Driving Business with Enterprise Ready AI and Open Source [Podcast]

Artificial intelligence and machine learning are revolutionizing the way businesses operate in the cloud. With the ability to process vast amounts of data and make predictions in real-time, companies are gaining a competitive edge like never before.

According to Gartner, the market for artificial intelligence (AI) software will reach almost $134.8 billion by 2025. Over the next five years, the market growth will accelerate from 14.4% in 2021 to reach 31.1% in 2025, considerably outpacing the overall software market growth.

In this video, we’ll explore how leveraging analytics in the cloud is providing companies with a valuable advantage and helping them adapt to the constantly changing market economics.

So sit back and get ready to learn how AI and ML in the cloud are driving success and competitiveness in the modern business world.

Summary of the podcast

Ian Brown, Head of Data Science at SAS, discusses the value of SAS software in providing analytics and AI capabilities globally. The availability and affordability of AI technology have made it cost-effective and have democratized AI technology.

However, open source requires governance and reliability in making business decisions, and Brown’s organization helps customers build solutions and address challenges such as silos and legacy architecture. They also help organizations make the most of their data while ensuring that data scientists’ models are being deployed and put into action.

The speakers discuss the benefits of modern machine learning in the banking industry, as well as the responsibility of custodians of data to be considerate of the data they provide to these models. The conversation concludes with Brown emphasizing the importance of data in training AI models, AI’s potential to be used in art and music creation, and his love for rock and alternative rock music.

Details about SAS Viya

In This episode of the Data Strategy Guru’s podcast,
I talked to Iain Brown, Head of Data Science at SAS Software.

His title on Linkedin says it all:
Head of Data Science | Adjunct Professor | Author | DataIQ Top 100 Most Influential People in Data

► You can join Iain on LinkedIn:
http://7wd.at/70c112

A great opportunity to have a chat with him on Data Governance, AI, ML, the Analytics evolution over the last 40 years, and how SAS is opening up.

Timing was perfect, with the SAS Viya launch on the Azure Cloud.

Read the Press Release: http://7wd.at/2be8e26

Want to try SAS Viya on Azure yourself:
http://7wd.at/e9f0ce

For the full details on SAS Viya and a free trial:
http://7wd.at/2c0cec9
– Full Analytics Life Cycle Support
– Transparent Insights
– Ethical AI Support
– Trustworthy Decisions

Detailed Summary:

• 00:00:00 In this section, Ian Brown, head of data science at SAS, speaks about the capabilities of SAS and its software in providing analytics and AI value across industries globally, using AI, data science, machine learning, natural language processes, and processing supply. Brown’s team focuses on understanding their clients’ use cases and challenges, creating processes that generate actionable decisions, and enabling organizations to leverage their data. Brown also discusses how the availability of AI and processing capabilities, along with increased data volume and variety, has democratized AI technology and made it cost-efficient, ultimately leading SAS to make their software much more open in allowing their customers to utilize an ecosystem approach.
• 00:05:00 In this section, the speaker discusses the benefits of open source and the need for governance and reliability when making business decisions. While open source provides innovation and flexibility, there is still a risk involved, especially with incompatibilities between versions when deploying models and using libraries between development and production. The speaker’s organization helps customers build solutions and ensure transparency, reliability, and explainability in their AI processes. They also address challenges such as silos, legacy architecture, and the need to democratize data access. Additionally, they help organizations make the most of their data while ensuring that data scientists’ models are being deployed and put into action. The evolution towards a more dynamic and cost-effective approach that is not reliant on certain skills or technical skills is underway, with the rise of low-code applications to help.
• 00:10:00 In this section, the speakers discuss how analytics is becoming more accessible to different roles within organizations, making it more cost-efficient and available to anyone. With SAS’s Sales Via now available on Azure Cloud and other marketplaces, it has become easier to access the capabilities used by some of the world’s biggest corporations. This move towards democratization and making analytics more widely available is helping organizations leverage their data and AI to the fullest. While there is still a risk of models not making it into production or failing to transition smoothly, decreasing barriers to entry and reducing dependence on expensive data science resources can only help innovation and move the industry forward.
• 00:15:00 In this section, the speaker discusses the importance of implementing an object-oriented modeling ecosystem that is similar to what open source does today, and how this can enhance API-driven connectivity. The speaker also recommends the rise of pre-trained AI and small- to medium-sized organizations partnering with data science consultant firms that can provide AI as a service type model. By deploying pre-trained models that have been previously used to solve challenges in organizations, AI can now be adopted and redeployed to benefit a wider range of entities. Moreover, the speaker provides two areas where AI can be used for good initiatives; one on AI creamy program for preeclampsia diagnosis for pregnant women, where AI provides feedback to clinicians, and the other on investigating risky communications between children to enable investigators to make an informed decision.
• 00:20:00 In this section, the speakers discuss the benefits of modern machine learning in the banking industry, where it can be used to personalize offerings and make better decisions about loans. While some are wary of AI replacing humans, the speakers argue that it can actually reduce risks associated with misusing information if used correctly. Transparency and accountability are key in ensuring that the decisions made by AI are understood and explainable. The speakers are optimistic about the potential for AI and encourage a pragmatic approach to its use.
• 00:25:00 In this section, the importance of data in training AI models is emphasized, as well as the responsibility of custodians of data to be considerate of what data they provide to these models and who they provide it to. The discussion also touches on the potential for AI to be used in art and music creation, and the conversation concludes with Ian expressing his love for rock and alternative rock music and providing information on how he can be reached online for further discussion on topics related to data science and artificial intelligence.

Full Transcript:

Yves Mulkers:
Hi and welcome to the Data Strategy Gurus podcast. Today we have again, very illuminated guest, Iain Browne. He works for a SAS Company and analytics Company already in the market since 1976. Hi and welcome, Iain. Maybe you can tell me a bit more and the audience about what you’re specifically doing at SAS.

Iain Browne:
Yeah. Well, thank you for invite me on. As mentioned, I work for the leading AI organization, SAS Software. Alongside that, I head up the data science team here, I also work still with universities. I’m a lecturer, adjunct professor at Southampton University, and fundamentally I run the division around data science. We focus across industries. I have a highly skilled, high technical group of people that I work alongside, much, much smarter and more accomplished than I am, and we help deliver value from AI, data science, machine learning, natural language processioning to all of our clients across the globe.

Yves Mulkers:
Yeah, you’re using the data science team to develop the products, or just see what are the use cases? How can I say you are heading up the data science team at the SAS?

Iain Browne:
Yeah. We’re blessed to have a wide range of capabilities from a software provider. As you mentioned, we’ve been in business over 40 years now providing analytics and AI value in software and services to organizations across the globe. My team obviously know the capabilities extremely well. We work very closely with our customers around understanding their use cases, their challenges.

We run something called an AI value accelerator with most customers to really hone in on what data they have, what challenges from a business perspective they face, and how artificial intelligence can actually be used to solve those problems, be it more standard supervised approaches where they have an outcome that they’re trying to predict, like churn, for example, within the telco space, where the customers are leaving, or risk, someone’s likelihood to default in the financial services sector, all the way through to some of those more unsupervised, more unstructured challenges of where you’ve got data in any form, from images, video, speech, all the way to text, and again, utilizing that information to generate insight, generate actionable decisions that an organization can then productionize, make best use of, and at the end of the day, service their customer.

Yves Mulkers:
Yeah. I mean, we know SAS a lot from the typical statistical models, very well known for these models. They have proven their value. We know they work. They’re widely used in insurance industries, in banking and so on. What did you see changing over the years, especially the last 11 years that you’re with SAS, and maybe some changes to the product? I know SAS has always been pretty close, there was a comment on that, but I recall when I went in 2018 to one of the Analytics conferences, they were slowly opening up and embracing open source. What’s your take on that? What are you seeing changing?

Iain Browne:
Yeah, that’s an interesting question to ask. As you mentioned, I’ve been at SAS over 11 years now. I’ve seen a lot of changes during my time. Actually, as an organization, we’ve offered capabilities around machine learning, deep learning models, neural networks specifically since the early nineties. These are things to SAS that aren’t necessarily new. What has changed, though, to your question, is the availability of these capabilities and the ability to process in a much more democratized way and a much more cost efficient way. With the rise of cloud computing, for example, it’s made it easier for organizations to leverage these types of models, to leverage the data, which again, has rapidly increased over the years, both in volume, variety and velocity, all the Vs of big data that we’ve seen over the last decade, that has enabled, and it is an enabler, organizations to start to do more with that.

Again, there are challenges with this, and as you mentioned, we’ve always historically, going back a decade ago, a bit more of a closed language. We’ve been proprietary, we’ve obviously had our own intellectual property that’s been embedded in the tool set, but we’ve had to change as well. We’ve made our software much more open in the later releases to enable our customers to leverage the most out of their ecosystem. As mentioned, open source is extremely powerful. There’s a lot of benefits around the community, around the flexibility, around the innovation that open source brings, but we still believe there’s more need for support and governance and reliability when a business decision is made. Again, that’s where we operate a lot with our customers today, is providing them that confidence that the decisions that they’re making are supported, are reliable, are governed, are transparent and explainable in terms of the AI processes that they’re running.

Yves Mulkers:
Yeah. You say governance, we talk about open source. I still recall somebody saying, “Yeah, I have about 50,000 programmers open source, the complete community that is contributing.” But on the other hand, like you say, there is a risk in there. I know as well that incompatibility between versions when you’re deploying models, when you’re using libraries between development and production, and that’s why you support, if I well understand, your customers as well in either even looking at the solutions, what you need to build best and how to do that, if I quite well understand, and as well on the governance side. So, how do you see the governance side, especially related to cloud and all the privacy related data which goes into the cloud?

Iain Browne:
Well, there’s been much more scrutiny on that. As an organization, we’ve dealt with regulations, legislation over the last 40 years, be it in the financial services sector or also in the healthcare sector, pharmaceutical sector. This isn’t necessarily anything new. What we have seen, though, is the rise of these applications across other sectors, other verticals, that are less regulated, but again, need to still be robust and governed in terms of their decision-making processes. Obviously, model ops, ML ops, AI ops, several different terms, all slightly different, has risen alongside this.

So, being able to turn these complex problems and models into decisions and generating intelligence from that decision is still a difficulty that a lot of organizations face, whether it’s siloed, whether it’s the issue around the right skills. As you mentioned, sometimes it takes a village to actually productionize these types of models, whether it’s through the data engineers, the data scientists, or then the decision makers that need to then action. Again, democratization goes both ways. There’s the ability to potentially do more, but at the same time it can bring more challenges to organizations to do the most with it.

They’re the areas that we’ve helped and supported a lot of our customers on their journey of evolution. As they’ve migrated to the cloud as well, being able to make that much more self-served and accessible is key to that process. You need to be able to get access to these tools when you need them at the right time. You don’t necessarily want a legacy architecture that’s running in the background 24/7 anymore. You need something that’s on demand. These are the challenges that we’ve addressed that our customers now are getting the benefits from as well.

Yves Mulkers:
You say that the typical architectures, and now you have more analytics on demand that definitely requires a different type of architecture, and that’s where you step in and advise among how to set up, because I see that when I’m talking to peers in the field and other architects, that’s still a big struggle. People are still thinking in traditional ways, even using cloud resources.

Iain Browne:
Yes. Yeah. Again, this is an evolution rather than a revolution. We’ve been used to these monolithic technology stacks that have been housed in-house within a lot of organizations over the years, and big IT team and infrastructure around that to support it. Again, as we move forward, we see the need for a more dynamic approach, let’s say. The ability to do this in a cost effective way as well is just as important. Storing data costs money, processing data costs money, and organizations need to make the most of that if they are going to make those right decisions.

No one really wants to have data scientists building complex, clever approaches, but then they’re not actually being deployed. It’s not beneficial to the data scientists, it’s not beneficial to the organizations. Putting in a root to live and a root to product is key for success on all sides. Again, they’re areas that we see organizations still struggle with today, but there is a lot of change in the sector as well to make that easier, make that more cost effective and make that process, in a lot of cases, not reliant on some of those skills or technical skills that we’ve had in the past. Things like loco no codes, applications do help that around the data scientist issue, but also around making it more accessible across business analysts, others in the business that would also want to ingest some of this insight and make the most out of it, right?

Yves Mulkers:
Yeah, so analytics becomes more available to a widespread of roles within an organization, more cost efficient, if I hear you talking. Now, since beginning of this year, if I’m correct, SAS Viya is now available on the Azure Cloud. Do you see a bigger reduction being involved into different type of solutions, more people reaching out to SAS Viya to build their analytics?

Iain Browne:
Yes. Certainly, being available on these marketplaces is much easier to get access to. This brings those capabilities, these capabilities, by the way, that are used by the biggest corporations in the world are now accessible to anyone. Anyone can get those through the marketplace. So, that’s a big change from SAS as well, that we’re bringing the power of AI that we’ve had over the years to anybody, and we can all individually download or get access to our own marketplace. I think that’s important around the message of democratization in a governed platform, and making it available. I mean, a lot of times also within data science, I’ve been through this myself on a number of occasions, it’s about testing, right?

It’s looking at ideas, failing fast, and being able to then decide on what you want to do to move forwards. Again, you don’t want too many barriers in your way to do that. You want to make that as simple as possible and as easy as possible. We’ve got access to some data, I can spin up in environments, I can test my theory. If it doesn’t work, I can turn it off again. If it does, I can then move it into a productionized process. Again, these are the types of things we see as important in the data science community, but also more important to organizations, leveraging the most out of their data and out of AI inevitably moving forward.

Yves Mulkers:
So, yeah, the possibility to quickly test a hypothesis, ramp it up, don’t be dependent on three months, six months set up of your infrastructure, of your models, whatever. Just try it. If you know this is proven, then you still can move it into productization, and having everything and the governance in place as it should be, that helps as well, and yeah, the shortage of skills as we typically call it for data scientists, for data engineers, for MLOPs engineers, AI engineers, whatever, call it type of role, pretty, pretty technical, do you think that that will even accelerate the type of solutions what we will build in the future?

Iain Browne:
I hope so. I think that the issue around data literacy, the data skills gap is well known, and that’s not just something we face in the UK. It’s something that’s global, a deficiency or gap between the skills that are needed and the skills that are available. So, again, bridging that gap and making that much easier for all is going to be part of this process. Again, removing barriers to any organization. AI I believe is important. I think there’s a lot of power in these approaches, in these methodologies that any organization can benefit from. I think reducing that barrier to entry, reducing the clients expensive data science resources, it can only help that, and help innovation. That’s the big stride that we’re making forwards in the UK government. On AI, innovation is at the core of our future policies around regulation, and making this available to all.

Yves Mulkers:
Yeah. What I see, I mean, you said business now has access to experiment themselves, ramp it up, try the hypothesis and so on. Isn’t there a bigger risk in even more models not making it into production? You typically see your model runs on a laptop and then you need to embed this in a process, have it constantly available, reliable, observable. But, yeah, there is still a risk. I saw that in the past with typical BI solutions, cost solution, that a great analyst has developed a model, but then they have a problem into moving it into production. Either it was a different technology stack development or just your laptop environment compared to the real production environment. What is SAS doing around that? There’s still a risk there.

Iain Browne:
Again, we’re talking about ecosystems that have evolved over decades, and I don’t think anyone would say as a solution to all of that. I do think greater connectivity is important. Something that we went through a number of years ago now at SAS was to open up our code base, but also base everything on an object-oriented modeling ecosystem, a bit similar to what open source does today. By that, you can then use things like API’s. API’s are very useful means of transporting information, connecting up devices and decisions together.

:
Again, I think that’s personally where a lot of this work will go around an API driven economy. One of the other ways the small to medium size organizations, we’re starting to see benefit from, is not just around accessing the tools and capabilities themselves, but the way they partner with other organizations. There’s a lot of data science consultants firms out there. We do consultancy ourselves that can actually provide this as a service type model. Either that’s hosted as a service platform, as a service, or actually delivering the answers or results of a model as a service with the skillsets that those organizations have.

:
I think that’s another way organizations should look to adopt this, as well as the rise of pre-trained AI, pre-trained models that are being delivered and deployed across cloud ecosystems. They’re already there. They work. They’ve been used to solve a challenge somewhere else, and then they could be redeployed again for challenges at your organization. I think there is more diversity here now than ever before around how AI is being used, but again, it’s driven by that cloud behavior, the ability to access things at any time in any way, and then deploying it back into system.

Yves Mulkers:
Yeah. They’re such complete different approaches as well, a lot of IT organizations have been used to in the past. I mean, I had it in mind of building those composable architectures, and that you can plug in something else, a new model comes along and new API comes along with new insights which performs better, and you embed it. You saw that now coming with Chat GPT, where a lot of platforms are already integrated Open AI into their interface and the benefits out of that. Typically, that’s what I see happening as well. If you say, well, can I name three use cases that wowed you in the last couple of years on analytics, that you say, “Wow, this is really great,” what kind of insights they now have which were not really possible before?

Iain Browne:
Yeah, most certainly. Again, I’ve been very lucky, I always think, in the role that I’m in, to have been involved in a lot of innovative approaches and applications over the years, as well as now my team who have been involved in many of the good AI for Good initiatives run at SAS. I think personally, some of the stuff we’ve done on the Data for Good side is widely shared and published around the University College Dublin, and the work we’ve done on their AI preemie program for preeclampsia diagnosis. Again, this is giving better time to recovery and understanding of delivery times for pregnant moms that are going through a difficult time using a mixture of natural language processing, as well as computer vision imaging to take all of that data and provide feedback to the clinicians, feedback to those that are practicing. From a doctor’s perspective, it puts that data in their hands in a much easier to use form.

Again, it’s those types of applications that bring across multiple sources of data and multiple different end users who traditionally may not have had any experience about… I think that’s really something special when that comes together and works. Again, we’ve seen similar things on child protection side, some of the stuff with leasing at the moment, enabling investigators to investigate risky communications between children. Again, that’s processing large volumes of unstructured data, large volumes of information that’s being creating in real time in most cases. Being able to understand that, filter that, and then provide those flags and those risks back to those police officers that can then handle that situation and then make a decision and action based on it.

Again, this is more of an augmentation. It’s not making the final decision by the model, but it’s providing us as humans more insight to make an informed decision on. I think that’s key as well around the message of AI. It’s helping us do our job better and those that need that information than it is replacing people. Then even on the risk side, if I was to give a third one, I have a credit risk type background. I used to work for a financial service provider before I joined SAS. Even in that space, we’re seeing dramatic increases in the application of more modern, more accurate machine learning methodologies that are still open and transparent and explainable, and actually in some cases reduce bias because they are able to understand the relationships and the data much better.

But the most important thing is they improve accuracy. It helps banks make the right decision at the right time for their customers whether to offer them a loan. If they do offer them a loan, how much loan is appropriate for that candidate. So, again, something that can benefit all of us is if organizations use that data to personalize their offerings to us to understand our behavior at the granular level. I think we can all as a society benefit and reduce risk from the bad use of information.

Yves Mulkers:
Just the other way around, I mean, the mistrust in AI, we’re all afraid that it’s going to replace… well, a lot of people are afraid that it’s going to replace the human. But the way you say it’s just going to counterfeit the risk which is in misusing either technology, misusing if you just follow the dialogues and use NLP to identify the relationships between what is being said, and that’s a new type of insight which was never possible at this scale.

Iain Browne:
Definitely. Obviously, I’m a bit of an AI optimist myself, as you can probably expect, but I do truly believe it. I believe this is a tool. It’s a tool for us to use, and it can be used in a good way, it could be used in a not so good way. I think the benefit here is for us to be aware and to be informed and to understand. I think everyone, that all of society is getting more informed about the benefits and also the risks of AI moving forward. You mentioned Chat GBT. I think the way that Open are AI doing it is excellent in informing society as to what is a possibility and also making it easy and accessible to everyone. Again, I think they’re the ways in which AI should be looked at. It shouldn’t be feared, it should be embraced, but an understanding is needed from everyone as to what it can do or what it should and shouldn’t do.

Yves Mulkers:
Yeah. I think just really showing the examples and the benefits and that you need to transition. I just read a white paper on one of the Harvard professors that used Chat GPT to do the exams. Well, it’s caught about 80%. It was not good on mathematics, but just the reasoning and everything behind that, it brought me back to one of the debates I saw with IBM Watson back in the days. It’s been five, six years, but if you see at what speed this technology has evolved, it’s amazing. Just like you, I’m pretty positive in augmenting our capabilities and being more productive in applying it.

Iain Browne:
We’re still at early stages of this as well. Even with Chat GPT, the architect of the generative algorithm it’s based upon is going to be re-released, there’s a new GPT four due to land this year. A lot of this we’re still even early stages of. I think that’s exciting in its own right. There is more to come from all of this. Again, the importance is to have that thought around governance and the expectation of what it should and shouldn’t do, and being transparent. All organizations need to be transparent. If they’re making decisions that they can’t explain to themselves, I would say that is a risky use of AI.

If you can’t understand it yourself, how is anyone going to understand it when you come to make that decision? That’s where the accountability, I believe, really lies as well. It’s who is making that decision and are they accountable and do they feel accountable for that decision given the approach that they’re using. All of this comes into the mix. I think we’ll see evolutions of this over time and how regulations unfortunately catch up to it as well. A pragmatic approach around the use of this should also be considered at this time as well.

Yves Mulkers:
Yeah, I don’t think that there is so much different than that is already related to AI. Still, I mean, traditional systems, legacy systems, it’s not always easy to explain how they come to the insights as well, because it has been running, everybody that has knowledge about the systems has left the company, has retired, whatever. For me, it’s kind of adapt, to it and just at this scale the AI can explain itself if we program it to do in such a way.

Iain Browne:
Yep. I think the conscious decision around how you feed the AI moving forward, it’s a phrase that I’m sure you’re familiar with and we’ve used for many years, garbage in, garbage out. The data that’s being used to train these algorithms is crucial to all of this process. Again, there should always be that consideration around what are you using to build the model. I think all of us should be considered of what data we provide these models as well, and who are the custodians of data, and what that data is going to be used for. I think that will change everything.

I think that next generations, they’ll value more the data they have and the data they provide. I think this will affect how AI is adopted and used as well. If there’s a mutual benefit to the user and the receiver, I think that will flourish. If one side gets more benefit than the other, I think that’s also at a detriment to AI as well. All of these things for consideration in the future.

Yves Mulkers:
Yeah. You see AI being used for art and creating images and so on, even some experiments in creating music, which is one of my favorite topics next to data and AI as well. Maybe to wrap up this conversation, what is your favorite band of or favorite type of music, Iain?

Iain Browne:
Wow. It is a good question. Unfortunately, unlike yourself, I’m not a great musician. I’ve tried over the years, but I’m always into rock, alternative rock in my youth, the likes of Foo Fighters, Red Hot Chili Peppers, Muse. Those types of bands I think I still… maybe considered dad rock now, I don’t know, but they’re the bands that I always used to listen to, I still follow even to this day.

Yves Mulkers:
Well, they’re still performing. We have a festival in Belgium called [foreign language 00:26:56], and I think every kind of year they come and present. Even this year Muse is coming again, Red Hot Chili Peppers. So, still well known, and you see still the traditional stuff sticks with a lot of people and talks to a lot of people. Iain, if people want to reach out to you online, how can they connect with you?

Iain Browne:
Yeah. I’m always open to connecting. I can be found on LinkedIn. I have a quite active Twitter presence. As of the start of this year, I’ve also picked up my roots in blogging. I used to blog quite regularly. I’m now using Medium as a platform to do that. So, I’m always interested to connect, share ideas and share thoughts on the topics of anything related to data science, artificial intelligence.

Yves Mulkers:
Okay. Iain, thanks for very much for this conversation. It was great talking to you. Yeah, let’s keep connected online.

Iain Browne:
Definitely. My pleasure. Thank you for inviting me on.