Campus Technology Insider Podcast January 2024
Rhea Kelly 05:20
I like that analogy. So, our panel thought we should define some terms to start with, and maybe I'll throw this to you, Howard. You know, can you kind of give a basic definition of, you know, AI, generative AI, LLMs? Like, what, what is that all? Where should we set a baseline for this?
Howard Holton 05:41
Yes. So artificial intelligence is using computers and math to analyze data and make decisions. And that is truly what it is. It's incredibly complex, but, but that's basically what we're trying to do, right? Generative AI is using a specific type of artificial intelligence that understands language, understands language, can determine intent, and uses that intent to generate a response. Large language models are, those responses are language. So, so you know, in most of them, as we think of them, they're really good at English, as an example. And we're now adding several, there's about six languages that they're very good at. But generative AI can do things like 3d modeling, right? Design a component that does these things, and you will get a reasonable CAD file that you can start with for 3d printing and rapid prototyping. The things that we can do with generative AI, we've barely scratched the surface of, right? Today we see things like ChatGPT that are that are large language models, we see DALL-E, right, that generates images, we see Copilot that generates software and code, but the kind of the sky is the limit as we get better and better and better at understanding intent. And then using that intent to feed into other systems that can create and generate a response. And, and really, they still require humans. They are not human replacement systems, they don't work that way. And they're not really designed that way. They're designed to make us faster and more efficient, not replace us. Assuming we're, you know, reasoning, logical, thinking humans. Some people are sheep. So I don't know what to do about that.
Rhea Kelly 07:29
One of the terms I like is the, being the human partner to AI. The AI needs the human partner to really be, I guess, the most effective.
Howard Holton 07:42
Yeah, it's a great personal assistant. Right? So if you need an, if there are things that you do that would be better done by someone, by an assistant, AI is like, it's great to think of that. But like I said earlier, right, think of it as a conversation. It's not a search engine. Right? If you just tell your assistant, "Go book me a flight to the UK," they will book you a great flight to the UK. It won't be on the airline that you want, it may not take off from the airport that you want, it may not be priced accordingly, it may not have the time consideration, and you're likely to be at a bulkhead, by the bathroom. Right? But if you have a conversation, you're going to get the result that you want. And AI is similar that way.
David Weil 08:21
Building off of what you said, just to sort of restate it slightly differently, if we're defining AI, I think it's really important that AI does not equal ChatGPT and Bing. AI is much greater than that. And I think that when we have these conversations back at our institutions with leadership or others, a lot of people are thinking AI equals ChatGPT. And it's not. And in fact, I, I would venture to say the way that most people will be interacting with AI is not through one of those, but actually embedded AI within applications that we're using to do sub tasks. And I think that's one way that we need to make sure we're framing this conversation.
Howard Holton 09:03
Yeah, I mean, at this point, everyone has interacted with AI. Everyone. Everyone in the room, everyone in the world has interacted with AI in some way. Right? The fact is, most of it has been passive to most of us. If you don't work in the field, if you don't work in the industry, if you're not actively developing these things and managing these tools, AI has impacted your life in a million ways, many of which are bad. Like Target was caught using artificial intelligence years ago to identify patterns of people that were pregnant that didn't know they were pregnant and advertise pregnancy-related things to them. There are, there were really awful, unethical things that have been done with AI all the way up to now, right? Banks approving loans based on AI selection that had horrible evil awful biases built in. Right? We can use AI to predict whether you'll default on loans. Anyone want to guess what the number one indicator that you're going to default on a loan? You've paid a bail bondsman recently.