Episode 62: A Conversation with Atif Kureishy

In this episode Byron and Atif discuss AI, deep learning, and the practical examples and implications in the business market and beyond.

-
-
0:00
0:00
0:00

Guest

Atif Kureishy is the Global VP of Emerging Practices at Think Big, a Teradata company. He also has a B.S. in physics and math from the University of Maryland as well as an MS in distributive computing from Johns Hopkins University.

Transcript

Byron Reese: This is Voices in AI, brought to you by GigaOm, I’m Byron Reese. Today my guest is Atif Kureishy. He is the Global VP of Emerging Practices, which is AI and deep learning at Think Big, a Teradata company. He holds a BS in Physics and Math from the University of Maryland, Baltimore County, and an MS in distributive computing from the Johns Hopkins University. Welcome to the show Atif.

Atif Kureishy: Welcome, thank you, appreciate it.

So I always like to start off by just asking you to define artificial intelligence.

Yeah, definitely an important definition, one that unfortunately is overused and stretched in many different ways. Here at Think Big we actually have a very specific definition within the enterprise. But before I give that, for me in particular, when I think of intelligence, that conjures up the ability to understand, the ability to reason, the ability to learn, and we usually equate that to biological systems, or living entities, and now with the rise of probably more appropriate machine intelligence, we’re applying the term ‘artificial’ to it, and the rationale is probably because machines aren’t living and they’re not biological systems.

So with that, the way we’ve defined AI in particular is: leveraging machine and deep learning to drive towards a specific business outcome. And it’s about giving leverage for human workers, to enable higher degrees of assistance and higher degrees of automation. And when we define AI in that way, we actually give it three characteristics. Those three characteristics are: the ability to sense and learn, and so that’s being able to understand massive amounts to data and demonstrate continuous learning, and detecting patterns and signals within the noise, if you will. And the second is being able to reason and infer, and that is driving intuition and inference with increasing accuracy again to maximize a business outcome or a business decision. And then ultimately it’s about deciding and acting, so actioning or automating a decision based on everything that’s understood, to drive towards more informed activities that are based on corporate intelligence. So that’s kind of how we view AI in particular.

Well I applaud you for having given it so much thought, and there’s a lot there to unpack. You talked about intelligence being about understanding and reasoning and learning, and that was even in your three areas. Do you believe machines can reason?

You know, over time, we’re going to start to apply algorithms and specific models to the concept of reasoning, and so the ability to understand, the ability to learn, are things that we’re going to express in mathematical terms no doubt. Does it give it human lifelike characteristics? That’s still something to be determined.

Well I don’t mean to be difficult with the definition because, as you point out, most people aren’t particularly rigorous when it comes to it. But if it’s to drive an outcome, take a cat food dish that refills itself when it’s low, it can sense, it can reason that it should put more food in, and then it can act and release a mechanism that refills the food dish, is that AI, in your understanding, and if not why isn’t that AI?

Yeah, I mean I think in some sense it checks a lot of the boxes, but the reality is, being able to adapt and understand what’s occurring, for instance if that cat is coming out during certain times of the day ensuring that meals are prepared in the right way and that they don’t sit out and become stale or become spoiled in any way, and that is signs of a more intelligent type of capability that is learning the behaviors and anticipating how best to respond given a specific outcome it’s driving towards.

Got you. So now, to take that definition, your company is Think Big. What do you think big about? What is Think Big and what do you do?

So looking back in history a little bit, Think Big was actually an acquisition that Teradata had done several years ago, in the big data space, and particularly around open source and consulting. And over time, Teradata had made several acquisitions and now we’ve unified all of those various acquisitions into a unified group, called Think Big Analytics. And so what we’re particularly focused on is how do we drive business outcomes using advanced analytics and data science. And we do that through a blend of approaches and techniques and technology frankly.

In my group in particular, we’re focused on AI and deep learning, so we’re working with large enterprise customers and enabling them to let’s say, do better manufacturing, reducing scrap and improving yield for instance, and how to apply computer vision for instance, to that fabrication process, is one example.

So, a lot of the challenges that we see, or a lot of the progress we see in research and applied research, we definitely consider those, but we have to understand how that would work in an enterprise setting that may be constrained in different ways, that may have different types of talent requirements, technology requirements, policy regulatory requirements. All of those things have to be considered.

So the listeners to this show love to hear about use cases, successes or failures when trying to apply AI to a business problem, to effect an outcome. It sounds like that’s your bread and butter. Can you tell us a war story or two of something you’ve done, a problem that was brought to you and how you applied data to it, how you applied AI to it, and what the outcome was?

For sure, as you can imagine, some customers are very vocal and open and transparent about when they engage with outside consultancies and there’s various motivations for either talking about those engagements or not. One motivation for talking about that is to definitely be seen as innovators and driving success and new ways of thinking, to the benefit of their own organization or their customers.

Now, on the other side, sometimes those capabilities may be a distinct competitive advantage and so customers are very hesitant to even allow companies to mention that they’re working together. So, some customers that have allowed us to speak in very great detail about the engagement that we’ve done with them, it’s in the areas of retail banking, and so, we kind of see two major flavors of engagement.

One is really around reducing cost, and that could be to automate aspects of business processes, so for instance, for loan origination or when we go to apply for a car loan or a mortgage, we have to go into a bank per se, and fill out applications and share deeds and salary statements and other types of information that allow the underwriter to make a risk-based decision on whether to issue the credit or not. And so, machine and deep learning can be applied in actually all aspects of that process so, number one, just the origination and onboarding of those various documents, and doing the checks of completeness and being able to understand the associated documents, can be done by a machine. I’m not talking about just OCR kind of very straightforward MLP tactics, there’s aspects of perception in the sense of, not all salary statements look the same or check stubs look the same or titles and deeds look the same, and so being able to understand, have a machine understand through different computer vision techniques, what those documents are and classify them, and then be able to do certain types of processes.

So, on the onboarding, once you move into the risk and underwriting decisioning, being able to express a very high dimensional feature space of your risk portfolio, and not just use pulled risk constructs, which is typically what banks and insurances do today, but really moving towards micro-segmentation where you’re looking at an individual, you’re looking at their portfolio in particular and potentially understanding what’s the risk for default and what have you by looking at the entire corpus.

But you’re not looking at it in the aggregate, you’re looking at that individual and applying very distinct wide and deep networks to make those decisions. Then finally on servicing and operating a loan, a lot of work that we’ve been doing is in the fraud space, so for any capability out there, you’ll always have adversaries that are looking to exploit it for nefarious purposes, and that could be insider threat or that could be external actors. So, a lot of the work is in different types of fraud mechanisms, in insurance, in money laundering, in credit cards and digital payments.

What’s fascinating is that deep learning has shown great success in again being able to take very large feature spaces, very large dimensionality and be able to detect different types of signals around anomalies. And that’s the essence of what transaction monitoring systems do, let’s say, in anti-money laundering. That’s the essence of what the credit card authorization systems do when they look to approve or deny a credit card transaction, and that has been really fascinating for us to see those capabilities and work with our customers to push those capabilities into production. What that means is you have to consider all of the privacy, legal regulatory compliance issues, like GDPR, like fair credit lending, like the various policies that you have to align to. That’s where a lot of the explain-ability or interpretability aspects come into play around these black box models.

Please talk a little bit about that last bit, because, a lot of times those systems use unsupervised learning and they’re looking at some data space and they’re clustering and they’re finding pockets in that data of people in the past who have defaulted or were fraudulent or what have you. And then the explain-ability may simply be: you’re in a cluster of people that are prone to default. How do you thread that needle and make it a robust system that’s also, like you said, compliant and explainable and all the rest?

A lot of times what you’ll see, I mean very rarely are we talking about one model, right? These are usually ensembles or stacks of models, in some instances that could go into the hundreds, okay, so being able to co-ordinate all this is the key critical aspect, but there’s different techniques for explain-ability. So for instance, some enterprises will use traditional machine learning, decision trees, boosted decision trees, to express certain concepts and that’s well understood by, let’s say, regulators. And that they have techniques for documenting and expressing how certain decisions are made, and then they’ll leverage deep learning to essentially do latent feature extraction, so, derive new features from these neural networks, that will then be put into a traditional machine learning model, so that they don’t have to worry about how to make the neural network itself, explainable.

So there’s a certain class of approaches and pros and cons with doing that. Another approach, which is something that we’ve done, is how do you drive interpretability into the deep neural network itself? And so there’s a whole science of this space, and so we’ve used different types of open source frameworks out there that allow us to do interpretation, and perturbative techniques that essentially say, “If I include noise into this model, and be able to express the amount of variants and the amount of contribution that these features have, ultimately to an output, let’s say a classification system, then I can understand which features of the model are influencing that output the most.” And that’s the technique to then say, “Which features in a probabilistic way, are the ones that are contributing the most?” And so that’s important. Let’s say if you’re making a determination of fraud, going back to our example, that you can cite that [a decision was made] because you’re associated with this actor or this certain amount, or coming from this geography or other types of attributes about the transaction. That’s the level of expression you need to pass GDPR or some other sort of audit and policy definition.

So, I’m curious, I really think about this particular question a whole lot, I guess because it’s so germane as well, but think of a search engine, and one of the big ones, and you go to them and say, “I sell ‘widgets,’ and for the search ‘widget sellers,’ I’m number 8 and my competitor’s number seven. Why are they higher than me?” That’s a hard question to answer, isn’t it? Even with all the techniques you have, because there may be 1,000 variables at play, right?

That’s right, and you know, maybe for the audience I should just kind of cite just a little bit about my background. The first half of my career was actually focused on two very scientific agencies. One was NASA, and as part of that, with the Earth observing system, we launched a satellite called TARA-AM1 and that’s had many sort of scientific instruments that started to determine global warming and other sort of phenomena of the atmosphere and what have you.

And then I worked very closely in the intelligence community, and the conversation of artificial intelligence, and built sophisticated systems and capabilities to serve the nation in different ways. At the crux of those were very mathematical and engineering oriented missions, and being able to take vast arrays of data and help with decisioning, and what was clear is that, just these techniques of interpretability are never enough, and so, there’s all these concepts, that eventually lead to traceability, right?

So when a system, or frankly when a human makes a decision, you want to be able to express what’s influencing those decisions and so, that’s naturally traceability. And so data lineage and data pedigree, all come into play and being able to ensure that there’s trust in the system. So, information security and authorization are also key aspects and every time that that data has been touched or manipulated or models have applied algorithms on top of that, you want to understand that at some level.

Now, clearly in a very complex system, you don’t want the user of that system to be inundated with all this complexity, but, allowing that information to come forward in the right way, is going to be more and more critical, because what’s really going to allow AI to thrive or not, is going to be about trust, right? Because if all of this sophistication and understanding is being pushed into machines, humans ultimately need to understand aspects of that and trust that system, that it’s going to do the right thing, on behalf of a certain decision, and decisions could include potential loss of life, so, there are risk-oriented types of missions, that have to obviously be considered.

So how do you think all of that’s going to shake out? Do you think we’re going to have something like GDPR in the U.S.?

I think fundamentally, it’s inevitable. I think as this technology and AI becomes more pervasive and immersive in our lives, we’re going to experience the benefits of that, and frankly we’re going to experience the downsides of that. So having the right levels of control and considerations around privacy and ethics, those are always an afterthought for a lot of organizations, but those types of enabling capabilities will be more prevalent and more important as we move forward.

So your company’s primarily interested in helping enterprises adopt this technology to drive a certain outcome. What enterprises would you advise the time is right for you to start thinking about this? Is there a size threshold or a market sector, or, if somebody who’s listening now, do they say, “Is it still early for my company to start tackling all of these issues?” Or is making better decisions, never something you want to put off?

I think that’s a great question. The reality is, I think, every industry has the opportunity to take advantage of AI, but that doesn’t mean every organization is in the right position to capitalize on that. And so what I mean by that is, there are degrees of sophistication that are needed to really embrace this technology, but, there are key building blocks that have been solved along the way that have allowed us to do this quickly and rapidly. So, obviously the accessibility of the cloud gives us the compute storage network to be able to onboard these capabilities very quickly, but that’s not always an option for all enterprises, for data gravity issues, for cost considerations, what have you.

But if you were to ask me, which industries are the most ripe? I think it’s definitely the industries of manufacturing, of financial services and insurance, communications/media entertainment, telcos and automotive in particular. And the reason I say those things it two-fold. One, they all have heavy investments and digitization efforts that have been going on, and data is the fundamental of machine and deep learning, so, having access to that is of course important. But two is that they’re all being disrupted in different ways, and so disruption is a key motivator for these organizations to really think differently about their business model.

So a lot of times we’d like to believe that organizations and enterprises have really smart people at the top who think about futures and how to take advantage and serve their customers or their constituents and shareholders in the best way, and no doubt that that’s true, but what motivates them even more is when there’s a competitive threat. We know what’s happening in the automotive and autonomous space, and so there’s so much spend and investment across all of the OEMs and Tier 1 suppliers, to really go after connected mobility and autonomous, in a much bigger way.

And of course at the center of autonomous is really deep learning at scale, and so, that’s why there’s so much pace of spend and really innovation occurring in those industries, and the same that’s happening in financial services, the same that’s happening in comms/media, and what have you, and then a close second for us in particular, is probably healthcare and life sciences and retail. A lot of folks may say that retail life science with healthcare is probably where a lot of the spend is happening, and that’s true, but I would say, that’s not where a lot of the innovations are happening, and again the reason is because a lot of the right compliance and ethical and policy aspects within the healthcare and life sciences space is modulating those successes.

Do you know the story of the guy in Japan that wrote the program and built the device to sort cucumbers, do you know that one?

I don’t, no.

So his Mom and Dad have this cucumber farm, and he noticed his mum would spend all day sorting cucumbers based on 4 factors, by their size or shape, their color and how spiny they are, because just some are worth more than others, and he looked at it and he said, “you know…” Basically he used Tensor Flow and a Raspberry Pi and Arduino and he built this thing that sorts these cucumbers by size and shape and color and all that. Now, that’s a fun example, it’s a true story, and that’s probably, you might have to do some math to figure out if that was a good use of that person’s time, because of the scope of the problem.

But I guess that’s what I’m trying to get at, is, if I’m a 100-person company and I manufacture something, should I wait until there are products that are made that I can buy and implement that use these technologies? Or is it to the point that the platforms are easy enough to use, that company X can actually solve its own problems without packaged software or SaaS solutions or something, does that make any sense?

It does and in fact, going back to the cucumber example, one of our customers that I’ve been talking very closely with does something very similar, except with potatoes. We talked about essentially, how do you apply these same techniques to help reduce the scrap associated with potatoes? And you would think, potatoes from farm-to-table or to your neighborhood McDonald’s is a fairly straightforward process. It’s actually a very complex one, and there are already techniques of computer vision within that manufacturing space, but those edge devices are very rudimentary.

So really the opportunity for applying more sophisticated AI, is to understand all of the upstream implications of when and how you buy potatoes, with the downstream implications of how much waste it produces when your potatoes are bad, or irregular shaped, or are diseased, and how do you influence potentially soil composition and other agronomy types of consideration. That’s really where I think the opportunities are.

Now, is that reachable, per se, for a small to medium sized business? It’s difficult for me to answer because we don’t typically engage with those types of customers, but I think between the framework, the cloud-based environments and the knowledge that’s out there, it’s becoming more and more accessible to get GPU compute, to be able to build data pipelines of imagery types of data, to label that and to be able to build and train models and deploy that. But it’s still somewhat complex.

So I think where the biggest breakthroughs in AI for the enterprise, I believe is in applying AI techniques to model building, data acquisition, data labelling, all those things that make it really hard to do this, in a much more agile and expeditious way. And so, if we’re able to acquire and land data and label it and build models in an auto-ML sort of fashion, do that hyper-parameter optimizations and sweeps, and be able to deploy this seamlessly into an operational environment, and do hundreds of models like this, very rapidly, then I think that’s where really we’re going to see the pace in innovation in AI pick up.

So are there any rules of thumb you use? Let’s say for enterprises, they are large and they manufacture, and they have logistics and they have all of this stuff. When they say, “We want to start using these techniques to improve our business.” How do you start looking around? Do you say, “Well show me… let’s talk about all the data you have, or let’s talk about where you think the most waste is, or let’s talk about things that look like games, because AI can play games well, because it constrains universes…?” How does that conversation begin with somebody who says, “Where do I start?” And again we’re talking about a large enterprise that can do anything.

Yeah, another great question. Really we see two types of opportunities: one, where you have an executive buyer and that executive buyer usually is in the state of: “Hey, my CEO or senior leader, told me that we need to be an AI-first company, and we need to go after machine and deep learning, and I need to do that, but I don’t know what to do…” They’ll never say, “I don’t know what to do,” but “I don’t know how to take the first step.”

So there’s this specific orientation that you’re going to engage with a buyer like that, and then the second type of buyer is more in the R&D space, which is probably CIO or CTO group where they know that they strategically have to go after this technology, and they’re very familiar with the frameworks and infrastructure and opportunities and they keep up with research and apply research, and they’re so fascinated by what’s going on, but how that translates into the lines of business and what problems they’re trying to tackle, and how, is a little bit distant.

I should mention that that buyer is very intimate about what’s the business comparatives, what’s the key challenges, where the investments have been, what are the capital investments, what are the budget constraints, and who’s doing what. And the reality is, you need to bring those two things together, and very rarely do both of those exist in the same organization, because if it did then they wouldn’t need folks like us. They would be driving success on their own.

So typically what happens for each one of those buyers is something that looks like a diagnostic that says, hey, let’s try to understand… what are the capabilities that you have today, what does that look like in terms of a maturity level? And you’re going to need, let’s say these seven or eight things, and let’s walk through and understand and assess how strong or weak those capabilities are.

And then, the second part of that, once you have that baseline understanding, you want to be able to tackle a high-value problem. And so, there’s many ways to do that, you can interview different folks in the business, you can look at 10ks or strategy documents and derive or understand where the key efforts and opportunities are. But, ultimately, you have to understand that deep learning is going to help solve this problem, and so, in the areas of video or imagery or natural language understanding, clearly there’s areas that machine and deep learning can be applied and then through that sort of diagnostic, you’re going to understand [whether] you have all the data assets in full fidelity to be able to do that, if you have access to infrastructure, whether that’s in the data center or on cloud, [whether] you have talent that could be able to build and harvest.

Now one thing I didn’t mention and I should have at the outset: it really matters in the acquisition strategy of how you want to acquire these capabilities. What I mean by that is, you could go buy it, and you mentioned that in terms of SaaS APIs or SaaS providers, and that’s a very low risk approach, but the reality is that it’s fairly inflexible and it’s not necessarily game-changing because all of your competitors most likely have access to that same SaaS API or SaaS provider. So if you want to buy capabilities like that, there’s a certain approach for that. If you want to build capabilities, which obviously [means] higher investment, higher cost, but higher risk. But the opportunity, it can be high impact and game changing, and so there’s definitely a specification and there’s definitely something in between, more of an integrated sort of model. So you very quickly understand what the organization is trying to do, whether it’s build, integrate or buy, and then you tailor a plan and approach of how to do that in a very incremental value-oriented type of way.

How pronounced is the talent shortage of people that are up on these techniques, because we hear that it’s like pro-sports level salaries, and it’s a huge amount of competition for the most qualified people, and that a few large companies have most of the talent locked up. Is all of that true or is it changing, what’s your experience?

I think there’s some truth to that and I think it’s changing. What’s occurring is, yes, there’s a density of talent that’s being consumed by, let’s say, the digital giants, and a lot of those companies as we know, are really companies rooted in advertising, and so they can do all of these adjacent innovations and be creative, because what’s keeping the lights on is something fundamentally different.

So there’s a great deal of creativity that can be applied and pretty interesting innovations. Now in an enterprise setting, you have a distinct initiative/objective, and you have to work within the constraints of the organization. No doubt you can be creative, but in order to bring talent into that environment, you have to have a lot of things in place, number 1 to attract that talent, and [then to] retain it. And so, where organizations and enterprises are now is, a lot of them have started up their ventures group or their innovation group and have a presence in all of the tech hubs in the Bay Area, in Boston, in various places across the U.S., and various places across the world.

So that’s their mechanism for tapping into that talent, but as you said, the compensation models are very different. In these Bay Area companies, they’re highly leveraged models, which means, accrued equity, stock rents and what have you, and a lot of the enterprises out there, that compensation package looks very different. And so, they have to fundamentally re-think the way that they attract talent, and retain it and compensate them, and give them an environment that gets them excited to come to work every day, and solve hard problems that they’re tackling.

I had this theory that if everything we knew about AI were frozen and there were no breakthroughs, nothing, that it would still take us a decade or more, two decades to apply what we know now universally throughout the whole economy. Does that have a ring of truth to it, to you?

It does, I mean in the sense of, when we think about how data and analytics-oriented all of the premier companies that are out there [are], we assume that they’re so very fluent, their leaders are completely in tune, they’re investing in the right ways, and they have these foundational capabilities that are very robust and they’re building upon it.

And the reality is, it’s mixed. There’s groups within those organizations that are very successful, but there’s invariably, culture aspects and technology types of aspects and other operating model aspects, funding aspects that all get in the way of being able to always do the most innovative things at the right time. And so, I agree with you in the sense of, if we just pushed pause, and we didn’t advance as a community in AI, we could probably still go 15 years to allow a lot of the organizations to really catch up, invest and build these capabilities, I would agree with that 100%.

The Gross-world-product, the sum total of the GNP of every country is, in really rough terms, $100 trillion. And the value of everything on the planet is about $400 trillion. And I once sat down to try to figure out: of that $400 trillion, how much of that wealth was created by the Internet. Because you know, it gave us Google and Etsy and Amazon and eBay and Airbnb and all that, but it also gave us a million small businesses and it helped… it did all of these things and it’s a really hard question to answer, but I came up with like $25 trillion of wealth has been created by the internet.

When I think about the consumer internet—the web is 25 years old—and so, in 25 years, just by connecting computers together with a common protocol so they can communicate, and it’s better communication, we created this enormous amount of wealth.

Would you feel like that AI is going to create multiples of that? Is AI like as big as the Internet in terms of economic impact, or is it twice as big or 10 x or half as much? You have any kind of a gut feel of that economic impact of these technologies?

I’m not sure, and the reason why I’d say that is because, even when I think about the valuations of companies today, it’s really predicated on cash flows in the future and anticipating what that looks like. When I think of AI, we’re really at the beginning of understanding what we can do with these different types of technologies and approaches, and that’s going to unlock, obviously, high-value companies that are able to monetize that in new and creative ways.

So the wealth that that will generate again for me is difficult to ascertain because there’s just so much volatility in potentially which way we could go and what that’s going to look like. But it is going to be transformational in the sense of, just like how digitization and the internet as you put it, has really changed the way we work and operate and live our lives, I think AI’s going to do something very similar, and it’s going to affect our everyday lives as consumers and citizens, and as employees and workers, and providers of different services and products. I think it’s going to touch every aspect of our lives.

You know if you just say something like, “Every business on the planet can be 10% better if they intelligently apply technology to data and listen to it,” and if that were true, then you’d say, “wow, in a $100 trillion economy, that’s $10 trillion of increased profitability.” And then you can put whatever multiple on that you want, and that doesn’t account for anything new or anything amazing.

It just says, “We live in such a poorly optimized world, that it’s like the sky’s the limit.” I only ask because you’re closer to it, because this is your bread and butter. So are you, overall optimistic or pessimistic about the impact of this technology on society as a whole (moving the lens out from the enterprise) its impact on employment, on health, on warfare, on everything else, when you net it all out, is this good for humans or is this bad for humans?

I’m definitely an optimist and [believe] that this is good for all of us. Before we leave the last topic, I will say, what’s going to enable all of those enterprises, let’s say, to be 10% more efficient is the barrier to entry. As that barrier to entry continues to lower, then of course that opportunity becomes more real.

I think what the threat is that AI can potentially pose to us, is sort of this, ‘winner-takes-all’ proposition, in which companies that can harness this technology invest in it, and really drive incredible innovation very rapidly. A lot of the competitors or other providers out there will get left behind very quickly. And so, you’ll have this consolidation of a great deal of capability and understanding in the hands of the few. And when I say a few, it doesn’t mean a few individuals, it could be a few corporations, a few industries or what have you.

If we can ensure equities and remain balanced where everyone feels that they still have opportunity, either as an individual, as a society, as an enterprise, then I think we’ll continue to thrive. When it becomes very, let’s say, asymmetric, and there’s distinct haves and have-nots, as that disparity is growing bigger, wealth being one of them, then I think it’ll be met with challenges that other implications will come into play.

All right, I’m with [you on] that, but you began that whole thing by saying you’re an optimist, so how do you think it’s all going to play out?

Yeah, so I think the value and potential of applying essentially massive scale, it’s been done in the sciences, if I go back to my undergrad, when we all learn math, not all of us have learned math and physics, but when we can understand and express natural phenomena, using science and math and technology, obviously there’s a great deal to benefit there.

Where I see the same occurring, let’s say in an enterprise setting where essentially [they’re] applying math at scale, and now we have the computation and framework to do that in much easier ways, no one would argue that that’s the right thing to do. So I’m definitely an optimist that it’s going to better the lives of all of us globally, but again I think if my time in the intelligence community taught me anything, it’s that there will always be actors that will try to exploit it for other purposes. And so, always being mindful of that sooner rather than later, I think it’s going to benefit all of us.

All right, well I think that’s a great place to leave it. I want to thank you for a far-ranging and fascinating conversation.

I appreciate it and enjoyed our conversation, thank you.

Byron explores issues around artificial intelligence and conscious computers in his new book The Fourth Age: Smart Robots, Conscious Computers, and the Future of Humanity.