00:06
Adam Stofsky
Hey, this is Adam Stofsky from briefly, and I'm here with Annette Hearst from Oric to talk about what everyone's talking about, which is generative AI. So, Annette, can you kind of just tell our audience a bit about yourself and what you do day to day over there at Oric? 

00:20
Annette Hurst
I'm an intellectual property lawyer, and I help clients solve problems related to deploying technologies and also using technologies where intellectual property rights, and especially copyright might be a blocker or a risk. 

00:37
Adam Stofsky
Generative AI. It just seems like this futuristic technology, it's at the cutting edge. Can you just summarize, why does your average company need to worry about this at all as a legal risk? What's going on here? 

00:50
Annette Hurst
Sure. First of all, for companies who are creating AI tools for the rest of us to use, there's a whole set of legal challenges around doing that. And then there's a set of legal challenges around people who just want to use the technology in their everyday business. They don't. They're not savants technically, you know, but they see the promise of the technology and they want to use it. But there are several inherent issues that can create legal risks associated with use of the technology for folks who don't. 

01:23
Adam Stofsky
Really understand, including maybe myself. So what is generative AI? How does it differ from other kinds of AI, like predictive AI? 

01:31
Annette Hurst
Well, the first thing to understand is it's not really AI, like Skynet or Hal. Right. It's very sophisticated software model that's generally referred to as a machine learning model. And there are different kinds of training that can be done to cause these models to learn data, to be able to generate outputs. And this is where the generative part comes in. They generate outputs that are same kind of content that humans can create. So text, images, music, video, all the sorts of things that humans create, these generative AI's can be trained to create as well. 

02:14
Adam Stofsky
What is predictive AI in comparison? 

02:17
Annette Hurst
Predictive AI is something that people are probably encountering in their everyday lives without really thinking about it being a predictive AI. And an example is recommendations that you get if you watch Netflix and you watch some shows, and then it recommends another one to you, that's a predictive AI. It's trying to predict what else you might like. And there are those kinds of matchmaking and decision making. Algorithms exist throughout our lives and society at this point, and they're often invisible, but they are definitely out there. 

02:53
Adam Stofsky
So predictive AI predicts what movie I might want to watch, whereas generative AI could maybe make the movie I want to watch. 

03:00
Annette Hurst
Exactly right. 

03:02
Adam Stofsky
Because these pieces of software create new things out of other things. This essentially makes what they create something kind of like what we call intellectual property. Is that sort of at the heart of what some of these problems are? 

03:17
Annette Hurst
Yeah, that is at the heart of it. But in a broader sense, it's also that all the mistakes that humans can make, these generative AI's can make also. And humans can make a wide range of mistakes when they're generating content that can lead to legal problems. And we've developed a lot of processes and systems over time to help mitigate those risks. With AI's, you have all the same set of risks, but it's a whole new landscape in terms of how do I mitigate them when there's no human in the loop? 

03:52
Adam Stofsky
So let's get back to our question about legal risk. Putting aside people who are making AI tools, if you're just an average company that maybe is going to use these tools, can you kind of summarize what are the major risks that we should all be thinking about? 

04:06
Annette Hurst
So in the intellectual property context, you have two different fundamental types of risks for users. One is freedom to operate. And freedom to operate refers to when you get an output from the tool, are you free to use it in a commercial way? Right. And there's an important line to be drawn here between internal company uses that are not, you know, not core to business operations or not being sold out in the marketplace, and things that you want to take and sell out in the marketplace. And so the first question is, do I have the freedom to exploit this, or might it be infringing on somebody else's intellectual property? Right. 

04:50
Adam Stofsky
Can I use it without a risk? 

04:52
Annette Hurst
Right. Can I use it without a risk? Good. Summary. The second problem is, can you own the output once you've created it? The hallmark of ownership is the ability to exclude others, to have exclusive use of something. And when you think about everyday important intellectual property, like movies and songs and other kinds of things, the value of those rests in the ability to control the exclusive use of it. And when something is created by an AI, it's not clear whether that ownership right, whether you can have that or not. 

05:32
Adam Stofsky
So to summarize, you're using a piece of software that's making something new. So the first question is, do I have the right to use what I've made because it might be composed of stuff other people own, music, images, names, whereas the other risk is, if I make it, what is like the legal status of this new thing that I kind of made, but a piece of software also kind of made. Is that a good summary? Exactly.