Campus Technology Insider Podcast January 2024
Noble Ackerson 15:51
Yeah, just to double tap on that. On the process side, you talked about the technology, the people. On the process side, what I'm sort of seeing a lot of are these two, two domains. Things that an organization can do to sort of automate certain decisions. And so we're talking about, you know, certain repetitive tasks that may be low risk, that don't particularly need a human in the loop, but may need a human on the loop to sort of help self-correct and further improve the models behind the scenes. And then more augmentative, this is what I actually see a lot more of a solution. So a lot of Microsoft's offerings with Copilot is more augmentative, in that, I mean, you have solutions that assist. And I, actually, another way to answer it, sort of bind this to the first question is: tools for thought, right? That's literally generative AI, I just thought about that, we should put that on a t-shirt or something. It's just a tool for thought, right? Like so for, from a process standpoint, what are the existing tools to augment within our organization? Rather than sort of thinking of building a whole new thing with generative AI, we've got a bunch of feedback from students, and how do we use some of these tools to either fast-track, like you said, our path to decision, or at least help basically decision-optimize our way to do it? And what I just want to sort of say is, all that said, generative AI isn't always the answer to that. Sometimes a traditional model, linear regression, like, you know, just simple math or statistics could solve the problem. But it all starts, starts with, you know, you know, leadership sort of looking through, what are the core problems that we need to look at, that we perhaps have to wait until some of our vendors incorporate? Or what are some things that we know would cause a little bit more complexity if we were to customize a vendor's offering, and just augment by building ourselves? And that's a hard decision for any leader to do, because there's going to be tech debt or sunk costs, and hard mistakes made along the way.
Rhea Kelly 18:24
Howard, what, what do you think is involved in that decision, of whether or not generative AI is the right solution or something else?
Howard Holton 18:34
I'm glad to kind of move on because they did a really good job of covering everything, but it brought up a few thoughts. The first is always use this the simplest thing to solve a problem but no simpler, kind of thing. If you can make a decision with four input points, you don't need AI. If you have 400, you need AI. Right? Human beings can't take in 400 inpoints, inputs to make a decision, but AI can. Right? It's very, it's even very hard to do that with just a standard statistical modeling. But if it's four, let's not overcomplicate it. AI is still a black box, right? I can audit the decisions that are made from four. Also, generative AI is not actually an application. ChatGPT is an application because they have taken generative AI and given it all the things that make it an application. Right? So generative AI, generative AI is effectively you interacting directly with AI for the first time. It's why it's part of the, the culture and our conversation right now. Right? But, but what it is is an intent engine, more than anything else, in a way that nothing's been successful at being an intent engine up to this point. There's all kinds of things that do a really good job of set of stage two. But since humans communicate in ways that only make sense to humans really, up to this point, and generative AI is the time where we can say we've actually made a computer understand, understand, the intent in what a human is saying, the ability to use that as, as a tool in a long line of tools, to get to an output that's simply better than we could otherwise, is the way to think about it. It is just a tool, it is just a component by itself. Other than kind of having something to bounce ideas off of, and, and like I was saying earlier, right, get, get, get kind of outlines, you really need something around it, right? You need information, you need systems, you need other pieces of automation, whether those are AI or not AI, to really take advantage of what generative AI can do. And generative AI is just going to try to understand the intent behind Standard English, right? So you don't have to think like a programmer or developer to get anywhere.