Generative Analytics

Unedited Google transcription from audio.

So, my audio recorder is on. So, once again, this is a session on generative analytics for the course, ethics analytics, and the duty of care module two, the module is applications of analytics. And the idea of generative analytics here is that an artificial intelligence system, doesn't just diagnose or even recommend things in recent years, especially we've seen a greater and greater capacity led by tools like GPT three for artificial intelligence to actually create new artifacts or new learning content.

And so, that's a significant change from what we've seen of artificial intelligence in the past. And so it's necessary to look at this new category of applications or uses of AI and analytics, generative analytics. As with the previous videos of the series, the method is going to be that.

I look at a series of types of generative analytics and discuss. Each of these briefly again, the purpose of this isn't to us a deep and definitive knowledge of any of these technologies in particular. Any one of these slides could be an entire course on its own. The purpose of going through an overview like, this is to give us as part of the overall, framing of the subject of ethics in analytics.

Someone understanding of what the applications are. And even in particular, some understanding of what the benefits of analytics are. So we can understand the motivation for wanting to use these systems and even to use these systems in contacts that some people might call unethical so to our list. So again, generating original content based on properties or parameters of the data combined with predictions or requirements.

For future data, is what gives artificial intelligence. The capacity to create new artifacts. Neither human's nor machines work from a proverbial blank slate. When they're creating something new, they create from something that already exists and then the anticipate or project or associate new properties, with the old thing, maybe combining reshaping.

There's variety of techniques that can be used in order to create new content. I think that understanding generative analytics will help us understand not only how computers can be of, but over time we'll also help us understand how humans can be creative. So, the first and probably the most famous examples of generative, content computer generated content in an online.

Environment is chatbots, you may all recall Julia, the chatbot from the 1990s but was pretty bad. There was once a chat bot that quote unquote ran for president of the United States. It was called Jackie. And today, chat bots have become sophisticated so customer service agents that are able to carry out.

I won't say an intelligent conversation with you. I've had some pretty unintelligent conversations with automated chatbots but conversations that are smart enough to detect. What it is that you're trying to understand. And at least make the effort to throw some resources, your own way. So, there's a several aspects to a chat but one aspect is the understanding of what you're saying.

And this is significantly difficult, It's not simply a case of audio to text transcription. Although it is that but it's also a case of being able to recognize what our significant or salient concepts in what you're saying and be able to associate those concepts with typical sorts of requests or questions that are person might ask in a sense, the AI that played against Ken Jennings.

Deep blue in jeopardy was really nothing more than a sophisticated chat button and pretty smart one too. Of course, not all content is going to be generated and real-time conversations. There's a wide range of applications already today. That automatically generate content that appears in leading newspapers. For example, the Washington Post has for a number of years used artificial intelligence to write sports stories.

These are fairly formulated stories that don't need. A lot of extra work and customization in order to produce, you know, a perfectly acceptable project. But this, as this capacity advances, the types of content that can be produced become more complex and and also more compelling. A reference here, an application called calf chai.

That is a machine writing an algorithm that can write articles from scratch. And the question is, it's as you see in the diagram, can a machine learn to write for the New Yorker. I would also include in this category, AI generated software. There are computer algorithms that support or actually write software for you.

I use a product called visual studio code by Microsoft. And over time, I've seen more and more useful software authoring assistance, appear as plugins for that environment. I've grouped a number of things under the heading of auto-generated animation, maybe I should change the title of this. But here, I'm thinking not just of cartoons that are created by AI or analytics but also images or video such as is produced by deep fake.

Any sort of animated content, what's significant here is that the AI is able to produce videos. People sounds, whatever that are sufficiently new as to be considered unique and there's a series of posts and examples out there. You know, they these people do not exist, this sign does not exist, this poem does not exist.

There's even an artificial intelligence that produces death metal on a 24-hour seventel week basis. And I might not be the greatest music in the world but this is certainly beyond the capacity currently of any human.

Intelligent and insightful conversation and content production may help artificial intelligence produce coaching applications. Now, what's interesting about coaching is that this is something that hasn't been available to, most of us for most of the time. Sure, athletes get coaching, it's expensive in time consuming, which is why the it's the best coaching is reserved for professional athletes.

Executives, get coaching. The highly paid executive has an expensive personal coach or mentor, to help them through those difficult business meetings or, as they say tough decisions, but for the rest of us were kind of on our own with AI and analytics. Backed coaching, we can access the same sort of resource that athletes and executives can access.

We're also always we're already beginning to see some of this and analytics tools that are diagnostic in nature that give us feedback on our performance. But now when these tools begin to offer comments suggestions recommendations training programs, encouragement, motivation and more. Now you have something that is very much producing the the same output and hopefully the same result as an actual coach.

They may be argued, you have to have a human to have a coach. But, you know, the choice isn't for most of us between having a human or having an AI, the choices between having an AI or having nothing and having. The AI coach is probably much better than having nothing learning analytics will not just provide coaching in the moment, but also coaching that helps us over the long term.

For example, on the slide here, I have the suggestion that they may help students develop self-regulated learning. Maybe I'll also me help with things like pattern recognition, critical thinking with negotiation conflict resolution other, things like that. What they call the soft skills that help people get by in an increasingly complex world.

Related to all of these is content curation now. It might be said and not in reasonably that perhaps content curation should be classified as a type of prescriptive analytics, focusing on the role of the curator as someone who recommends content but arguably. And I would argue here is about much more than selecting and presenting content.

There's also very often and act of interpretation and presentation. That happens in, this does involve the production of new content, everything from the creation of metadata to the writing of those short, synopsis that show up on the little cards beside artwork in a museum to the creation of a collection that may mix different things.

All these are risk for the mill of an AI content curation engine and so we can we can imagine seeing in the future. New ways of saying content. Presented that give us an almost unhuman or yeah unhuman not in him but on human perspective on different artifacts from different cultures with different backgrounds.

And like for one, I'm looking forward to seeing that and of course everyone's bug. Bear is the whole idea of artificial teachers or artificial tutors. And right now we're not predicting robots talking to us in front of a classroom teaching us. Instead of a human I wouldn't make a whole lot of sense for a variety of reasons but we can easily imagine even today artificial teaching assistants.

And in fact there was a product called Jill Watson, that was used as an artificial tutor in a university class that actually fooled the students into thinking that she was human. Should I call a robot teacher? She, it's hard to say, isn't it? These teaching assistants are already being deployed in fields, like law medicine and banking, and we will begin to see them in other less profitable courses and programs in the future.

And it's one of these things. Again, we're not going to go from nothing to artificial teachers. We're going to progress slowly as the capacity of these AI supported tutors in teachers is slowly increased changing and in some ways eliminating many of the traditional duties of human teachers. So those are some of the generative analytics and you know, you may well, think of many more ways and which an artificial intelligence can be creative if you do.

Then I recommend that you go to the all applications page and module two and make your suggestion for a new type of analytics under the category of generative analytics. So, that's it for this video in the next video. I'll be talking about somedayantic applications of analytics by day on tech.

What we mean, is they go beyond making recommendations beyond creating new content and into the realm of telling us what should be done? What is right and what is wrong. And I think that's where some of the most interesting applications of analytics are coming up. So that's it for now.

Thank you for joining me. I'm Stephen Downes.