JN: What role do you think academia plays in AI innovation?
Dr. Benson: Academia plays a very important role. Naturally, academia is where a lot of the theoretical and experimental work has been done in AI over the years. However, it’s only one piece of the puzzle, and it’s the interaction between academia and industry which is really fueling AI innovation and the explosion of AI in both consumer and enterprise products we’re seeing today.
What academia lacks, and what industry is inundated with, is real world data. About ten years ago we started experiencing significant growth in machine generated data and the ability to collect and store more data volumes than we had ever been able to before. This created a treasure trove that could be used to power all of the theoretical machinery that had been developed in academia, and realise real recommendations and real predictive applications.
The increased availability of data influenced the theoretical aspects of AI and created this symbiotic relationship between academia and industry.
JN: How do you think educating the talent of the future has furthered your career and SnapLogic as a business?
Dr. Benson: Working both with SnapLogic and continuing my academic work has been incredibly enriching for me and I’d say it’s been a mutually beneficial relationship.
Industry and academia have very different cultures. You’re driven by different goals in academia than you are in industry. I remember when I first started at SnapLogic in 2010, I was asked to write some code for the product, and I was incredibly nervous about getting that code in front of my colleagues, because in academia you just write code to get stuff done, you’re not writing it with a specific audience in mind or to be reviewed and scrutinized! Thankfully it actually fared well. I learned a lot. I learned how to adapt to that type of coding environment.
I’ve brought those experiences into my classes. Some of the things I teach now, especially in distributed systems and reliable systems, are things that you maybe wouldn’t see in academia. I’m teaching some of the realities of building production software and hopefully better preparing them for a future in industry.
JN: What interests your students most about future technology developments and specifically AI?
Dr. Benson: I think it’s the expansiveness of future technology. They’ve grown up with ubiquitous computing at their fingertips, connectivity and the idea that a computing platform is the at the center of the world. It’s a totally different worldview to the one I had when I sat down with an Apple ][+ in 1980. Back then it was me and a computer. It was what the computer could do as a single entity, it wasn’t the collective brought on by the Internet that I think the upcoming students view computing as.
I think that the potential that students see is that there’s this massive, interconnected worldwide computing platform and that, when framed in that way, everything’s possible.
Even when it comes to AI and machine learning, which of course takes years of study to become proficient in, there’s no question that students are excited about this world-changing potential.
JN: How far along the AI journey do you think companies in the tech industry are?
Dr. Benson: The answer is that different companies have different levels of development. At SnapLogic we’re fortunate because, as I’m an academic, and because of my background, we’ve been able to bring in students with machine learning backgrounds and accelerate what we’re able to do. I’m not sure all companies are in that position yet.
But there’s huge interest in industry, there’s huge interest in academia. There’s data science, analytics, and a pipeline full of students coming through with a strong interest in AI. Students who will eventually go out into industry and so we’ll see increased activity in AI products in the coming years.
Read more: FICO Chief Analytics Officer: How to build AI that averts disasters, not creates them
When it comes to leaders in the field, we’re talking about the big companies like Google and Facebook. Google, for example, couldn’t recruit and hire enough people with AI and machine learning backgrounds, so it trains and teaches in-house. It has a whole organisational department where it brings people in and they get them up to speed. So, obviously if you’re a company like Google or Facebook you have the resources to do that kind of training.
JN: Is AI really going to revolutionise the business world?
Dr. Benson: I absolutely think the applications of statistics and classification and what makes up machine learning, and eventually AI, is going to have a huge impact.
Are we at an inflection point? These things are hard to predict. I’d say that we’re going to see year-to-year incremental updates and improvements. At SnapLogic, we’re doing just that with Iris.
On a global level, as more and more companies add more and more machine learning, things are going to get easier, things are going to begin to think for you to a certain degree.
But it’s not going to be that tomorrow the AI robots are on, we sit back, our drinks are delivered and we don’t do anything. It’s going to be a progression of improvements, with AI augmenting what we do as humans, and I believe that it’s going to continue at a steady pace.
JN: Does AI pose a serious threat to jobs?
Dr. Benson: I think you can point to some areas, like automation in manufacturing, which will dramatically change some job functions. In the technical space, however, I don’t think you’ll see entire job descriptions wiped out because of AI.
In reality, companies are going to get more effective at delivering their products and AI will make businesses more efficient. The net result is that businesses can use those resources, those savings they make in one place and apply to another place, and enhance their products and offerings faster and in new ways.
Once businesses save time and money in one place, they’re going to focus on another place that’s going to require human capital to achieve desired goals. In a very narrow sense, AI might replace some jobs as we currently know them, but in a long-term macro sense, it’s more like a reorganisation than an elimination.
JN: What are the main features of Iris and why is its development so important for SnapLogic’s customers? What are the next steps for Iris and AI within Snaplogic?
Dr. Benson: In its current form, the first main Iris feature we released is the ability to significantly improve user productivity in terms of building out data integration pipelines. Pipelines consist of “Snaps” which are connected together, with each snap representing a particular functionality. So you place these kind of Lego blocks together to create a data integration path. This is very visual, very tactile and previously it was very explicit. You needed to know exactly the next piece in the puzzle that you wanted to use.
Now that we’ve been collecting data for several years, we were able to use that data to fuel a set of machine learning algorithms that allow us to give you a pretty accurate recommendation or prediction of what the next piece in the pipeline should be. Iris can already get 80-90% accuracy recommending the next Snap or Snaps, and this will only get better with time.
When we look at the data, and it’s only been a month and a half since launch, we’re seeing around 40% of all Snaps that are put onto the canvas to build one of these pipelines are coming from Iris. The amazing thing about that number is that only about 20% of our users have Iris enabled, as we’ve only just started rolling it out. So as soon as we get more users switched onto the feature, and as we continue to build out Iris’s capabilities, it’s going to have an even larger impact. We’re seeing great product engagement and great productivity gains from this first set of Iris features.
Read more: A fork in the Bitcoin road – Good news for investors, bad news for blockchain
When I first looked at those statistics I was pretty blown away and pretty excited. I’m excited about our future. We have a whole set of AI features that we’re working on, and with each release of our product we’re going to extend Iris in very significant ways that will hopefully see similar if not greater productivity increases that we have seen with the first release of Iris.
So, we’ve got some exciting plans to keep us busy for the next couple of years.
