In this inaugural episode of the "Earley On…” podcast, hear Seth Earley, CEO of Earley Information Science, discuss the topic of machine learning and artificial intelligence. Seth discusses what enterprise organizations need to know about artificial intelligence and how to leverage AI for real world business value.
Welcome. This is Seth Earley. I am CEO of Earley Information Science. And this is a little bit of an experiment. I have actually not done a Podcast before and I'm going to give it a shot. So Earley Information Science as many of you know is a consulting firm that's been around for about 20+ years and we work in the area of information architecture and taxonomy and metadata, around knowledge management, around search, around information management of all sorts. And all types of information management. So, what I'm going to talk about today is machine learning, machine intelligence. And a lot of people hear about this term, they hear about artificial intelligence, they hear all of this stuff. And the question really is, what's--first of all, what's meaningful? You know, why do we want take care about this? Obviously, you know, you know hear about Siri and you hear about Cortona, and there is Amelia and there is Alexa, and all of these different tools and all these different systems are forms of intelligence that are computer generated and it's all about this field of our artificial intelligence and of course we have a self-driving cars and so. But what does this really mean for organizations today? And we can think of this from the perspective of machine intelligence. There are several classes of functionality. They are really about detecting patterns and matching patterns. And you can think of this as pattern discovery. Do you think of it as pattern recognition? You can think of it as pattern matching. So, obviously if you're looking at identifying faces, that's a recognition and if you want to find a new face that's similar to this other face, then you're doing a pattern matching. Many times you're also trying to discover patterns. So anomaly detection is an example of a pattern discovery. And so machine learning is a very broad class of technology that incorporates either processing of data where you take outputs of one operation. And those outputs form the input for another. And to this way, this offer learns through approximation, you increase accuracy each time you have an approximation and you bring that into another cycle. And so machine intelligence is pattern recognition, pattern identification, pattern matching. It can also do pattern prediction. Machine learning is about iterative inputs and approximations. Artificial intelligence is really a number of applications that take machine learning and machine intelligence, they put them together. What's the goal of AI? Is to allow for more understandable interactions with computers. It's a--it's a way of letting us emulate cognitive approach as to solve problem solving. So we want to get computers to think more like people, cognitive computing is another term people talked about. And it's used to describe process of machine intelligence and computer interactions and types of problems that can be applied and addressed, and dealt with using artificial intelligence. So, we have these ideas of artificial intelligence, machine intelligence, machine learning. We also have something called an inference engine. And inference engine is a type of--well, first of all, inference is a type of information that can be derived or implied from content in database identification of patterns or application of logic, right? So you can start to see where machine learning can help us with those patterns. So data is derived from--inference isn't necessarily there for humans or machine depending upon the application. So if you're processing a large amount of data, you can make inferences about the nature of the data, the contents, the quality, the security, the structure and so on. If you're processing content or documents you can make inferences about the relationships or the structure of data that's contained in that information. Now, we call that the Corpus. So we have a Corpus, we have content, we have data, we have documents. We derived something, you want to imply something. And you can actually apply this a lots of different types of situations. You can apply this to data quality. So inference engines can help us discover data sources for example and discover relationships. We can identify security risk. We can measure quality. We can make data sources more consumable and more discoverable for people. So there's a lot of really interesting things that we can do when we are leveraging those sources and when we are applying machine intelligence and machine learning. Now, we actually came across of financial consulting firm that needed a solution to tag and organize their documents and data feeds and metadata, so that they could store more efficiently. They could discover, they can access it. And if needed to do a data extraction on documents because a lot of the documents had--they're like reports, financial reports. They had incoming documents and then, they had data feeds. Then, the idea is they needed to be able to extract meaning from all of this information. So when you think about it, when you have lots of unstructured content, deriving meaning and making decisions based on that meaning is really what it's all about. So you can take documents and you want to have a human read them or you can have machine read them. And if the machine reads them, the machine can then organize to make them tag them, they can structure them. And then, we can also pull the data if there's data in say financial reports from that structure content, and then we can put them into systems and make them more readily available to humans. So, when you start to think about this whole life cycle of a content and information that's coming into an organization, especially financial services and analysis firms, you know, we have to have a mechanism for doing some kind of aggregation for being able to again, extract something from that information. There's no simple way to do this. There's no single way to accomplish this. We have to be able to do with multiple processes. We have to have some human curation. We have to have some automated processes. So there's lots of different onboarding processes for this. And at the end of the day, we have a combination of manual, we have a combination of assisted and then, we have automated mechanism. And what these different tools will do, this different approaches will leverage different classes of a machine intelligence. And at the end of the day, what we're trying to do is we're trying to make sense of information. We're trying to do something differently. So when you look at say a marketing organization, a marketing organization is trying to make decisions based on information about their customers. They are trying to make decision based on the signals, the vapor trials that this different applications are throwing off and they're always throwing off information. They're always throwing off data. And yeah, you walk around with your cellphone and your cellphone is constantly streaming data. Although of course someone, you know, selling that and making money from your cellphone and your data, but when organizations are investing in all of these new sources of marketing technology. They basically have to do something one of the data. They have to make decisions and they have to be able to do something with the signals, right? A customer comes to your site, they're interacting in some way, what's that doing? That's giving you a signal. They are electronic body language. And based on that, you do something. You give them an offer. You change, you know, the navigation. You give them sort of a piece content or then, they'll do something else. They take that content where click on an offer or they register or they make a purchase, or, you know, or they put something in the basket. All of those things are signals. And what we want to do is we want to take those signals and you use them to make decisions. What are the decisions we're trying to make? We're trying to make decisions to improve the conversions. We're trying to make decisions to improve our relationship with those customers. So all of these things are being leveraged to different degrees by organizations and many cases they're not doing a very good job of this. So, one of the things that I want to do is talk about those types of issues, those types of challenges that organizations have with their data, with leveraging, unstructured content, structure data and looking at the various tools that are emerging in order to do that. So that's what I help you accomplish with a Podcast series. And this is the first one. Rambled on a little bit but I'm sure I'll get better. Anyway, if you do have any questions or you want to give us a topic, just send a note to firstname.lastname@example.org, that's E-A-R-L-E-Y, don't forget the E before the Y. And of course our website which is www.earley.com. Okay. Thank you and I will look forward speaking with you again.