I think everything else is at such commodities of scale that, you know, it's rounding errors for them a lot of the cases. But I think there was a couple other regions here. Japan I kind of vaguely remember had a little bit higher pricing and then maybe it was Brazil, the Sao Paulo region I think had slightly higher pricing for compute. And then yeah, everything else is really typically Internet connectivity is different depending on the region and the lack of dark fiber or undersea cables, et cetera in those areas.
Our last Amazon story. They're taking the cdk, which is the software development framework for defining cloud infrastructure and code and provisioning it through AWS cloudformation.
They're taking that and they're modifying because it currently consists of two primary components, the construct library that you use in a programming language to model your AWS app and a cli. The construct library synthesizes a model of your application to directory on disk and CLI reads that directory file to deploy your application on aws. Starting next month, the cdk, CLI and CDK construct library will no longer be released in lockstep and will be moved into different repositories instead. They will have their own independent release cadence, which means their version numbers are going to diverge and there were no impact to the CDK API or the user experience. They're doing this as they have matured the library. They have found that changes to the different components proceed at different paces and require different testing strategies. And this change gives them the ability to make those changes through these cadence of one subject project without affecting the other, giving the entire project more agility over time. So nice.
[00:18:56] Speaker C: Yeah, I mean it's. I'm sure there's reasons behind the scenes for this. It's sort of one of those things I was trying to think through.
[00:19:02] Speaker A: Like the dev team needed four pizzas and needed to divide it teams, so.
[00:19:07] Speaker C: Or it's like they do have four pizzas and then the 1:2 pizza team is like, hey, waiting for the other one all the time and they're sick of it. But I don't know, it's sort of weird.
So it's like. It also like there's cloud formation, there's all kinds of things, you know, that this is part of an ecosystem of. So it's probably fine. You know, it's kind of strange.
[00:19:31] Speaker B: It's the kind of thing I would expect them to do if they were also going to announce another new thing, like another like terraform support or something else like that.
[00:19:42] Speaker A: I support Terraform and cdk.
[00:19:43] Speaker C: Really?
[00:19:43] Speaker B: Terraform? I didn't know.
[00:19:45] Speaker C: Oh really? I didn't know that. Yeah, I thought it was just cloudformation.
[00:19:48] Speaker A: Yeah, you can do Terraform too.
Doesn't support Pulumi last I checked.
[00:19:54] Speaker C: Well, yeah, I mean it would be a direct competitor for Pulumi, I think, like they kind of offer the same thing. I mean, I don't know. I've never really. I've really tried over and over and over to get into the CDK model and it just doesn't work for me. And I think, I wonder if it's just because I was sort of a sysadmin that turned into a programmer over time, if it came from that direction or if it's just my utter hatred of Typescript.
[00:20:18] Speaker A: It's your utter hatred of Typescript?
[00:20:20] Speaker C: Yeah.
Like just fundamentally not compatible as a person with cdk.
[00:20:26] Speaker A: That's my issue with it. I, I've tried a couple times, you know, and I. Part of the Cairo. No cloudformation. So I'm like, you're not saving me any time because you're still outputting me cloudformation. I already know how to do that. So like, you know, you're not really saving me anything that I didn't know. When you get into like really repetitive tasks, like, hey, I need to create 100 launch groups for this really complicated architecture for some reason. Like I can see where that. The CDK is really interesting there because now you're using programming methods to basically do repeats and to do modifications based on certain parameters. I can see the value of that in that scenario.
[00:20:57] Speaker C: But I'd have a lot of questions about the format. Like why are you building a template with hundreds of.
[00:21:02] Speaker A: Yeah, I know. The problem is that what I'm saying to you makes no sense in an auto scaling world. But I can see if you have the need why that would make sense. And then. Yeah, I would hope you would smack me and be like, what are you doing, you fool?
[00:21:13] Speaker C: You're going to update all those at once.
[00:21:15] Speaker A: Okay, cool.
[00:21:18] Speaker C: That'll go well.
[00:21:19] Speaker A: You know, I'm a masochist in my coding, so.
[00:21:21] Speaker C: Yeah, yeah, it's just, it's. It's one of those things like I feel like this is. I know there are people that feel like, you know, terraform or. Yeah, configuration code is like manual labor and they want to automate it away. It's just I've never really seen the advantage and I only see the introduction of another abstraction layer where issues can come in.
[00:21:45] Speaker A: Agree.
All right, let's move on to gcp, who is starting to gear up already for Google Next.
We're shockingly only a few short months away. It is the beginning of April after all, and we're already midway through January.
[00:21:57] Speaker C: Had the rabbit. Yeah, exactly.
[00:22:00] Speaker A: I was looking at myself like I need to book travel soon.
So Google is getting ready by starting to launch their things to check out at Google Next blog posts. And so these will probably be happening for the next several months as they kind of hype up things. And the first one up was security. For all your security people who are going to be attending Google Next, they've got a bunch of stuff that they're doing for them that I think is pretty cool. So they have access to a dedicated security lounge in the expo where you can meet security leaders engineering Google cloud, secure by design, platform and products. So it just means there's a booth area that has security people. But I like they called it a lounge. Maybe there's a couch.
There'll be an interactive security operations center to see Google SecOps from the eyes of both the defender and the adversary. Mandy. And Threat Space where you'll learn from frontline defenders and incident responders. There'll be all kinds of overviews on securing your AI experience and talks that they're doing. They'll be doing a capture the flag challenge where you can test and hone your cybersecurity skills with real world data. Random notes and information from the dark web simulate a real world threat hunt. They have security tabletop exercise drills where they'll where you can role play and analyze aspects of hypothetical but realistic security incidents and burden of other sessions. Plus over 40 security breakout sessions. And for your CISOs, they have a dedicated programming track to equip CISOs and other security leaders with insights and strategies that they need to navigate the evolving threat landscape. So overall, lots of good security content coming up this year. The joy of having Mandiant now as part of the Google family. Security is a much bigger deal.
[00:23:24] Speaker C: Yeah, I'm really hoping with security lounge has like bouncers and metal detectors and.
[00:23:28] Speaker A: Probably like if you don't have security on your badge, like you're not get out of here. Like they do for the executive lounge, which you don't. They don't allow you into.
[00:23:37] Speaker B: I figured they were just curating a lounge for it so that everyone else could just stay away from those people.
Like here. Hey you guys. You guys, come over here.
[00:23:45] Speaker A: And the security lounge is actually located at the Mirage Hotel. Now for those of you who know anything going on in Vegas, the Marriage Rock Hotel is actually closed to become the Hard Rock Hotel. But that's where the security people are down there, away from all of us here at the Mandalay Bay. Yeah, I'm sure that's not the case, but I like the idea of it.
Let's go to Google is announcing the general availability of the Vertex AI RAG engine, a fully managed service that helps you build and deploy RAG implementations with your data and methods. Google's AI RAG engine allows you to adapt to any architecture from models, vector databases and data sources that work for your use case. Evolve your use case so you can add new data sources, update your models and or adjust retrieval parameters through simple configuration changes and evaluate in simple steps with different configurations find out what works best. For use case you can do AB testing. There's multiple other features available in the RAG engine, including DIY capabilities to tailor their solution to your solution to mix and match different components. There's search functionality, there's connectors for Google Cloud storage, Google Drive, Jira, Slack or local files, enhanced performance and scalability of Vertex AI search for you, simplified data management opportunities such as websites, bigquery datasets and cloud storage of sources and improved LLM output quality for your retrieval augmentation capabilities. And it's all customizable through customizable parsing and retrieval customizations available to you as well. And what I learned through Reese's article is that I really hate the name rag.
I didn't really like it before, but when they mentioned it a thousand times in this blog post, I was like, rag is just, it's like, I don't.
[00:25:14] Speaker C: Know, it's an ugly sounding word.
[00:25:16] Speaker A: It's an ugly sounding word. And I just, I don't know, I just don't like it.
[00:25:22] Speaker C: Yeah, I mean I'm, I'm looking forward to this because this is right where I'm stuck now in my ar. Journey is like, okay, this is cool. I like this tool. It does all these magical things. Then I want to do this specific thing with this specific data. And it's not something that you're training a model for. How do I make it do that? And that's not easy, right? Because the. And it's not really to do with AI. It's the same problem machine learning's had forever, which is how do you get access to the data in a safe manner, in a scalable manner that doesn't cost a million dollars and yeah, it's hard. Data's everywhere. Put it all in a data lake, do a ETL or you know, try to figure out how to, you know, use it in place. Which, this is really helpful. I'm hoping this fixes some of the issues that I'm running into right now.
[00:26:05] Speaker B: This must be really tough. I think being a service provider in this industry right now because things are changing so quickly. It's like, well, do we launch this vertex AI RAG product or do we wait three months? And this paper we just wrote about Titans, which is kind of like a slightly modified architecture that sort of separates episodic memory, like specific facts that you must remember as facts in themselves from the general training sort of pool of the network. And so that, that will help address hallucinations, it'll help ground things better because that's like they published this paper a few days ago.
This could really kind of get rid of RAG in a way because if you make the memory kind of pluggable then you could, you could train a very small model with, with your actual data and you would store that, saw that information. You wouldn't need RAG or wouldn't need as much rag. You could just like plug in like a memory module into a pre trained model and have it work. It's kind of cool.
[00:27:06] Speaker C: It's fascinating. But the stickiness of RAG and getting these components and integrations all in there too, like it would just naturally fit in there. So yeah, it doesn't stop you from moving on to the next greatest thing, Pastrag as well.
[00:27:19] Speaker B: So it's, it doesn't, but it's, you know, do you move now? Do you wait a bit? Do you invest it in time and time now and, and hope that it doesn't change fast enough?
[00:27:27] Speaker C: Like they started building this forever ago. Right. So they've already made the investment.
[00:27:32] Speaker B: Yeah, but as a, as a customer of Google's, like, well, what do you do? Do we start now? Do we wait six months? It's, it's pretty tough.
[00:27:39] Speaker A: I don't think there's any reason to wait on any of these projects. I think, I think the reality is that if you wait, you'll never get there because you're always waiting for perfection. And I don't think perfection is anywhere close at this moment. So yeah, you, you had to kind of move forward and solve for what you need today. And if the technology changes, just like you know, people did react or sorry, what was angular? And then angular became passe and we have remember to react and you know, I'm sure at some point React will be considered too buggy. And so there's another new web framework that everyone's going to be excited about. So it's, I think the same thing we're going through. AI is like, yeah, you, you know, you can build for today and you just need to keep evolving as you go. As long as you don't get stuck in one model and one idea, I think you're okay.
[00:28:22] Speaker C: Yeah.
[00:28:23] Speaker B: Yeah. I think we're almost reaching a point if we haven't reached it already where, where the, the, the latest a AI metal. I hate saying AI AI models are getting good enough that, that it's, it's, it's like Hitchhiker's Guy, you know, with deep thought creating the, the next computer that could actually solve the problem and it didn't. But I, I kind of get the feeling that, that Things like Sonic 3.5 is being used to generate good data, really good data to train the next OPUS model and then the OPUS model will be used to create the next silent model. And I'm sure Google's doing a similar thing and it's kind of like self improvement iteratively. Once we build a system that can improve itself then I think the pace of innovation is going to be huge. Which is the only reason I would think about pausing. Implementing something like RAG right now when there's potentially other hugely improved technology is like right on the horizon.
[00:29:18] Speaker C: Yeah, it's interesting because it's the, I guess it's all, you know, when you think about an investment, time and money like it's, I think it just varies on, you know, what you're trying to accomplish with the tool. I guess because it's, it does seem weird to stop, but. Yeah, especially as fast as everything's going now. Like I feel, I still feel like I'm just barely holding on by my fingernails to all the announcements and different types of things. Like I still haven't used the vast majority of models. I'm still trying to figure out how to make my one use case sort of work kind of, you know.
[00:29:54] Speaker B: Yeah. The problem is now, now we've got this power at your fingertips.
There's just so many things you could do like picking which one you're actually going to work on is, is the hard bit. Yeah, I, I feel like I've, I've become the constraint now. I've got all these wonderful tools available. I've become the constraint in the things I can work on because respond to one of them at a time. So I'm like really need to figure out how to replace the human part of these interactions. I did a bunch of work with Claude over the weekend and I kind of looked back at the conversation and I started off with a pretty solid idea.
98% of the conversation content is Claude responses, 2% is me writing stuff back. And it's not like analyzing what I was actually writing. It's like you redo this again or I think this could be done a slightly different way. Or I've tested this, let's move on.
[00:30:41] Speaker C: I think there's a mistake in your logic. Yes. If you go by the number of characters and who's writing what, I agree. But you're reading all of that and you are molding those ideas and having to respond back to TRA to tell the. I think that is the vast majority of the, the construct in that model, which you can't measure.
[00:31:01] Speaker B: It is. But I think that's, that's just, just think now you've got all the free users of Chat, GPT and Claude and oh, whatever, Deep Seek and everybody, everyone is there providing the human input, the human answers to these responses being generated. And that's going to be absolutely valuable in training the next kind of generation of models which do that for you.
[00:31:22] Speaker C: Yeah, I just think you always have to have something, you have to have like fresh water coming into the system at all times. Right. And I think that that thought and that prompt back is, is that right now? I, I, I don't know, like, it'll be fun because the minute you get rid of that, we really are kind of superfluous and we'll just get wasting oxygen at that point. So we'll see.
[00:31:40] Speaker B: Yeah, definitely.
There are a lot of cloud cost management tools out there, but only Archera provides cloud commitment insurance. It sounds fancy, but it's really simple. Archera gives you the cost savings of a one or three year AWS savings plan with a commitment as short as 30 days. If you don't use all the cloud resources you've committed to, they will literally put the money back in your bank account to cover the difference. Other cost management tools may say they offer commitment insurance, but remember to ask, will you actually give me my money back? Achero will click the link in the show Notes to check them out on the AWS marketplace.
[00:32:24] Speaker A: All right, let's move on to, uh, Google's automotive AI agent. They're apparently unveiling the Automotive AI Agent, a new way for automakers to create helpful generative AI experiences built using Gemini with Vertex AI. The automotive AI Agent is specially tuned to allow automakers to create highly personalized and intuitive in car agents that go beyond vehicle voice control. This allow you to ask via natural conversations things like is there an Italian restaurant nearby? As well as follow up questions like does it have good reviews? And what's the most popular dish at that restaurant? Mercedes Benz is among the first to implement the automotive AI agent in its M Bucks you virtual assistant. Coming to the new Mercedes Benz cla later this year. And I'll be avoiding all Mercedes Benz clas because if it can go beyond the driving voice control to driving, I don't want to be involved.
[00:33:10] Speaker C: Well, I keep thinking about like you know, the, the manufacturer, specifically GPS interfaces. Like that was a terrible choice because it was immediately out of date and not getting updates and then everything just shifted to a mobile device that you can keep up to date. And this is going to be no different.
[00:33:27] Speaker A: Why?
[00:33:28] Speaker C: This is not a good idea.
[00:33:30] Speaker B: No, I mean my key is 10 years old and I haven't updated maps I think in 10 years because they want $400 for map update.
[00:33:37] Speaker A: They want some crazy amount of money for map updates.
[00:33:39] Speaker B: Yeah, okay, sure. We could get it from BitTorrent probably. I didn't bother. Don't use it. Use a phone. It has basic voice controls but is like change a radio channel, do pointless stuff. Nothing to do with the actual drying bit, nothing to do with navigation. And I'm thinking like, I'm thinking for the Mercedes thing like so great you can ask these questions. You can only do that by just saying, you know, hey Google, hey Siri, then you.
[00:34:04] Speaker C: Yeah, it's, it's a weird thing and I'm sure it's a takeoff of that. I'm sure they're, they're thinking a little bit about the updates on that but because of the, you know, the, that experience with updating maps and the whole thing and it's not even trying to get it illicitly, like they're not created half the time. Like there, there isn't an updated version for my 10 year old car. Like it just doesn't exist.
[00:34:26] Speaker B: Yep.
[00:34:26] Speaker C: And it's you know, it's sort of a ridiculous thing that you know, it's like I guess, yeah, everything's, you can do over the air updates and you can do a lot different things than you could 10 years ago that are a little bit more ubiquitous. But I don't know, this isn't a selling feature to me at all. It's like just avoiding makes me want to not buy a Mercedes Benz.
[00:34:43] Speaker B: I guess the Advantage is going to be is okay, is there an Italian restaurant nearby? Yes, navigate there and then that integrates with the car's nav system or whatever and then puts the programs of routine or something. I can see that being valuable, but Google haven't even done integrated Gemini with phones properly yet. Like you, you can't do a lot with the AI. You can chat with AI for sure and Gemini's actually got pretty good. But it doesn't control things. Like it doesn't integrate nicely with Google home stuff. It doesn't integrate with anything. If you ask it to start a timer, it opens a web page with a timer on. If it close the webpage, the time is gone. You know, it's not the same as starting actual timer with the, with the old assistant, which was not natural language. Presumably we're not quite the same as the AI stuff. I don't know, like they're jumping ahead and investing in car stuff where they haven't even got, you know, they have the whole Android ecosystem with millions of devices and they've done nothing with that yet. So.
[00:35:39] Speaker C: I mean, Siri is finally getting to the point where it can actually do things on your phone versus just sort of like, I'm sorry, you'll have to unlock that. You know, like it's incredibly frustrating. And so it's like, yeah, until they really build in that level of stuff where the volume control really is sort of like an API driven microservice in your car. Which cool. Now the machine can do it, but can I, you know?
[00:36:05] Speaker A: Right.
[00:36:05] Speaker C: It's just like, yeah, well, you know, kind of thing.
[00:36:10] Speaker B: Wouldn't mind me saying this anyway. Especially not just because he had Google integrated chat.
[00:36:15] Speaker C: So Justin doesn't pay me enough to get a Mercedes.
[00:36:19] Speaker A: Fair enough.
Not that expensive. They're like 40 grand. Like, come on. Like, I mean, not a good one. I mean it's 40 grand CLA. The CLA actually was 44,000. I looked it up earlier because I was curious. I was like, how much is this car with gentic AI in it? I actually don't know if that's the right one or not, but yeah, anyways, all right. Last year Google released a Veo and Imagen 3 and creators use those AI tools to bring their ideas to life with the help of those models. Now Google is pleased to announce the latest version of Veo in VO2 and the latest version of Imagen 3, which did not become Imagen 4, which is just confusing. Both of which achieve state of the art results. These models are now available In Video Effects, Image Effects and their latest experiment, whisky, the VO2 can create high quality video in a wide range of subjects and styles. And head to head comparisons judged by human raters. VO2 achieves state of the art results against leading models out there. VO2 will deliver resolution up to 4K and extend it to minutes and length. And you can specify things like the lens to use, blur out background or focus on subject by putting a shallow depth of field all into the prompt. While many video models hallucinate unwanted details like extra fingers or unexpected objects, the VO2 produces these less frequently, making the outputs more realistic. Imagen 3 is improving, including brighter, better composed images. It can now render more diverse art styles more accurately, from Photorealism to Impressionism and from abstract to anime. Whisk is their newest experiment and it lets you input or create images that convey the subject, scene and style you have in mind. And you can bring them together and remix them to create something uniquely your own. From a digital plushie to an enamel pin or sticker. Whisk combines Imagen 3 with Gemini's visual understanding and description capabilities. Now, I tried to try Whisk three or Whisk here with Imden three because I was curious and only can make digital plushies, ENL pins or stickers.
[00:38:03] Speaker C: Oh no, oh no.
[00:38:06] Speaker A: Literally choose one of those three things and then what image would you like to use? And then here, here's your result, which I thought was. I'm like, well that's not very helpful, but I like the idea of what they're trying to do there. And then I did have a chance to check out Image FX earlier. I created a diagram that you might have seen in the chat room with an elephant in it. That was Image Effects.
[00:38:28] Speaker C: I was pretty impressed with how quick you whipped that out too.
[00:38:31] Speaker A: It also was terrible at words, which is my problem with all of these things. As I said, I gave it some very specific ones that I knew, like NetApp and a couple others. Then I said, then put other computer words in there and it made up a bunch of weird stuff after that. That was a mistake in my prompt. I just left it at what I gave it. But yeah, these are good. Blog post itself actually has a couple of examples. They did, they have a scientist basically, you know, going and looking at inside of a telescope microscope, some sample. And that's actually generated by VO2.
[00:39:01] Speaker B: That's fantastic.
[00:39:02] Speaker A: Which looks pretty good. And there's a couple others. If you scroll through different examples of things that it generated, they all look a little AI. Still, to Me too. So they also don't look real, but you know, for you being fooled for a brief moment, I could see it. Uh, but if you scrutinize it for more than, you know, two seconds, you realize pretty quickly it's AI just based on it.
[00:39:22] Speaker B: I don't know. I, I, I thought the, the lady in the yellow. I guess it's like a biohazard suit kind of thing. Looking in the Microsoft, that was really good. I think what gave that away is the teeth on the zipper.
[00:39:33] Speaker A: Mm.
[00:39:34] Speaker B: Kind of changed the. I thought the dog swimming underwater was probably the best one they had.
[00:39:38] Speaker A: Um, I, so the one that I, I thought was actually the best was the bees. So if you had little.
[00:39:42] Speaker B: Oh yeah, oh yeah, I saw the bees.
[00:39:44] Speaker A: And I think the reason why the bees work so well is because it's, it's a lot of moving objects, so you can't really focus on any one edge to notice that it's artificial versus where you look at the flamingos. The flamingos are very clearly too bright for the scene that they're in. And the lighting is not right.
[00:39:59] Speaker B: Yeah, the dark's pretty good though. Maybe the edges of the dog's ears aren't, aren't as smooth, smooth out as you would expect. But the bowls floating from the bottom kind of like the physics simulations of that and the light rays on the bottom of the pool, it's all pretty good. I think it would fool most people.
[00:40:15] Speaker A: Yeah, the, the, the coffee pouring and the pancakes are also quite good too.
[00:40:20] Speaker B: Yes, the physics of the pancake stack wobbling side to side is, is nice.
[00:40:25] Speaker A: Quite well done. So, yeah, definitely. I know we're on a podcast with audio only.
Definitely. Take a look at our show notes and click on the link here. As you can see, all the things we just talked about, but yeah, some of them are pretty convincing and you can definitely see the future is coming quicker than Hollywood is probably ready for.
[00:40:41] Speaker B: I requested access to it. Haven't got an answer back yet.
[00:40:44] Speaker A: Yeah, it's Friday they all holiday weekend.
[00:40:47] Speaker B: This was weeks ago. I need to like talk about our podcast a little bit more with them and tell them how our listeners would love to see us generate cool stuff to promote their products.
[00:40:55] Speaker C: Yeah, there you go.
[00:40:56] Speaker B: Yeah.
[00:40:58] Speaker A: Well, the UK CMA has announced that they will be assessing whether Google Search has strategic market status or SMS under the new digital markets, competition and consumer regime and what new requirements Google Search may need to follow. Google plans to engage constructively to lay out how services benefit UK consumers and businesses as well as trade offs of new regulations that they may impose on top of Google search. We'll keep an eye on this one. This would be probably a fun story because what Google wants and what the UK wants is probably completely different things.
[00:41:24] Speaker C: Yeah.
[00:41:25] Speaker A: And this will probably then to turn into an EU issue as well.
[00:41:28] Speaker C: Yeah, I did like the constructively language. Like, yeah, I get why they put it in there and I get what they're trying to say, but it's also like, are you sure?
[00:41:38] Speaker A: Yeah, I'm sort of. Sort of curious because this is what I think. Apple started here with a similar thing from the uk and then that turned into an entire EU issue and now they're owing like billions of dollars to the EU. So, yeah, we'll see how these things evolve.
TechCrunch had an interesting article about NotebookLM feature from Google and apparently they had to teach the Google LM podcast host to be less annoyed.
[00:42:04] Speaker C: So.
[00:42:05] Speaker A: So I missed this. In the December announcement, they added the ability to call into the podcast and ask a question, essentially interrupting the AI host while they're talking. So it's basically an interactive model. They're talking about a topic and you have a question. You can basically jump in and ask your question. When the feature first rolled out, apparently the AI host seemed annoyed at such interruptions and would occasionally give snippy comments to human callers. Like, I was getting to that. Or as I was about to say, which felt maybe a little bit adversarial.
No big. LM's team decided to do some friendliness tuning and they posted on X that friendliness tuning was the things I never thought I would do on my job but are a category of things. They tested a variety of different prompts and landed on a new prompt that is more friendly and engaging. TechCrunch reported that they tested the fix and said that it was working. And the host even expressed surprise coming. Whoa. Before politely asking the human to chime in, like, we have a guest. Like, whoa, we have a guest coming to the call. Is that like a service we can subscribe to? Is that a. I don't know.
I like the other way. Maybe we can have Notebook LLM call into us and ask us questions. That'd be good.
[00:43:04] Speaker C: Yeah, that's a good idea.
[00:43:05] Speaker A: Yeah, that's what we need.
[00:43:07] Speaker B: So, yeah, actually, can we make calls out from Riverside when we're recording? I know people can call in as guests. What if we can call out?
[00:43:16] Speaker A: I don't know. I mean, I'm sure you could hook it up with some other VoIP system and do I'm sure there's so much.
[00:43:22] Speaker C: More that we could be.
[00:43:22] Speaker B: Oh, we'll Totally call Japanese ChatGPT's 1-800-number.
Keep it on hold if you got any questions.
[00:43:34] Speaker A: That's right. Forgot about that just like two weeks ago and I already forgot. They had a call in number for ChatGPT. ChatGPT movie phone.
[00:43:43] Speaker C: That's right.
[00:43:44] Speaker A: Apparently an article here from I'd never seen before called Fierce Network, which apparently is a trade press that follows telecommunications companies. And their headline said Google Cloud could overTake Microsoft's number two cloud position this year. And I said, yeah, right.
But then I read through the article and I'm still not convinced. But they did put some interesting perspective on it. So analyst Jack Gold attempted to zero in on cloud hosting revenue for the big three hyperscalers. And he concluded that Google Cloud's pure cloud hosting revenue is likely much closer to Azure's pure cloud hosting than Microsoft wants it to be. In fact, he estimates it to be within a billion dollars of each other. He said at current growth rates, he predicts that Google cloud will be 55% greater than Azure within three years. So that would be shocking. He has all the data and all the evidence of why he did it. It's a bit of a work because Amazon and Google and Microsoft don't always publish the true actual pure cloud because they lump it in with Office365 or Dynamics licenses. And so you're trying to kind of like reverse engineer all of that with their earning statements, which is not going to be easy to do. So based on their best guess and what their investigations did, they think they got pretty close and that's why they think they're a billion dollars off. But interesting.
We'll see if they're right. I'll take this as a prediction from them for 2025 that I don't think is going to pan out, but maybe it will. We'll see.
[00:45:05] Speaker C: Yeah, I mean, I disagree with the time scale and if you extend the time scale out too much longer, you just have to assume everything sort of stays the same. And there's so many things that can change things, you know, like a. There's a. I'm sure there was a huge bump from AI for Microsoft, you know, a little while ago. And has that been really spread across the other cloud providers? I don't really know if they caught up. It's kind of yeah, then. And it's so not an apples to apples comparison because of the way these numbers are reported. It's sort of like, and I still don't see a lot of, I don't meet a lot of people working in GCP comparatively to Amazon obviously. And, but Azure, there's more people in Azure mostly because they have to.
[00:45:51] Speaker B: But yeah, where exactly do you draw the line then between you know, a managed cloud Service and a SaaS product like Office 365?
[00:46:04] Speaker A: Like, I mean this is a real question for Oracle right now.
[00:46:07] Speaker B: Yeah, like fundamentally, you know, really is, is offering Office360 as a service that you can subscribe to on a month to month basis any different than offering computers a service or anything else as a service?
[00:46:17] Speaker C: You know why? Because one I go to portal Azure.com and the other one I go to Office.
[00:46:23] Speaker A: So if they put Office 65 into the Azure portal, would you feel differently?
[00:46:28] Speaker C: Yes, I'd be like, oh crap, they, they screwed up my argument and I, and I feel they would have done that just to spite me and, and they would. Well, I would feel important. I would also be angry about.
[00:46:38] Speaker A: I'm pretty sure that there is a way to actually get to the Office 65 portal from inside the Azure Portal. Like it's a.
[00:46:44] Speaker C: There is, but it is clearly not the same product.
[00:46:48] Speaker A: You clearly pass a perimeter that you notice. Yeah, same thing happens in Google actually with, when you go into Google workspaces from Google Cloud. There's definitely a clear like, oh, this UI changed.
[00:46:58] Speaker C: Yeah. I mean by the same math, I also feel like, you know, quicksight is not part of Amazon because it's whatever they screw up that I'm like, nope.
[00:47:10] Speaker A: I mean I think you had to look at Oracle too because Oracle's moved all of their consumption based Oracle database pricing into their cloud intelligent cloud revenue. So I mean they're also doing this too. So really the problem is Amazon doesn't have a competitive mail and Office product.
[00:47:24] Speaker C: No kidding.
[00:47:25] Speaker A: Yeah, so if they had that, then we can compare the apples to apples.
[00:47:29] Speaker C: Maybe that's why they're going after Microsoft so hard.
[00:47:32] Speaker A: Maybe.
Doubt it though. I think they're just probably better reasons.
Well, it is the beginning of the year and 2025 we've seen several CEO letters to their employees basically talking about how 2025 should be the toughest year ever and get buckle down and get to work. And you know Zuckerberg, we're laying off the bottom 5% of you losers, get the hell out. But we have one from Satya Adela from Microsoft, so I thought I would share this one. He basically sent this as an update to Microsoft employees last week and he publicly shared it because it would got leaked anyways. So good choice.
Satya indicates that they are heading into the next era of the AI platform shift and 2025 will be about model forward applications that reshape all application categories. Unlike previous platform shifts, this impact will impact every layer of the application from GUI servers, cloud, Native databases and APIs all being done at once equivalent to being 30 years of change compressed into exactly three years. And this is their third year so they've been at for two years, they said. He says they will build a gentech application with memory entitlements and action space that will inherit powerful model capabilities and will adapt these capabilities for enhanced performance and safety across roles, business processes and industry domains. This will lead to what he calls the AI first App Stack one with the new UI UX patterns, runtimes, build with agents, orchestrate multiple agents and reimagine management and observability layers. And so does imperative that Azure must become the infrastructure for AI. And while they build AI platforms and developer tools spanning Azure, AI Foundry, GitHub and VS Code on top of it, the good news for Satya is that they've been working on this already for two years and have learned a lot in terms of the systems, apps, platforms and tools required in the AI era. And to further advance this roadmap and everything I just said, they're creating the new core AI Platform and Tools team. This new division will bring together Dev Div, which I don't know what that is, a platform and some key teams from the Office of the CTO including AI Supercomputer, AI Agentic, Runtime and Engineering Thrive with a mission to build the end to end Copilot and AI stack for both first party and third party customers to build and run AI apps and agents. This group will also build out GitHub and Copilot, thus having tight feedback loop between the leading AI first product and the AI platform to motivate the Stack and its roadmap. And this new core AI team will be led by Jay Paric, evp.
[00:49:43] Speaker C: You know I was really excited to hear the thoughts on the AWS CEO and now that I'm reading this I'm no longer interested because it's like it's. I guess I agree, you know and I you know this is a strategic alignment within the business that if I was within the business maybe it would make sense. Or not.
[00:50:05] Speaker A: Yeah, Dev Div is definitely something internal like it was in the article and I was like I don't know what that is that Typo on the article and I, yeah, I didn't go to Google it but I was like, oh, this is an internal demo that they shared publicly. So they didn't think to change it for what all of us layman people will know what that is. But yeah, I assume it's kind of like developer productivity or something. It's my guess.
[00:50:23] Speaker C: That would be my guess too. Yeah. Is that it's. They're in like internal sort of platform engineering.
[00:50:30] Speaker B: Yeah, it is a pl. It's, it is a platform engineering team developer division. We build the tools, platforms and services that build Microsoft and the world.
[00:50:38] Speaker A: Well that's a, that's a charter if I ever heard one.
[00:50:42] Speaker B: Yep.
[00:50:43] Speaker A: We're a cloud pod.
[00:50:44] Speaker C: I mean it's, it's kind of neat though like if you think about that and then they putting that with the AI agentic team and that like it could be really. Because I mean it is, that is my day to day life. Like it's that as I challenge how do I get AI here and there's so many hurdles.
[00:51:02] Speaker A: It's interesting to me that the GitHub copilot team is actually going to move underneath this which means that the copilot's basically getting divorced from GitHub and the GitHub dev team I assume than to do this.
So that's kind of interesting because.
[00:51:15] Speaker C: Well I, I wonder. Because I Wonder if the GitHub team is actually outside of this already or if the. Because that, that could be the thing too is the co pilot team might be separate.
[00:51:25] Speaker A: Yeah, it's always been separate.
[00:51:26] Speaker C: There's like 9,000 copilots. So I don't know which is why.
[00:51:29] Speaker A: Well, I mean I thought it was just branding thing. I don't think that actually meant the copilot team that built GitHub co pilots doing all the other copilots too. I didn't, I never made that logical leap so.
[00:51:38] Speaker C: But it might be a separate AI team.
[00:51:40] Speaker A: It could be, yeah. I don't know. Again, this is inner workings of Microsoft politics that I don't really care about. But I do think it's interesting that they've, you know, they feel so strongly about this that they're creating this core AI team to really focus on platform and tools and how to be AI first in their internal uses as well as how to then make that available to us as customers. Interesting stuff. Curious to see how it turns out this year and is Satya right or not? He's a smart guy. He thinks a lot.
[00:52:03] Speaker C: He is, you know, he's Done wonders for that business.
[00:52:06] Speaker A: And yeah, I mean the Steve Ballmer. Could you imagine if Steve Ballmer was still in charge of AI at Microsoft right now?
[00:52:15] Speaker C: Let's just be a robot that yelled at us. Yeah.
[00:52:19] Speaker A: Let him just run the clippers. Like stick to that.
And finally, one last story that I'm putting in here. Just because I'm a child and I will never not laugh at the fact that point of sale is abbreviated to positive and in this particular case, Oracle is supercharging their retail operations with their new pos.
[00:52:38] Speaker B: Yes, yes.
[00:52:43] Speaker C: I bet they are.
[00:52:44] Speaker A: There is nothing crowd related here other than a couple of little mentions, but apparently this new X Store POS system is containerized and will run on top of OCI and OCI container instances or on your OCI Roving Edge infrastructure inside of your store. For those of you who want to run containers remotely from in the store from your central corporate headquarters, God bless you. That would not be me. But yeah, xdor POS new and improved supercharged for retail operations from Oracle. So yes, and I will continue chuckling all the way out of the show.
[00:53:18] Speaker C: Well I can't wait to see what this POS delivers.
[00:53:21] Speaker B: Yep, how long till stinks.
[00:53:26] Speaker A: Even the screenshot they used for this, I was just like that was the screenshot you chose is a terrible screenshot. Like anyways, yeah I like the idea.
[00:53:34] Speaker B: Of the, the roving edge stuff is kind of cool and containerized things at the edge is kind of cool but it's, it's so painful going, going to CVS or a gas station or any place and the technology is not working the way it's supposed to and they, and they're on the phone calling their centralized IT support team which is probably 5,000 miles away or more. And yeah, we're still number 1, 2, 3, 4 and you know I've got a line of customers here waiting and nothing's working. What do I do? Like I just trying to troubleshoot Kubernetes apps over the phone with, with the cashier at 3 o'clock in the morning.
[00:54:07] Speaker C: I hear you but I don't think this is going to make that better.
[00:54:10] Speaker B: It is not.
[00:54:11] Speaker A: No.
In many, many, many, many, many, many moons ago I worked at Best Buy as a tech, as a squad member before they got the white shirts and all that. So that's how long it's been. And you know as being a geek swag member and being technical and if the positive Best Buy broke we were in charge of calling help desk and I can tell you a couple of things that I learned. One was the entire store ran on a VM via like a small ESX host in the back. And then if there was a problem, one of those ESX VMs that ran on top of it had an image server on it. And literally you call and say, hey, this POS isn't working. And they literally like, like 40 seconds later it would reboot and start reimaging itself. And that was pretty much the standard response for any problem you're having. A positive was we just reimage it.
And so I can imagine the ability to do a container really simplifies that, which makes some sense to me, especially with VMware's Broadcom pricing structure. I'm sure a lot of retailers are thinking containers sound great in the store.
I was just looking at the article again as we were talking about it, as I do sometimes, and I noticed there was another acronym that I missed. And I'm disappointed because I didn't introduce you guys to BOPIS or Buy Online Pick up in Store. And so the Xdor POS Bopus support is all available to you as well. And that's just great.
[00:55:32] Speaker C: Bopus over there and over and pick it up.
[00:55:35] Speaker A: You know, that's what I'm going to do.
[00:55:36] Speaker C: Yeah.
[00:55:37] Speaker A: Well, that's a fantastic way to end the show, you guys. I hope you have a great rest of your week and weekend and we'll see you here on the Cloud POD next week.
[00:55:47] Speaker B: See you later, guys.
[00:55:48] Speaker C: Bye, everybody.
[00:55:52] Speaker B: And that's all for this week in Cloud. We'd like to thank our sponsor, Archera. Be sure to click the link in our show notes to learn more about their services.
While you're at it, head over to our
[email protected] where you can subscribe to our newsletter, join our Slack community, send us your feedback and ask any questions you might have. Thanks for listening and we'll catch you on the next episode.
