Instant Genius - Meet the computer scientist teaching an AI to play Dungeons and Dragons

Episode Date: March 15, 2021

Many of us have had a one-to-one interaction with artificial intelligence. Whether that’s through an automated chat service for customer service, or trying our hand at beating an AI built to play chess. But these experiences aren’t flawless, they’re not as smooth as our interactions with other human beings. One researcher trying to improve the language abilities of AI is Lara Martin, a postdoc at the University of Pennsylvania. More specifically, Lara is trying teach AI to tell stories. Let us know what you think of the episode with a review or a comment wherever you listen to your podcasts. Subscribe to the Science Focus Podcast on these services: Acast, iTunes, Stitcher, RSS, Overcast Read the full transcription of this episode [this will open in a new window] Listen to more episodes of the Science Focus Podcast: Rana el Kaliouby: What if computers could read our emotions? Aleks Krotoski: What happens to your data when you die? Robert Elliott Smith: Are algorithms inherently biased? Pete Etchells: Are video games good for us? Jamie Susskind: How technology is changing politics Jim Al-Khalili: Why AI is not the enemy Hosted on Acast. See acast.com/privacy for more information. Learn more about your ad choices. Visit podcastchoices.com/adchoices

You said this place was steps from the water. We just haven't found the steps yet. How much did we save? Enough. Enough to get lost. Or you could book a stay with Hilton. Welcome to your ocean front room. Just steps from the water.

The Hilton sale is on now. Book on Hilton.com or the Hilton app and save up to 20% to get the stay you expected. When you want savings, not surprises. It matters where you stay. Hilton, for the stay. Study And play.

Come together on a Windows 11 PC. And for a limited time, college students get the best of both worlds. Get the Unreal College deal, everything you need to study and play with select Windows 11 PCs. Eligible students get a year of Microsoft 365 premium

and a year of Xbox GamePass Ultimate with a custom color Xbox wireless controller. Learn more at Windows.com slash student offer. While supplies last, ends June 30th, terms at AKA.m.S. College PC. Peak pollination season, and my business is scaling fast. To keep the nectar flowing, I need a phone plan with top priority data speed.

That's why I chose GoogleFi Wireless. My connections stay strong even when the hive is buzzing. Plus, unlimited plans started $35 a month. Now that's a deal that doesn't stay. Explore Google Fi Wireless plans today. Plus taxes and government fees. Google Fi Wireless is not subject to data traffic deprioritization during times of high network usage. This podcast is sponsored by.

by Name, Audio and Focal. Streaming has made music more accessible than ever, but true listening is about more than ease. It's about quality. British audio experts name audio, alongside French acoustic specialist Focal, combine handcrafted tradition with cutting-edge innovation and high-end materials,

delivering digital precision with analogue warmth, so you can experience exceptional sound at home. Music just as the artist intended. Visit nameadio.com to learn more. the Science Focus podcast from the BBC Science Focus magazine team, with the UK's best-selling science and technology monthly, available in print and in several digital formats throughout the world. Find out more at sciencefocus.com or look out for us in your app store.

Hello and welcome to the Science Focus podcast. I'm Amy Barrett, editorial assistant at BBC Science Focus magazine. Many of us have had a one-to-one interaction with artificial intelligence, whether that's through an automated chat service for customer service, or trying our hand at beating an AI built to play chess. But these experiences aren't flawless and they're not as smooth as our interactions with other human beings. One researcher trying to improve the language abilities of AI

is Lara Martin, a Computing Innovation Fellow and a postdoc at the University of Pennsylvania. More specifically, Lara is trying to teach AI to tell stories. I'm really pleased to be joined by her today and I believe you have an AI-generated story to share with us. Yes, I have a story. here. And I think what's good about this story is it kind of shows the strengths and the weaknesses of the system that I built for my PhD. So it goes like this. Christopher attacks Harriet. The king tries to take him away. Christopher is taken away, but is quickly saved by Agner. The king is able to escape the tower. Christopher and Christopher are able to get Christopher out of the tower, where the tower is still on the king. Penelope and the King are able to save Christopher.

So what's interesting about this story is that it's starting to make causal and coherent sense until it reaches a point where it just spits out garbage. So there's, so the way that I've been doing my work is I'm trying to combine. neural network output. So like GPT, there's these really large language models that people use for a lot of natural language generation work. And I combine that with more classic AI symbolic methods. And so here I am trying to make the story, I'm fighting with GPT to make it more coherent, but there's still work to be done. With the Christopher and Christopher sentence, things just, my parser just breaks. And so it's a work in progress, but that's what research is.

Yeah, absolutely. We should really be impressed by hearing that story, shouldn't we? you know, it's one thing to hear it with the frame of reference of all these amazing stories that come out of humans. But for AI, this is kind of really novel, isn't it? Right. So what's interesting about having an AI tell stories is that if you think about what you're feeding in an AI. Or so there's, I don't know how familiar everyone is with neural networks, but neural networks are a way of, of having the computer generalize over a large amount of data. So they learn to pick up these patterns over time.

And the way people do this, so like what I mentioned, GPT, GPT is where they took this massive amount of data. Like they're like it's just what we didn't think was possible a couple years back. and just crammed it into one neural network. And the neural network just learned patterns of this data. And so I take that and I tweak it a bit so that it learns to tell, like I have a corpus of science fiction stories. So I'm like kind of retrained it or they call fine-tuning it to,

to tell science fiction stories. But going back to what you said, what's interesting is that so now the AI has this great idea of what a story is, but that doesn't give it any idea of what it's like in the real world. So there's this thing called

the Principle of Minimal Departure by Mary Laurie Ryan. And I love this because it really sums up this idea that when you tell a fictional story, you try to match it as close as possible to the real world. So even when you're listening to a story, you make assumptions that it's going to be as close as possible to the real world. So I don't have to reintroduce gravity every time I tell a new story. That's going to be an assumption you make. I think in the book she talks about how, like, Don Quixote doesn't have to talk about how to use currency and waste time with that.

Because these are just cultural or physical assumptions that we make. And so, therefore, they're not in the stories. and that makes training these neural networks really difficult. Because the neural networks aren't able to understand what is and what isn't needed for a story? Well, it's that, so they end up learning kind of temporal relations. But that can mean anything. It could mean, like, I have to load the guns. before I shoot it or it could mean

like Susan jumped up and down and then Evan flew to Paris like it's like maybe that's that's a thing that happens maybe but um so it's it's learning these it's learning what sequences happen across time but that doesn't mean that it's like cause and effect um and so So what I do is I kind of inject that into these systems to guide the generation and have it make more sense.

So we know that, you know, AI are capable of playing specific games that we've told it to or following kind of instructions. But why is it important for AI to be able to tell stories and can. convincing stories. Yeah. So, well, I mean, I kind of like to think of a world where imagine you have this, instead of using like a keyboard and mouse, maybe your interface is talking to the computer, like kind of like Alexa. Never mind. He's listening to you right now. Apologies if any listeners have any kind of Amazon technology there. Oh, I turned mine off, so hopefully it turned off for everyone else. But imagine having a full-blown conversation with her or saying, like, you could tell a story, like, okay, so I need to plan a birthday party.

Well, in this birthday, in my birthday party, I'm going to have these wonderful balloons and then all the guests will arrive. And so you're telling this, it's like almost all conversations and all types of talking that we do is actually kind of telling a story. So it makes, I think it would really improve our interface, our interfacing with the computer if we were able to tell stories. with it or at it. So it needs to be able to both understand what we mean when we tell a story and also tell a story back. Yeah, so telling a story back is another step. So when I do story generation, it is doing story understanding. And then, well, it's kind of bulk of story understanding.

and then it tells, it generates its own story. And being able to generate its own story is just like, okay, did you understand me? Are you on the, are you on the same page as me? Will you plan this birthday party the way I asked you to? And, but I think, like in the more recent or in the closer future, I think that it would be really great to see systems that, like, that will, like, that train you in certain scenarios.

Or if you have writers block and you, actually a lot of people are working on this, where maybe you stop for a while and the system, and gives you prompts and how relevant are those prompts and does the user use them and how do they use them, things like that. And so there's different ways that stories can be used that people don't realize, partly because we are telling stories all the time. So it is kind of fundamental to what it is to be kind of human within our relationships. yeah, I often start out my talks with, like, we've, like, learned to tell stories before we could write.

I mean, this is part of our human nature, and it's not easy. It, like, for a computer to learn how to tell stories or learn how to understand stories, it needs to know all this other basic information as well. That's why a lot of natural language processing, which includes linguistic information, is necessary in order to build up systems like this. So while there are certain rules that go along with certain stories, are there kind of basic patterns that you can teach the AI? Yeah, so there's going back to, well, there's a couple of ways. to teach the AI certain patterns, some of which is like going back to what I said about inserting symbolic methods,

which meaning that, so when I say symbolic, I mean like, like, feeding in the, to the system these, like, discrete tags. So instead of saying, dog is this like string of numbers it is I'm just using the word dog and this is what I'm keeping feeding into the computer and the computer does not understand this but it can operate on it because I am I gave it these rules to operate on it so if you see a dog then get a treat so so there's been work by Nan Young Peng, who is now at UCLA. And she has looked at having these keywords. So a user enters in

a sequential keyword. So like eat, leave, drive or something. And the system will create a story, like using these keywords and having like each sentence contain the words in order.

So it becomes like Charlie, Charlie ate his breakfast and he left his house and drove to work. And so that work is actually done mostly just with neural networks. And there's stuff on the more symbolic side. where you can have a lot of control over the types of stories that you're generating because these systems are basically hand-engineered. And so instead of having just this data that who knows what the model is learning from it, you're actually just creating these nodes.

They're these like plot points of the story. And then you have another system that will plan paths across these plot points to create a story. And that that is, that type of symbolic story generation has been going on since at least, the 70s, these neural techniques are more modern and, yeah, not a lot of people have worked on the neural story generation until about 2018 or so. So it's a very kind of, it's a research in its infancy. Right. what's interesting is that like you said like there's these systems used to have a lot of control we used to have a lot of control over them and so we knew what types of stories they're making and now we're diving into this other end where who knows what who maybe we don't even know what data it's feeding in this is actually a big problem that is across all of artificial intelligence where we have this huge amount of

of data and there could be like racist, sexist, like all kinds of garbage in this data that we just don't know is there because it's so huge. And we're just training on it. And so we end up having these systems that it's like a like gambling trying to see what it outputs. So where does the data come from? It comes from a variety of sources. So I have personally, I have scraped, like for my science fiction data set, I scraped from fan wikis because science fiction TV show nerds are very thorough with their plot summaries.

And so that was a great source of data. But people have been like scraping all of the end. internet, basically, to create these data sets. And the internet can be a pretty toxic place. Yeah, gosh, it's scary to think that the fate of artificial intelligence could potentially lie in the hands of the people on the internet. Yeah, there was actually a pretty big disaster that Microsoft had a few years back, where they had a Twitter bot learn from, the interactions that she had on Twitter. And she quickly became a Nazi.

And it was not good. So, yeah, and they took that down very quickly. But it, like, it made a ripple. And things like this keep popping up. And even though I don't research bias in data in particular, I like to be informed about it and try to be careful about what data I'm using. So there is a responsibility on your end? Absolutely. And I think all AI researchers should have this, or do have this responsibility, whether or not they work towards making things

better. So for the system that you've built, what would you say is kind of the ultimate test for this kind of thing? Is there something that's like the Turing test that you can do? Okay, yeah. So, yeah, I think that Dungeons and Dragons would be a really great. It's not, I wouldn't call it a turning test. It's because, yeah, it's, but it's like, it, it would really show how open and mid and flexible and useful these types of systems can be. because if you just think about the types of things that you do when you play these games is pretty remarkable. Like you're not just telling a story, but you're telling a story with like three or four other people.

And if you're the dungeon master, it's a, it might even be like a harder task. I think it is kind of a harder task because you're creating a whole world and you have to relay this information onto other people to make, to allow them to create a theory of this fictional world that you're trying to share. And so it's, there's a lot of theory of mind going on. So trying to see, are you all on the same page? Do I understand what you think the world is like?

There's a lot of intrinsic reward. So unlike playing Go, for instance, which has an extrinsic reward, which says that like if you do this, I actually don't know how to play Go. If you do this and this, you get points for that. stuff and maybe I'll stick to the Atari games because I've done that. So like if you shoot all the like

aliens in an invader game then you get points for each alien that you kill and then you have this like extrinsic reward that you get at the end that's like very clear like you can't debate

that. Whereas if you have an intrinsic reward it's like well maybe my character will get I maybe my character will will get um a reward for or like feel good about um taking care of this orphan that we run across um and that's not something that you can necessarily like build into the system it's something that um it's something that's both kind of personal and also like personal to the character. So it's yeah, there's just like a ton of these challenges that we haven't even begun to look at. It's pretty remarkable.

Do you think it's something that, you know, you and I could see in our lifetime an actual AI dungeon master? Well, I mean, there is the AI dungeon, but to actually see something that acts like a human or, and there's like different levels of being a dungeon master, maybe someone's good dungeon master and someone's a poor dungeon master. So to see a good dungeon master AI in our, lifetime, I think. I don't know. I'm a little skeptical that'll happen. Yeah. And can can an AI

ever be creative with what you've given it or is it just going to kind of rehash the things that you've told it? Can it actually come up with new ideas? So yeah, computational creativity is a really fascinating field in general because there's just so many kind of philosophical questions that we don't know how to answer. Like if it creates something that's brand new. So like there's this idea of there's different types of creativity. There's personal creativity. So maybe I like if like a two year old drops something on the ground and they discovered that their their cup. reacts to gravity in this way. So that's like personally creative.

But historically creative is a lot more complicated because you have to compete basically with the rest of the population. And so if an AI comes up with something that's like historically creative or at least like new and interesting, and then it brings up the questions like, well, was that the AI's work, or was that the work of the developer that created it, the researcher that created it? Because there's there's like every agent that you make, you're putting your imprint on it, whether you like it or not.

And the more, the more on the like rule-based symbolic side you head towards, the more of your imprint is into this agent. So to say that, like, can an agent come up with something that's creative by itself is a tricky question. I mean, if it's like purely by itself, like, quote unquote, it might just be like it did something completely random. And in that case, it wasn't intentional, you know?

and it's is that creative then there's just a lot of philosophical questions here that are yeah just open-ended and I'm not a philosopher so and of course there's the ethics side of it as well like there's kind of a lot this is really a new route for exploration yeah absolutely and what's the kind of most surprising thing that is thrown up for you there's been some pretty interesting things that my systems have come up with over the years um like the one of my early systems uh had a story about um like a horse becoming a lawn chair entrepreneur and I just think that that concept is really interesting when I think about that every now and then.

And so there's these really like quirky, funny things that they come up with. But like in that system, for instance, that was random. Like I had, well, it was partially random. Like I had these tags that said like, okay, I need an agent, some kind of job maybe and something. And so it filled in, but then I was, it was pretty early on. So I just had it fill things in randomly.

And it just happened to come up with this interesting concept. But most of the other times it would come up with these weird things that like I don't even know what like a really specific type of flower or something. like because it would just pull this from the database it had and so it might be it it's it's kind of um it's not as it's not as surprising when you can't understand it yeah but i can definitely see kind of a ai terry pratchett style thing happening in the future with a horse who sales, volunteers. It's amazing. Do you think that there's kind of, I mean,

we're talking way into the future. So this is probably kind of speculative fiction right now. But will, you know, art and storytelling be the next big frontier that sees an AI boom?

Are we going to see AI books, you know, are publishers on the shelves or are we going to see an Oscar-winning AI in the future? Well, I think it would be great to see AI being like creative AI being made. I am not as like having like an Oscar winning AI I think is not not as great because I

I think that like it goes into that problem. It's like who who has the ownership of this? Like is it the AI's like did the AI win the award or was it the people that created it or was it the data that it was trained on? Like the people who created that data like it's there's a lot of weirdness there and the computer is not a living thing. And I think that it's really important for people to realize that computers are not as smart as they think they are. Because it's, they're not, they're not people.

They're not, they don't have agency. They're just tools that other people have used to work on these things. And so I think the best use of computational creativity or creative AI is co-creativity and using it as a tool to augment human creativity. Because computers are really good at looking through large spaces of data so they can come up with things that you've never seen before or never thought of. connected with this. But humans are really good at making those connections, connecting ideas that the computer might present to them. So going back to the example of the horse being the launcher entrepreneur,

the computer knows nothing about what that means. It's just spinning out stuff. but having a human take that and like run with it maybe they go and make a story about this horse like that'd be fantastic like this is a this is a really interesting idea and humans have that ability to connect these things and I think that I think that's a really good symbiotic relationship that's needs to be used more. Yeah, I guess I've never really thought of it that way, but they're kind of tools and aids to our own creativity as opposed to having their own agency. Yeah.

And for you, what's kind of the most exciting thing about your research? Is it where it could go or some particular use? So it's interesting because I am very proud of the work that I've done during my PhD, but it is not the work that I wanted to do when I started my PhD. So I started, I came into the program wanting to add speech and like spoken technologies to something cool. And I came across Mark Radell's work on storytelling. and I thought that would be really, really great. And then I realized, oh, this is not ready for speech.

And well, maybe it was more like Mark told me that this, we're not quite ready for speech yet. And but now that I'm done, it kind of opens the door to all these different possibilities of what can I do with, with just computational creativity and both like language understanding and speech technologies because in speech there's this thing called prosody which is basically all of the extra information that you convey to someone in addition to the words so it's like the intonation and the like the melody the timing and a lot of times this has meaning behind it because people talk in a particular way or you can sound sarcastic and you don't pick that up in text. And so I think that's really fascinating.

And I think that there can be a lot of work there too. I mean, there is a lot of work going on in speech technologies. but yeah, I think that all these things are really interesting and I'm very, very curious to see where my career takes me. And that sort of thing, you can see kind of the benefit of, or doing this research from an kind of AI perspective, will actually help us as humans, like for people who, maybe struggle with understanding intonations or, you know.

Right, right. Yeah, I tend to, I consider myself a cognitive scientist, and so that includes, like, artificial intelligence and linguistics. Because I did linguistics in my bachelor's degree, in addition to computer science, and I kind of combine the two. And then there's other things in cognitive science, like, philosophy and psychology and it's a very humanistic way of thinking of science and I think it's I think that by looking at AI through a cognitive science lens really puts more of the

focus on the human and so yeah you can make more tools that either act like a human or or you use what you learn about humans to make better AI to work with humans. That was Laura Martin from the University of Pennsylvania talking about her research, teaching AI to tell stories. If you've enjoyed listening to this episode of the Science Focus podcast,

please do leave us a review and check out the February issue of BBC Science Focus magazine, which is out now. In this issue, we explore how you, your brain creates reality. We're looking to the baffling science of dark boson stars, and as always, our panel of experts answer your questions. Of course, there's much more inside and on sciencefocus.com. Thank you for listening to the Science Focus podcast from the BBC Science Focus magazine team. With the UK's best-selling science and technology monthly,

available in print and in several digital formats throughout the world. Find out more at Sciencefocus.com or look out for us in your app store. This podcast is sponsored by Name, Audio and Focal. The texture and emotional depth of music can be lost through digital sources or poor signal. Name Audio believes you can have digital precision with analogue warmth. Alongside French acoustic specialist Focal, Name creates high-end audio systems, combining innovation with craftsmanship, so you can listen to music, just as the artist intended. Discover more at name audio.com.

Ambition comes in all shapes and sizes. At First Citizens Bank, we roll with your goals because we're built for what you're building. Fit for your ambition for Citizens Bank. Hank joined BJ's wholesale club the day he became a father of 30. I coach football. Now Coach Hank saves up to 25% off grocery store prices. 30 pounds of pasta, three cases of protein bars, 75 sports drinks.

And that's just pregame. He knows teamwork and BJ's knows savings. This is your home, coach. Home of the Save. Join for just $20 at BJs.com slash SW Fort Worth and save 10 cents a gallon on gas for six months. Now open, limited time offer new members only.

BJ's Home of the Save.