.banner-thumbnail-wrapper { display:none; }

AI

As enterprises face hybrid IT complexity, new management solutions beckon

The next BriefingsDirect Voice of the Analyst interview examines how new machine learning and artificial intelligence (AI) capabilities are being applied to hybrid IT complexity challenges.

We'll explore how mounting complexity and a lack of multi-cloud services management maturity must be solved in order for businesses to grow and thrive as digital enterprises. 

Listen to the podcast. Find it on iTunes. Get the mobile app. Read a full transcript or download a copy. 

Here to report on how companies and IT leaders are seeking new means to manage an increasingly complex transition to sustainable hybrid IT is Paul Teich, Principal Analyst at TIRIAS Research in Austin, Texas. The discussion is moderated by Dana Gardner, principal analyst at Interarbor Solutions.


Here are some excerpts:

Gardner: Paul, there’s a lot of evidence that businesses are adopting cloud models at a rapid pace. There is also lingering concern about the complexity of managing so many fast-moving parts. We have legacy IT, private cloud, public cloud, software as a service (SaaS) and, of course, multi-cloud. So as someone who tracks technology and its consumption, how much has technology itself been tapped to manage this sprawl, if you will, across hybrid IT.

Teich

Teich

Teich: So far, not very much, mostly because of the early state of multi-cloud and the hybrid cloud business model. As you know, it takes a while for management technology to catch up with the actual compute technology and storage. So I think we are seeing that management is the tail of the dog, it’s getting wagged by the rest of it, and it just hasn’t caught up yet.

Gardner: Things have been moving so quickly with cloud computing that few organizations have had an opportunity to step back and examine what’s actually going on around them -- never mind properly react to it. We really are playing catch up.

Teich: As we look at the options available, the cloud giants -- the public cloud services -- don’t have much incentive to work together. So you are looking at a market where there will be third parties stepping in to help manage multi-cloud environments, and there’s a lag time between having those services available and having the cloud services available and then seeing the third-party management solution step in.

Gardner: It’s natural to see that a specific cloud environment, whether it’s purely public like AWS or a hybrid like Microsoft Azure and Azure Stack, want to help their customers, but they want to help their customers all get to their solutions first and foremost. It’s a natural thing. We have seen this before in technology.

There are not that many organizations willing to step into the neutral position of being ecumenical, of saying they want to help the customer first, manage it all from the first.

As we look to how this might unfold, it seems to me that the previous models of IT management -- agent-based, single-pane-of-glass, and unfortunately still in some cases spreadsheets and Post-It notes -- have been brought to bear on this. But we might be in a different ball game, Paul, with hybrid IT, that there’s just too many moving parts, too much complexity, and that we might need to look at data-driven approaches. What is your take on that?

Learn More About

Hybrid IT Management

Solutions From HPE

Teich: I think that’s exactly correct. One of the jokes in the industry right now is if you want to find your stranded instances in the cloud, cancel your credit card and AWS or Microsoft will be happy to notify you of all of the instances that you are no longer paying for because your credit card expired. It’s hard to keep track of this, because we don’t have adequate tools yet.

When you are an IT manager and you have a lot of folks on public cloud services, you don't have a full picture.

That single pane of glass, looking at a lot of data and information, is soon overloaded. When you are an IT manager, you are at a mid-sized or a large corporation, you have a lot of folks paying out-of-pocket right now, slapping a credit card down on public cloud services, so you don’t have a full picture. Where you do have a picture, there are so many moving parts.

I think we have to get past having a screen full of data, a screen full of information, and to a point where we have insight. And that is going to require a new generation of tools, probably borrowing from some of the machine learning evolution that’s happening now in pattern analytics.

Gardner: The timing in some respects couldn’t be better, right? Just as we are facing this massive problem of complexity of volume and velocity in managing IT across a hybrid environment, we have some of the most powerful and cost-effective means to deal with big data problems just like that.

Life in the infrastructure

Paul, before we go further let’s hear about you and your organization, and tell us, if you would, what a typical day is like in the life of Paul Teich?

Teich: At TIRIAS Research we are boutique industry analysts. By boutique we mean there are three of us -- three principal analysts; we have just added a few senior analysts. We are close to the metal. We live in the infrastructure. We are all former engineers and/or product managers. We are very familiar with deep technology.

My day tends to be first, a lot of reading. We look at a lot of chips, we look at a lot of service-level information, and our job is to, at a very fundamental level, take very complex products and technologies and surface them to business decision-makers, IT decision-makers, folks who are trying to run lines of business (LOB) and make a profit. So we do the heavy lifting on why new technology is important, disruptive, and transformative.

Gardner: Thanks. Let’s go back to this idea of data-driven and analytical values as applied to hybrid IT management and complexity. If we can apply AI and machine learning to solve business problems outside of IT -- in such verticals as retail, pharmaceutical, transportation -- with the same characteristics of data volume, velocity, and variety, why not apply that to IT? Is this a case of the cobbler’s kids having no shoes? You would think that IT would be among the first to do this.

Dig deep, gain insight

Teich: The cloud giants have already implemented systems like this because of necessity. So they have been at the front-end of that big data mantra of volume, velocity -- and all of that.

To successfully train for the new pattern recognition analytics, especially the deep learning stuff, you need a lot of data. You can’t actually train a system usefully without presenting it with a lot of use cases.

The public clouds have this data. They are operating social media services, large retail storefronts, and e-tail, for example. As the public clouds became available to enterprises, the IT management problem ballooned into a big data problem. I don’t think it was a big data problem five or 10 years ago, but it is now.

That’s a big transformation. We haven’t actually internalized what that means operationally when your internal IT department no longer runs all of your IT jobs anymore.

We are generating big data and that means we need big data tools to go analyze it and to get that relevant insight.

That’s the biggest sea change -- we are generating big data in the course of managing our IT infrastructure now, and that means we need big data tools to go analyze it, and to get that relevant insight. It’s too much data flowing by for humans to comprehend in real time.

Gardner: And, of course, we are also talking about islands of such operational data. You might have a lot of data in your legacy operations. You might have tier 1 apps that you are running on older infrastructure, and you are probably happy to do that. It might be very difficult to transition those specific apps into newer operating environments.

You also have multiple SaaS and cloud data repositories and logs. There’s also not only the data within those apps, but there’s the metadata as to how those apps are running in clusters and what they are doing as a whole. It seems to me that not only would you benefit from having a comprehensive data and analytics approach for your IT operations, but you might also have a workflow and process business benefit by being an uber analyst, by being on top of all of these islands of operational data. 

Learn More About

Hybrid IT Management

Solutions From HPE

To me, moving toward a comprehensive intelligence and data analysis capability for IT is the gift that keeps giving. You would then be able to also provide insight for an uber approach to processes across your entire organization -- across the supply chains, across partner networks, and back to your customers. Paul, do you also see that there’s an ancillary business benefit to having that data analysis capability, and not ceding it to your cloud providers?

Manage data, improve workflow

Teich: I do. At one end of the spectrum it’s simply what do you need to do to keep the lights on, where is your data, all of it, in the various islands and collections and the data you are sharing with your supply chain as well. Where is the processing that you can apply to that data? Increasingly, I think, we are looking at a world in which the location of the stored data is more important than the processing power.

The management of all the data you have needs to segue into visible workflows.

We have processing power pretty much everywhere now. What’s key is moving data from place to place and setting up the connections to acquire it. It means that the management of all the data you have needs to segue into visible workflows.

Once I know what I have, and I am managing it at a baseline effectively, then I can start to improve my processes. Then I can start to get better workflows, internally as well as across my supply chain. But I think at first it’s simply, “What do I have going on right now?”

As an IT manager, how can I rein in some of these credit card instances, credit card storage on the public clouds, and put that all into the right mix. I have to know what I know first -- then I can start to streamline. Then I can start to control my costs. Does that make sense?

Gardner: Yes, absolutely. And how can you know which people you want to give even more credit to on their credit cards – and let them do more of what they are doing? It might be very innovative, and it might be very cost-effective. There might also be those wasting money, spinning their wheels, repaving cow paths, over and over again.

If you don’t have the ability to make those decisions with insight, without the visibility, and then further analyze it as to how best to go about it – it seems to me a no-brainer.

It also comes at an auspicious time as IT is trying to re-factor its value to the organization. If in fact they are no longer running servers and networks and keeping the trains running on time, they have to start being more in the business of defining what trains should be running and then how to make them the best business engines, if you will.

If IT departments needs to rethink their role and step up their game, then they need to use technologies like advanced hybrid IT management from vendors with a neutral perspective. Then they become the overseers of operations at a fundamentally different level. 

Data revelation, not revolution

Teich: I think that’s right. It’s evolutionary stuff. I don’t think it’s revolutionary. I think that in the same way you add servers to a virtual machine farm, as your demand increases, as your baseline demand increases, IT needs to keep a handle on costs -- so you can understand which jobs are running where and how much more capacity you need.

One of the things they are missing with random access to the cloud is bulk purchasing. And so at a very fundamental level, helping your organization manage which clouds you are spending on by aggregating the purchase of storage, aggregating the purchase of compute instances to get just better buying power, doing price arbitrage when you can. To me, those are fundamental qualities of IT going forward in a multi-cloud environment.

They are extensions of where we are today; it just doesn’t seem like it yet. They have always added new servers to increasing internal capacity and this is just the next evolutionary step.

Gardner: It certainly makes sense that you would move as maturity occurs in any business function toward that orchestration, automation and optimization – rather than simply getting the parts in place. What you are describing is that IT is becoming more like a procurement function and less like a building, architecture, or construction function, which is just as powerful.

Not many people can make those hybrid IT procurement decisions without knowing a lot about the technology. Someone with just business acumen can’t walk in and make these decisions. I think this is an opportunity for IT to elevate itself and become even more essential to the businesses.

Teich: The opportunity is a lot like the Sabre airline scheduling system that nearly every airline uses now. That’s a fundamental capability for doing business, and it’s separate from the technology of Sabre. It’s the ability to schedule -- people and airplanes – and it’s a lot like scheduling storage and jobs on compute instances. So I think there will be this step.

But to go back to the technology versus procurement, I think some element of that has always existed in IT in terms of dealing with vendors and doing the volume purchases on one side, but also having some architect know how to compose the hardware and the software infrastructure to serve those applications.

Connect the clouds

We’re simply translating that now into a multi-cloud architecture. How do I connect those pieces? What network capacity do I need to buy? What kind of storage architectures do I need? I don’t think that all goes away. It becomes far more important as you look at, for example, AWS as a very large bag of services. It’s very powerful. You can assemble it in any way you want, but in some respect, that’s like programming in C. You have all the power of assembly language and all the danger of assembly language, because you can walk up in the memory and delete stuff, and so, you have to have architects who know how to build a service that’s robust, that won’t go down, that serves your application most efficiently and all of those things are still hard to do.

So, architecture and purchasing are both still necessary. They don’t go away. I think the important part is that the orchestration part now becomes as important as deploying a service on the side of infrastructure because you’ve got multiple sets of infrastructure.

Learn More About

Hybrid IT Management

Solutions From HPE

Gardner: For hybrid IT, it really has to be an enlightened procurement, not just blind procurement. And the people in the trenches that are just buying these services -- whether the developers or operations folks -- they don’t have that oversight, that view of the big picture to make those larger decisions about optimization of purchasing and business processes.

That gets us back to some of our earlier points of, what are the tools, what are the management insights that these individuals need in order to make those decisions? Like with Sabre, where they are optimizing to fill every hotel room or every airplane seat, we’re going to want in hybrid IT to fill every socket, right? We’re going to want all that bare metal and all those virtualization instances to be fully optimized -- whether it’s your cloud or somebody else’s.

It seems to me that there is an algorithmic approach eventually, right? Somebody is going to need to be the keeper of that algorithm as to how this all operates -- but you can’t program that algorithm if you don’t have the uber insights into what’s going on, and what works and what doesn’t.

What’s the next step, Paul, in terms of the technology catching up to the management requirements in this new hybrid IT complex environment?

Teich: People can develop some of that experience on a small scale, but there are so many dimensions to managing a multi-cloud, hybrid IT infrastructure business model. It’s throwing off all of this metadata for performance and efficiency. It’s ripe for machine learning.

We're moving so fast right now that if you are an organization of any size, machine learning has to come into play to help you get better economies of scale.

In a strong sense, we’re moving so fast right now that if you are an organization of any size, machine learning has to come into play to help you get better economies of scale. It’s just going to be looking at a bigger picture, it’s going to be managing more variables, and learning across a lot more data points than a human can possibly comprehend.

We are at this really interesting point in the industry where we are getting deep-learning approaches that are coming online cost effectively; they can help us do that. They have a little while to go before they are fully mature. But IT organizations that learn to take advantage of these systems now are going to have a head start, and they are going to be more efficient than their competitors.

Gardner: At the end of the day, if you’re all using similar cloud services then that differentiation between your company and your competitor is in how well you utilize and optimize those services. If the baseline technologies are becoming commoditized, then optimization -- that algorithm-like approach to smartly moving workloads and data, and providing consumption models that are efficiency-driven -- that’s going to be the difference between a 1 percent margin and a 5 percent margin over time.

The deep-learning difference

Teich: The important part to remember is that these machine-training algorithms are somewhat new, so there are several challenges with deploying them. First is the transparency issue. We don’t quite yet know how a deep-learning model makes specific decisions. We can’t point to one aspect and say that aspect is managing the quality of our AWS services, for example. It’s a black box model.

We can’t yet verify the results of these models. We know they are being efficient and fast but we can’t verify that the model is as efficient as it could possibly be. There is room for improvement over the next few years. As the models get better, they’ll leave less money on the table.

We’re also validating that when you build a machine-learning model that it’s covering all the situations you want it to cover. You need an audit trail for specific sets of decisions, especially with data that is subject to regulatory constraints. You need to know why you made decisions.

So the net is, once you are training a machine-learning model, you have to keep retraining it over time. Your model is not going to do the same thing as your competitor's model. There is a lot of room for differentiation, a lot of room for learning. You just have to go into it with your eyes open that, yeah, occasionally things will go sideways. Your model might do something unexpected, and you just have to be prepared for that. We’re still in the early days of machine learning.

Gardner: You raise an interesting point, Paul, because even as the baseline technology services in the multi-cloud era become commoditized, you’re going to have specific, unique, and custom approaches to your own business’ management.

Your hybrid IT optimization is not going to be like that of any other company. I think getting that machine-learning capability attuned to your specific hybrid IT panoply of resources and assets is going to be a gift that keeps giving. Not only will you run your IT better, you will run your business better. You’ll be fleet and agile.

If some risk arises -- whether it’s a cyber security risk, a natural disaster risk, a business risk of unintended or unexpected changes in your supply chain or in your business environment -- you’re going to be in a better position to react. You’re going to have your eyes to the ground, you’re going to be well tuned to your specific global infrastructure, and you’ll be able to make good choices. So I am with you. I think machine learning is essential, and the sooner you get involved with it, the better.

Before we sign off, who are the vendors and some of the technologies that we will look to in order to fill this apparent vacuum on advanced hybrid IT management? It seems to me that traditional IT management vendors would be a likely place to start.

Who’s in?

Teich: They are a likely place to start. All of them are starting to say something about being in a multi-cloud environment, about being in a multi-cloud-vendor environment. They are already finding themselves there with virtualization, and the key is they have recognized that they are in a multi-vendor world.

There are some start-ups, and I can’t name them specifically right now. But a lot of folks are working on this problem of how do I manage hybrid IT: In-house IT, and multi-cloud orchestration, a lot of work going on there. We haven’t seen a lot of it publicly yet, but there is a lot of venture capital being placed.

I think this is the next step, just like PCs came in the office, smartphones came in the office as we move from server farms to the clouds, going from cloud to multi-cloud, it’s attracting a lot of attention. The hard part right now is nailing whom to place your faith in. The name brands that people are buying their internal IT from right now are probably good near-term bets. As the industry gets more mature, we’ll have to see what happens.

Learn More About

Hybrid IT Management

Solutions From HPE

Gardner: We did hear a vision described on this from Hewlett Packard Enterprise (HPE) back in June at their Discover event in Las Vegas. I’m expecting to hear quite a bit more on something they’ve been calling New Hybrid IT Stack that seems to possess some of the characteristics we’ve been describing, such as broad visibility and management.

So at least one of the long-term IT management vendors is looking in this direction. That’s a place I’m going to be focusing on, wondering what the competitive landscape is going to be, and if HPE is going to be in the leadership position on hybrid IT management.

Teich: Actually, I think HPE is the only company I’ve heard from so far talking at that level. Everybody is voicing some opinion about it, but from what I’ve heard, it does sound like a very interesting approach to the problem.

Microsoft actually constrained their view on Azure Stack to a very small set of problems, and is actively saying, “No, I don’t.” If you’re looking at doing virtual machine migration and taking advantage of multi-cloud for general-purpose solutions, it’s probably not something that you want to do yet. It was very interesting for me then to hear about the HPE Project New Hybrid IT Stack and what HPE is planning to do there.

Gardner: For Microsoft, the more automated and constrained they can make it, the more likely you’d be susceptible or tempted to want to just stay within an Azure and/or Azure Stack environment. So I can appreciate why they would do that.

Before we sign off, one other area I’m going to be keeping my eyes on is around orchestration of containers, Kubernetes, in particular. If you follow orchestration of containers and container usage in multi-cloud environments, that’s going to be a harbinger of how the larger hybrid IT management demands are going to go as well. So a canary in the coal mine, if you will, as to where things could get very interesting very quickly.

The place to be

Teich: Absolutely. And I point out that the Linux Foundation’s CloudNativeCon in early December 2017 looks like the place to be -- with nearly everyone in the server infrastructure community and cloud infrastructure communities signing on. Part of the interest is in basically interchangeable container services. We’ll see that become much more important. So that sleepy little technical show is going to be invaded by “suits,” this year, and we’re paying a lot of attention to it.

Gardner: Yes, I agree. I’m afraid we’ll have to leave it there. Paul, how can our listeners and readers best follow you to gain more of your excellent insights?

Teich: You can follow us at www.tiriasresearch.com, and also we have a page on Forbes Tech, and you can find us there.

Listen to the podcast. Find it on iTunes. Get the mobile app. Read a full transcript or download a copy. Sponsor: Hewlett Packard Enterprise.

You may also be interested in:

Inside story on HPC’s AI role in Bridges 'strategic reasoning' research at CMU

The next BriefingsDirect high performance computing (HPC) success interview examines how strategic reasoning is becoming more common and capable -- even using imperfect information.

We’ll now learn how Carnegie Mellon University and a team of researchers there are producing amazing results with strategic reasoning thanks in part to powerful new memory-intense systems architectures.

Listen to the podcast. Find it on iTunes. Get the mobile app. Read a full transcript ordownload a copy. 

To learn more about strategic reasoning advances, please join me in welcoming Tuomas Sandholm, Professor and Director of the Electronic Marketplaces Lab at Carnegie Mellon University in Pittsburgh. The discussion is moderated by Dana Gardner, Principal Analyst at Interarbor Solutions.

Here are some excerpts:

Gardner: Tell us about strategic reasoning and why imperfect information is often the reality that these systems face?

Sandholm: In strategic reasoning we take the word “strategic” very seriously. It means game theoretic, so in multi-agent settings where you have more than one player, you can't just optimize as if you were the only actor -- because the other players are going to act strategically. What you do affects how they should play, and what they do affects how you should play.

Sandholm

Sandholm

That's what game theory is about. In artificial intelligence (AI), there has been a long history of strategic reasoning. Most AI reasoning -- not all of it, but most of it until about 12 years ago -- was really about perfect information games like Othello, Checkers, Chess and Go.

And there has been tremendous progress. But these complete information, or perfect information, games don't really model real business situations very well. Most business situations are of imperfect information.

Know what you don’t know

So you don't know the other guy's resources, their goals and so on. You then need totally different algorithms for solving these games, or game-theoretic solutions that define what rational play is, or opponent exploitation techniques where you try to find out the opponent's mistakes and learn to exploit them.

So totally different techniques are needed, and this has way more applications in reality than perfect information games have.

Gardner: In business, you don't always know the rules. All the variables are dynamic, and we don't know the rationale or the reasoning behind competitors’ actions. People sometimes are playing offense, defense, or a little of both.

Before we dig in to how is this being applied in business circumstances, explain your proof of concept involving poker. Is it Five-Card Draw?

Heads-Up No-Limit Texas Hold'em has become the leading benchmark in the AI community.

Sandholm: No, we’re working on a much harder poker game called Heads-Up No-Limit Texas Hold'em as the benchmark. This has become the leading benchmark in the AI community for testing these application-independent algorithms for reasoning under imperfect information.

The algorithms have really nothing to do with poker, but we needed a common benchmark, much like the IC chip makers have their benchmarks. We compare progress year-to-year and compare progress across the different research groups around the world. Heads-Up No-limit Texas Hold'em turned out to be great benchmark because it is a huge game of imperfect information.

It has 10 to the 161 different situations that a player can face. That is one followed by 161 zeros. And if you think about that, it’s not only more than the number of atoms in the universe, but even if, for every atom in the universe, you have a whole other universe and count all those atoms in those universes -- it will still be more than that.

Gardner: This is as close to infinity as you can probably get, right?

Sandholm: Ha-ha, basically yes.

Gardner: Okay, so you have this massively complex potential data set. How do you winnow that down, and how rapidly does the algorithmic process and platform learn? I imagine that being reactive, creating a pattern that creates better learning is an important part of it. So tell me about the learning part.

Three part harmony

Sandholm: The learning part always interests people, but it's not really the only part here -- or not even the main part. We basically have three main modules in our architecture. One computes approximations of Nash equilibrium strategies using only the rules of the game as input. In other words, game-theoretic strategies.

That doesn’t take any data as input, just the rules of the game. The second part is during play, refining that strategy. We call that subgame solving.

Then the third part is the learning part, or the self-improvement part. And there, traditionally people have done what’s called opponent modeling and opponent exploitation, where you try to model the opponent or opponents and adjust your strategies so as to take advantage of their weaknesses.

However, when we go against these absolute best human strategies, the best human players in the world, I felt that they don't have that many holes to exploit and they are experts at counter-exploiting. When you start to exploit opponents, you typically open yourself up for exploitation, and we didn't want to take that risk. In the learning part, the third part, we took a totally different approach than traditionally is taken in AI.

We are letting the opponents tell us where the holes are in our strategy. Then, in the background, using supercomputing, we are fixing those holes.

We said, “Okay, we are going to play according to our approximate game-theoretic strategies. However, if we see that the opponents have been able to find some mistakes in our strategy, then we will actually fill those mistakes and compute an even closer approximation to game-theoretic play in those spots.”

One way to think about that is that we are letting the opponents tell us where the holes are in our strategy. Then, in the background, using supercomputing, we are fixing those holes.

All three of these modules run on the Bridges supercomputer at the Pittsburgh Supercomputing Center (PSC), for which the hardware was built by Hewlett Packard Enterprise (HPE).

HPC from HPE

Overcomes Barriers

To Supercomputing and Deep Learning

Gardner: Is this being used in any business settings? It certainly seems like there's potential there for a lot of use cases. Business competition and circumstances seem to have an affinity for what you're describing in the poker use case. Where are you taking this next?

Sandholm: So far this, to my knowledge, has not been used in business. One of the reasons is that we have just reached the superhuman level in January 2017. And, of course, if you think about your strategic reasoning problems, many of them are very important, and you don't want to delegate them to AI just to save time or something like that.

Now that the AI is better at strategic reasoning than humans, that completely shifts things. I believe that in the next few years it will be a necessity to have what I call strategic augmentation. So you can't have just people doing business strategy, negotiation, strategic pricing, and product portfolio optimization.

You are going to have to have better strategic reasoning to support you, and so it becomes a kind of competition. So if your competitors have it, or even if they don't, you better have it because it’s a competitive advantage.

Gardner: So a lot of what we're seeing in AI and machine learning is to find the things that the machines do better and allow the humans to do what they can do even better than machines. Now that you have this new capability with strategic reasoning, where does that demarcation come in a business setting? Where do you think that humans will be still paramount, and where will the machines be a very powerful tool for them?

Human modeling, AI solving

Sandholm: At least in the foreseeable future, I see the demarcation as being modeling versus solving. I think that humans will continue to play a very important role in modeling their strategic situations, just to know everything that is pertinent and deciding what’s not pertinent in the model, and so forth. Then the AI is best at solving the model.

That's the demarcation, at least for the foreseeable future. In the very long run, maybe the AI itself actually can start to do the modeling part as well as it builds a better understanding of the world -- but that is far in the future.

Gardner: Looking back as to what is enabling this, clearly the software and the algorithms and finding the right benchmark, in this case the poker game are essential. But with that large of a data set potential -- probabilities set like you mentioned -- the underlying computersystems must need to keep up. Where are you in terms of the threshold that holds you back? Is this a price issue that holds you back? Is it a performance limit, the amount of time required? What are the limits, the governors to continuing?

Sandholm: It's all of the above, and we are very fortunate that we had access to Bridges; otherwise this wouldn’t have been possible at all.  We spent more than a year and needed about 25 million core hours of computing and 2.6 petabytes of data storage.

This amount is necessary to conduct serious absolute superhuman research in this field -- but it is something very hard for a professor to obtain. We were very fortunate to have that computing at our disposal.

Gardner: Let's examine the commercialization potential of this. You're not only a professor at Carnegie Mellon, you’re a founder and CEO of a few companies. Tell us about your companies and how the research is leading to business benefits.

Superhuman business strategies

Sandholm: Let’s start with Strategic Machine, a brand-new start-up company, all of two months old. It’s already profitable, and we are applying the strategic reasoning technology, which again is application independent, along with the Libratus technology, the Lengpudashi technology, and a host of other technologies that we have exclusively licensed to Strategic Machine. We are doing research and development at Strategic Machine as well, and we are taking these to any application that wants us.

HPC from HPE

Overcomes Barriers 

To Supercomputing and Deep Learning

Such applications include business strategy optimization, automated negotiation, and strategic pricing. Typically when people do pricing optimization algorithmically, they assume that either their company is a monopolist or the competitors’ prices are fixed, but obviously neither is typically true.

We are looking at how do you price strategically where you are taking into account the opponent’s strategic response in advance. So you price into the future, instead of just pricing reactively. The same can be done for product portfolio optimization along with pricing.

Let's say you're a car manufacturer and you decide what product portfolio you will offer and at what prices. Well, what you should do depends on what your competitors do and vice versa, but you don’t know that in advance. So again, it’s an imperfect-information game.

Gardner: And these are some of the most difficult problems that businesses face. They have huge billion-dollar investments that they need to line up behind for these types of decisions. Because of that pipeline, by the time they get to a dynamic environment where they can assess -- it's often too late. So having the best strategic reasoning as far in advance as possible is a huge benefit.

If you think about machine learning traditionally, it's about learning from the past. But strategic reasoning is all about figuring out what's going to happen in the future.

Sandholm: Exactly! If you think about machine learning traditionally, it's about learning from the past. But strategic reasoning is all about figuring out what's going to happen in the future. And you can marry these up, of course, where the machine learning gives the strategic reasoning technology prior beliefs, and other information to put into the model.

There are also other applications. For example, cyber security has several applications, such as zero-day vulnerabilities. You can run your custom algorithms and standard algorithms to find them, and what algorithms you should run depends on what the other opposing governments run -- so it is a game.

Similarly, once you find them, how do you play them? Do you report your vulnerabilities to Microsoft? Do you attack with them, or do you stockpile them? Again, your best strategy depends on what all the opponents do, and that's also a very strategic application.

And in upstairs blocks trading, in finance, it’s the same thing: A few players, very big, very strategic.

Gaming your own immune system

The most radical application is something that we are working on currently in the lab where we are doing medical treatment planning using these types of sequential planning techniques. We're actually testing how well one can steer a patient's T-cell population to fight cancers, autoimmune diseases, and infections better by not just using one short treatment plan -- but through sophisticated conditional treatment plans where the adversary is actually your own immune system.

Gardner: Or cancer is your opponent, and you need to beat it?

Sandholm: Yes, that’s right. There are actually two different ways to think about that, and they lead to different algorithms. We have looked at it where the actual disease is the opponent -- but here we are actually looking at how do you steer your own T-cell population.

Gardner: Going back to the technology, we've heard quite a bit from HPE about more memory-driven and edge-driven computing, where the analysis can happen closer to where the data is gathered. Are these advances of any use to you in better strategic reasoning algorithmic processing?

Algorithms at the edge

Sandholm: Yes, absolutely! We actually started running at the PSC on an earlier supercomputer, maybe 10 years ago, which was a shared-memory architecture. And then with Bridges, which is mostly a distributed system, we used distributed algorithms. As we go into the future with shared memory, we could get a lot of speedups.

We have both types of algorithms, so we know that we can run on both architectures. But obviously, the shared-memory, if it can fit our models and the dynamic state of the algorithms, is much faster.

Gardner: So the HPE Machine must be of interest to you: HPE’s advanced concept demonstration model, with a memory-driven architecture, photonics for internal communications, and so forth. Is that a technology you're keeping a keen eye on?

HPC from HPE

Overcomes Barriers 

To Supercomputing and Deep Learning

Sandholm: Yes. That would definitely be a desirable thing for us, but what we really focus on is the algorithms and the AI research. We have been very fortunate in that the PSC and HPE have been able to take care of the hardware side.

We really don’t get involved in the hardware side that much, and I'm looking at it from the outside. I'm trusting that they will continue to build the best hardware and maintain it in the best way -- so that we can focus on the AI research.

Gardner: Of course, you could help supplement the cost of the hardware by playing superhuman poker in places like Las Vegas, and perhaps doing quite well.

Sandholm: Actually here in the live game in Las Vegas they don't allow that type of computational support. On the Internet, AI has become a big problem on gaming sites, and it will become an increasing problem. We don't put our AI in there; it’s against their site rules. Also, I think it's unethical to pretend to be a human when you are not. The business opportunities, the monetary opportunities in the business applications, are much bigger than what you could hope to make in poker anyway.

Listen to the podcast. Find it on iTunes. Get the mobile app. Read a full transcript ordownload a copy. Sponsor: Hewlett Packard Enterprise.

You may also be interested in: