Quantum is Here! – Then More on re:Invent and Data Protection – Infrastructure Matters EP64

Welcome. Uh, this is Diane Hinchcliffe, uh, VP of CIO practice at the Futurum Group. And we're here for another episode of Infrastructure Matters.

This is episode 64, and I have my co-host, uh, Kimberly Bates with us. Welcome, Kimberly. Good day, and welcome to the holiday season.

It is the holiday season and also means a little bit less travel for us analysts, which is kind of nice. But boy, have we been busy. A la a week and a half, two weeks, months, six months.

I, months, months, months, months, months, months. In fact, Keith is not here because, well, he took some time off. He, he needed to take some time off.

So there We go. Oh, I'll Mrs. Cowboy hat.

I'm sure others will too. Yeah. As well.

Um, so let's dive right in. Um, there was, um, a tremendous amount of news, uh, out of Las Vegas last week with the biggest show that caps off the end of the year. So, uh, that was Amazon's AWS's, uh, reinvent, uh, week long extravaganzas.

I spend more time at that event than any other. I was there all week was unusual for me. Um, Kimberly, you and I even sat together at some of the, uh, announcements at the analyst summit, and they just kept coming fast and and thick too much for us to even cover here.

Uh, but let's dive into what we thought, uh, was the most significant or interesting announcements. Kimberly, you wanna go first? Sure.

Um, I'll just cover some top level things that I thought was pretty cool. The, and you know me on data storage. So they announced for S3, some very, very significant items.

Um, they're, they are promoting S3 as the data lake. Bring your data to the da, their data lake, and we're gonna enable this for your entire AI strategy. Now, that is to say you still may need luster to power through your GPUs, but the S3 is the, is the source for data.

The two big things they announced is being able to do tables with the end there for, for your iceberg, as well as the metadata processing. And they're doing quite a bit within the S3 to improve the performance kind of level on there. I'm gonna take a look at my notes here.

Um, in terms of the data layout as being optimized, um, they've done some things to simplify the ta, the table security. Um, they are optimizing garbage collections and row deletion that you need to do for iceberg. So I'm kind of getting into the technical wes here, but some really, really nice stuff, items that they have done.

There's, there I am with that technical term stuff. The second piece of that was the S3 metadata. That's gonna be part of the iceberg table.

So there's, I mean, there's an entire sessions on that, so if you can go up and take a look at it, you know, either I'd reinvent or we're probably gonna be publishing on a little bit more in terms in depth with our research on there. Well, I really liked what the, the metadata announcement because, you know, S3 was the number two service they ever released, and they, they shared some of the, the usage data about how much people actually store in there. And they're, and people are cra in the kitchen sink into S3, and they've admitted it.

It actually tells you what you have. Uh, you can actually go in there and throw everything to a data lake and then go, well, what's in there? And it will actually tell you.

And I thought that was, that's really gonna help make S3 much more self-service, much more, um, uh, you know, useful to people who are not necessarily developers who've, you know, or data architects who just wanna work with their, their, the data they have in S3 and understand what's in there. I thought that was made, um, was a really good announcement. Absolutely.

I think the number they put up there was 400 trillion objects today scored in three. So it's kind of like, oh, there's a bit up there. Um, the other two things that was really cool, and I'm gonna go into, again, I'm looking at my notes here, was some two really cool customer stories.

And I'll start with the first one, which is Merck. Merck has been there, and it's non-AI. One is Merck has been there for 13 years.

Um, they started, um, what they called the Blue sky project Infrastructure as code, um, project they were doing, what they were looking at is they just could not keep up with development. And so the pretty much what they gave is AWS responsible for the maintenance, the patches, all those kind of areas, while they could focus on Modernizer apps. And which they did is, um, they retired.

They quoted a thousand apps in the modernization process that they've done. Mm-hmm. Wow.

Um, so now they feel like they, what they can do now is really focus on governance and guardrails, especially as they go into the AI space. The other one that was super cool was Novo Nordisk, um, which is a pharmaceutical out of Belgium, if I get that right Now, They're famous, Very famous. They are known.

They have, they have 50% of the insulin business. We won't even talk about that. That's another issue.

Oh. But anyway, so what they talked about is how they're using AI for regulatory filing. So there are about 200 to a thousand documents that have to be prepared for filing, and that would be to the FDA or Francis, you know, FDA or England's, FDA, et cetera.

And it can take them, um, 12 to 15 weeks to prepare and to submit. Um, what they have done is they have applied ai, um, that, and this is taking this down to ready for mm-hmm. 10 minutes.

Oh wow. 10 Minutes. Um, they have reduced their staff that's doing the writing from five to three medical writers.

They have redeployed those into other things. They have reduced a total of 27 full-time equivalents. What they used is Bedrock, MongoDB, anthropic, and some other foundation models.

Um, and has the tagging and the metadata, um, on the documents that they have that changed how they've organized the company in terms of submitting this. So there's more collaboration between the scientists, I guess they're doing the clinical trials and the people that are doing, so it's a real, real big, you know, change. I mean, what they've looked at is they've moved from being content writers to tech experts on the content.

Um, great case study about what AI can do. So when people ask me, is this stuff real? That to me is a real fabulous, Oh yeah.

We're starting to really see the, the first real concrete stories about AI transformation, uh, that's been successful. And I think, yeah, I agree with you. That's, that's one of the top ones.

And, and, and it was, uh, it was interesting, uh, interesting how a Amazon's really come from being on the back foot with ai. Um, you know, they were trying to play catch up and they were, you know, late with Bedrock, arguably compared to their Hyperscaler competitors. Uh, and, uh, they, but they've been working very hard.

And that was on evidence at, at, um, reinvent last week. Uh, and they've also announced their new, their own set of models, the, the new NOVA models, uh, of, and of various different sizes. And they're really focusing on, on trying to create cost effective models that provide highly accurate, grounded answers.

And this is what a, what a lot of folks are, you know, working on. Um, and from, from a CIO perspective, which is, you know, really my primary focus, the announcements that really caught my eye was, um, Amazon Q Index. Now, Amazon Q is not a household end, but it's their enterprise, um, ai, it's kind of their chat version, their version of chat GPT, except that it's, it's, uh, taught with your enterprise knowledge.

And so Q Index, uh, is a way of, of creating, uh, access through a natural language interface to all your enterprise data. And they, and they, they came out of the gate with 40 different connectors to all your top enterprise systems, whether that's ERP or CRM or whatever. And you now have this single place you can use Amazon Queue to go and safely access all your corporate knowledge across all these major systems.

And more connectors are coming in and it'll, it'll take years, um, for them to, to fully realize that vision. But Q Index is how, if you're gonna, if you are an Amazon customer and AWS is still the, the top dog in cloud, um, how you can bring all your data together into one place and start using it, start delivering that to customers. I thought that was, uh, one of the most important announcements from an enterprise AI perspective.

How, how you're really gonna deliver AI infrastructure across all your silos. And so Q Index was the, the big, uh, news around that. And it, it had the same thing with SageMaker, which is their, uh, very successful analytics product.

Uh, they now have, uh, SageMaker index the same thing, allowing you to access all your analytics across SageMaker in a single place. And so this is now they're starting to show their hand on, on a more comprehensive enterprise class vision, not just point AI solutions, but actually starting to think about enterprise AI across all of your IT systems. Uh, and this is the kind of their early maturity we need to have that CIOs need to have in order to really start delivering AI value to, to, um, to their internal stakeholders and to, uh, external customers.

So tho those are some of the ones that really caught my eye. Yeah. And did you take a look at the Q developer one, which is their, their tool for the developer side of the house That time?

Yes, I know it. I mean, we can spend the whole show on that. It was, Yeah, we could.

Absolutely. Anyway, uh, guys, it was just some phenomenal show. Um, so much in there.

It, it was overwhelming. My poor brain was hurting. Yeah, no, exactly.

But it, I think that they've, they've now, uh, you can safe to say, well, the things like bedrock guardrails, where they're really gonna say you can run all your different models that you have. Everything is, that's been, that's been, um, uh, you know, uh, certified to work with Bedrock, you now, it can put guardrails on it to make sure that it's all safe and follows the same AI policy. And when it goes out of, uh, you out of permissions or out of policy, uh, you'll catch it.

And you can do that consistently across all your models. This is exciting stuff, uh, from an AI perspective, uh, for those who have to make it work, actually deliver it. Right.

So, uh, I was really thrilled to, to see all that. But that's, that's, um, that was reinvent. Um, I have a, a super thread on, uh, on X that you can see.

It has every single announcement, uh, that I was able to track. I, I'm sure I missed a few. But, um, uh, and we also have the RUM has, uh, full coverage, uh, coming out, um, uh, starting this week and, uh, next week on, on Reinvent.

com. Uh, we also, so, uh, Kimberly, what's up next? Uh, you, uh, you were, uh, uh, uh, talking about, uh, the American Society for ai.

Tell us a little bit about that. Yeah, American Society for ai, they, um, they're a fairly new, um, organization. They've been around for a little bit over a year and a half, two years.

Um, it is a by invitation, kind of a think tank group. There's 125 members that come from all over. I mean, we're talking tech people, we're talking state, local, um, federal government, you know, I think we have three House of Representatives on their military.

Um, we have, uh, in this session here, we had a, one of the, um, lieutenant generals, um, that was there that had been involved with ai. Um, we have professors from MIT, Oxford, George Washington, you know, Stanford, you name it, they're all there. Yeah.

We've not Legal cross section of, uh, of who's who, which is what you need if you're gonna have any kind of real, real society. Right. And artist and sociologist.

Oh, nice. So, I mean, a really, really, um, some amazing folks that bring on being and, you know, just super smart people. I'm, I'm like the dumbest person in the room.

I just was like, also VCs were there. So I'm Sure that's not true. But as you know, from a think tank group, it was afraid.

And one of the things that we do is debate topics. Um, so we do a five minute, five minute debate questions and that kinda stuff. And we had a couple things.

And what I wanna touch on is when you bring this very diverse group together, what are the things that they felt like were the big issues? And where are those going? Um, one of them open source of LLMs, should we or should we not?

And, um, you know, it went on both sides of it, arguing it, where it's going. Another who owns the data that's in those LLMs who owns the data that comes out of the LLMs? And this is especially true for when we started talking about the artists and copyrights, those kind Of things.

And well, my content goes into an LLM. Yeah. I, I want fair compensation.

I think that's only reasonable. Yep. And then there's other people that talk about differently.

Another thing is the accountability for the actions of the agent or the agent assistant. Where's the liability? That was a lively discussion with the lawyers in the room and us in the room and where this goes and that kind of stuff.

And, you know, basically right now we're using negligence principles, you know, have you been negligent with what you're doing? But what happens when we let these things go and start making the decision, you know? And, and we came back and started talking about whamo.

We started talking about auto driving, you know, where we have no tolerance for an accident where you and I have probably had maybe an accident. I know I have. Um, so, and, and there's, there's in how that plays out.

Um, another one was something called hybrid ai. And I know you're gonna talk about quantum here in a bit, but talking about how, you know, what we do have now Transformers, symbolic type of technologies, mixing that with your rule-based technology and mixing it with quantum and how we're gonna probably see these things come together to address some of the stickier problems. Like, and one of the stick problems we talked about was logistics and gentlemen got up and talked about how they're using hybrid, um, and they would never have been able to do what they're doing.

Um, which is a company called Savant X, which is doing logistics management. Yeah. Well, you, you can tackle traveling salesman problems and other things that are, that are too difficult for traditional or classical computing to handle if you augment it with quantum.

Yep. Um, well, we talked about that and then talked, um, the talent issue was a big discussion as well. Um, and then talking about how, I mean, the professors that were talking about how they're tr trying to train up these people to come up, you know, the Gen Zs or whatever, and the difference between you and I who have been around this technology for a long time, and what it means for Gen D to come up in here in terms of this world, and how do we apply, you see, so those are all debated topics, and there's enough time to hear, to kind of go into those things.

But you know, what the, the society will do is start to push things out in terms of recommended policies, research kind of things that we'll see over time. It's self-funded. There's no, uh, you know, there's nobody dropping money in here, so we're not being influenced by, you know, the government or some sort of Lobby.

Yeah. That's, and that's why those associations of societies are so important, is that they can be impartial, arbiters, um, and bring a lot of expertise, uh, that doesn't have necessarily a, you know, a giant tech company behind it, for example. Right.

And we have the rules are in terms of anonymity, they, you know, we, whatever those rules are, where you can, you can't bring who said what, but you can bring what was said, Uh, chat, the old Chatham rules. Uh, Chatham rules. Thank you very much.

Yes. It was definitely Chatham rules there. Um, so which was, you know, made for very lively discussion, especially when we had, um, you know, somebody from the state of ca uh, California, the heads up with AI stuff, talking about what they're doing and how they're, how they're experimenting with it and what, you know, the areas that they're going.

Um, you know, last time we had, uh, folks from New York City and when we were in New York talking about it. So it's, It's, well, I mean, these are all the conversations of the day. And, and in fact, we're doing a big study on CIOs and AI right now at rum.

Um, and one of the things that's come as we talked to a bunch of healthcare, uh, CEOs, uh, what's really come out is, uh, there's a big tension between the clinicians, uh, will be, will be using their AI models to determine what the, what the care should be. And the insurers will have their own model about whether it's something should be treated or not. And their, the, the biggest concern that's emerging is, is how do you decide whose model is right?

Which is the right one? And so these, that has massive societal implications to everyone, uh, receives healthcare. And pretty soon all of that is going through models.

We heard all United Health, uh, you know, has, uh, used AI models to deny care. And that's been quite controversial. Uh, and that's gonna be a very hot topic, and hopefully your society there can help us navigate that, that challenge there for us.

Well, and that gets back to the LLM open source piece of it. Should, I mean, should that be open source enough so we can peer in to see how they have made those decisions and made those l you know, made those things. I mean, we talked about, you know, for instance, like you were thinking genomics shouldn't genomics be an open source LLM except for then we pop into saying, well, what if the bad guys get ahold of that?

Well, of course. And then you've had the whole commercialization of scientific research is a whole other topic about, uh, you know, a lot of that and information is no longer freely available anyway, so what do you do? Mm-hmm.

Right. So, mm-hmm. So, uh, moving along, uh, uh, Kimberly, you have other news, uh, for us, uh, uh, Cohesity.

Um, will you walk us through the, the news there, Right? Um, data. So we had two big announcements from data protection companies.

Uh, I know it's kind of boring data protection. The last thing on the thing, but kind Of, sorry, there was the number one thing out of my CIO survey was, uh, was all cyber topics is the hottest thing you go Yeah. Right.

And that's why we're seeing, um, a transition of people up upgrading or changing out their data protection systems, et cetera. So what just happened is Cohesity, which is a, you know, has, is a, was a start, has been a startup, pretty small and Veritas, long time data protection guy, been around very long time. Yeah.

New et cetera. Um, they had announced about a year ago that they were going to join Ha join. Um, what happened is that Veritas split two, um, so like things like the enterprise side, uh, technology, MBU and a few of the others is now part of Cohesity.

Cohesity is the big company. So Veritas was owned by a PE firm. So this is the way for, you know, that moved over there.

And then another firm was firm formed called Acter, which owns the backup exec technology, as well as a few other technologies. So we now have this Veritas Co, it's now the name is gonna be Cohesity. Um, Veritas is bigger than Cohesity.

So you're looking at, okay, so how is this gonna work? What's gonna be happening here is Cohesity has done quite a bit of work with ai. So what you're gonna look at is saying, okay, so can I take all that Veritas data that's stored in, you know, in backup kind of stuff, and can I start using that in terms of a o option for starting to use for training?

And that's kind of where this is going over the long haul. Um, so not only just protection, but also in that file system kind of environment. So that was pretty, pretty cool.

The other one that was really big, and you kind of go Holy crud, um, is a company called Veeam. And they are about a billion dollars in size firm. Um, they just issued a new equity offering, which is a secondary offering.

Oversubscribed. Yeah, Interesting. Their valuation is now 15 billion, expect them to go public sometime in the next 18 months.

Um, but they have made some huge bets on where we're going with, um, things like three six protecting 365 Salesforce and some of the other as a service kind of, um, software offerings. And that's where we're kinda expect them to explode in, in the business that they do. So kind of like, wow.

Yeah. Right. Well, I, I think all of that, uh, you know, in what's old is new again, in terms of infrastructure.

Uh, and we saw this with, uh, with quantum computing at, you know, GPUs made compute exciting again. Uh, and now quantum computing appears to be perhaps on final approach to actually getting into the data center. Now they, we've had these big bespoke, uh, quantum, uh, compute, uh, devices in the data center in hyper in the hyperscaler clouds for a while, but they have very limited access.

They haven't really been able to scale it up and there hasn't really been able to look lot of demand because the chips can't do much yet. And in fact, they've just now arrived in chip format, right? So the last few years, uh, but Google had a major announcement.

Their new Willow quantum ship, um, uh, has a a hundred and, uh, four qubits, I think it is, uh, it's in the low hundreds. Uh, and uh, it, they, they, they made a splash 'cause they announced that they believe that it, it, it's performance is so high, it must prove that, uh, that the, the multiverse exists and that that quantum computing is the, is the, is the, is the way to access it. I dunno if you saw that, but it created quite a bit of hoopla with both scientists on both sides saying yes and no.

But, um, you know, we see with IBM's, her and Chip, which has, uh, you know, only a handful, only a couple dozen more, um, qubits we're, we're getting the, the, the, the, uh, the the point where we now have quantum chips that can begin to do things beyond just simple, you know, random number generation and things like that, because that's all you can do. A hundred and some odd qubits is not very much. And China this week also announced their own, um, uh, quantum chip with 504 qubits, uh, which is four, four times more than both Google and, and IBM have.

So Quantum's really developing these, these new chip packaging of, of quantum is really important because that's gonna allow them to enter the data center. Uh, and we'll see, we were talking about hybrid ai, uh, and be able to, uh, you know, supercharge AI by adding quantum to the, to the compute mix at, at the infrastructure level, uh, in our data centers and in the cloud. Um, it's getting close to being real.

The error rates are dropping to the point where you actually can use these for business applications. Not quite yet, but we see you plot all the trajectory forward. I, I would agree with, uh, Arvin, Krishna, the CEO of IBM, that we're likely to see these really being a usable offering, uh, in the next couple years.

And these are all great proof points for that. I dunno what you think, what do you think, Kimberly? Well, as I said, I in at the conference, the American Society for ai, there was, um, a VC there that's, he's, that's all he focuses his company focuses on is Quantum.

And then there was also one of the members is a, on the CEO of a company called Savant X. Lots of airtime here, um, which is Quantum, and they're implementing, they have implemented systems that are doing logistics management for, um, the la I believe it's LA Harbor Container, um, trucking. Um, they're doing something that he won't talk about for, I think it's the Air Force.

Um, and they, they're about to do a, um, deployment for one of the, um, trains train. Um, I think it's, uh, one of the train systems. Um, and that is about loading on the, you know, the container loading onto the, the trains and scheduling them.

Um, so I mean, these are real apps. I mean, these are real kind of, uh, you that are creating some significant value proposition for these organizations. So it's real, it's here.

Um, No, that's here. That's, well, that's what well, I title the show. The Quantum is here, And what scares Quantum is here, what scares me a bit is hearing you talk about 500 qubits with China, because I, if I recall correctly, the, um, government has moved up the timeframe in terms of when people needs to be, have quantum be quantum secure, um, because of this.

And so I, one time I thought it was, uh, 2030, and now they're moving. I think if they moved it to 20, You have to because, because you only need about a thousand qubits, maybe 1200 cubits to break basic RSA security. Oh, Geez.

Okay. So we are right. We were right there.

We're right there. Yeah. Yeah.

We we're, we're one or two gen chip generations away from where the, that's the basic RSA 2 56 can't be broken for a while yet, but still. Um, so The CIOs that you work with, are they very, are they aware of this? I mean, especially since cyber, It's on the risk.

If you, if they have a list of risks, um, they know that all of a sudden everything, all their secrets will become readable, um, here in about five years. That, that's how they're, that's how they're looking at it. So yeah, it's on the, the, the list of risks that they're managing, are they investing a lot in it?

No, they're expecting, um, their vendors to be able to, to dig them out and, and, you know, re-encrypt everything or whatever it has to, whatever they have to do To be quantit secure. Wow. Yeah.

This, well, and, and the, the actually, the whole thing with, with China is really heating up because, uh, we're now really blocking China's access to AI chips. Uh, and we just had new regulation come out this week, um, to block third party or third, uh, countries from, you know, being intermediaries for that. So we're really trying to tamp it down as a result.

Uh, China, uh, or I think at the beginning of the week, blocked our access to their, um, their sources of antimony, gallium, and other things that we need to make our, those chips, uh, the AI chips. So a real battle is, you know, at a slow level right now, but it's brewing in terms of, uh, China's now where we're gonna have to build our own AI chips is we're not gonna be able to source 'em from the West. So, um, we, we'll, we're gonna see what happens here, you know, uh, might be the West gets to GPUs and on, uh, quantum is done with China, but we'll see what happens.

Yep, yep. So it's, uh, and then all the stuff with Intel kind of like blends right into that same kind of discussion there that's happening. So anyway.

Okay, cool. Yeah. Well, to, uh, round off the show this, uh, this week, unless I can't believe, unless you have anything else after this mm-hmm.

0 was out. Uh, this is a major model, uh, upgrade. It's fully multimodal.

Um, it, um, I'm watching it, uh, uh, uh, climb the AI leaderboards. Uh, but what's really significant is, uh, they have associated with it, uh, it's something called Project Mariner, um, which, uh, uh, shows off the agent-based capabilities. A agent-based AI is the hottest topic right now.

A, uh, uh, AI doesn't just produce content for you, but does things for you. 0 really focuses on that. And Project Mariner is an interesting plugin.

0, and it will use your browser with all your access that your browser has and passwords and everything to go out in the internet and do things for you. And they, they claim it can handle complex multi-step tasks across multiple applications to do things on your behalf directly through the browser. Uh, and why I asked you a lot of reservations about directly handing AI control of our IT systems, uh, it's still an impressive tech demonstration.

Uh, I watched what, uh, what it can do. And so agent-based AI is here, folks. The question is, can we control it?

Can we protect ourselves from it? So it's gonna be interesting when You say it does something for you, what kind of things will it do? You can ask it to, to book the cheapest vacation trip to, um, to Europe, you know, so go and search all the different sites.

Get me hotel and flights. Uh, send me an email just to confirm everything, uh, and I'll give you a yes or no, go off and do it, and then you go get coffee. Right.

You know, So you still have the expert, and that's what I call that the person that looks at what happens and says, okay, go do that. It's not automatically executing for You. It'll, no, it will absolutely could.

Yes, yes, you can have it do that for you. You can say, go find the, the cheapest place to, uh, to buy this item and buy it. Buy it for me, and it will do it without asking you if that's what you want.

Well, and well, that, and that gets back to the negligence principles and the legal issues and everything else that we talked about is there we go. Okay. Wow.

Okay. If they, you know, I feel sorry for people that are not in the tech industry to a bit, that are spending time looking at this things. I mean, one of the beauties of this podcast is I learned so much from you guys.

You know, I know we don't have Keith here, but I would learn, you know, open my brain, pour a little bit more in it, and, but, um, exactly. There's no way you can track all this. It's just crazy.

No, Crazy. Even if you do it full time. And we do.

I mean, that's the, that's what's so funny, right? Yeah. Yeah.

Well, well, anyway, it was a amazing week in infrastructure. Um, uh, and, um, uh, we hope that, uh, you'll, you'll join us for our next episode. Kimberly, when is our next episode?

Are we, I think we're recording on Friday next week. Okay. Yep.

Great. I'll be there too. Okay.

Alright, well thanks for stopping by everyone. Have a great week. Okay, thank you.

Bye.