Tina Eliassi-Rad
Appearances
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And that's basically what is going on. You know, that's the reality of it, you know. And and so so there's a lot of that kind of engineering going on as opposed to like really thinking about what is the phenomena that I'm interested in? How is the data coming to me? What are the sources of noise? Should I how should I take them into account? Should I even take them into account?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And what are the uncertainties in terms of the predictions that I am outputting?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah, so basically you create a bunch of data and you get a buy-in from the community that these are good data sets to test a machine learning or an AI model on. And then there's a leaderboard and you want to be number one. Right. And so you hack the systems that exist or you hack your own system. You create your own to be number one, you know, as as much as possible.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And that's basically what is going on. And I like that.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
this metaphor so my colleague um barabaschi said it's like there are two camps there's like a toolbox it's a finite toolbox right and the machine learning the ai people the engineers put tools into that toolbox and because it's finite it's very competitive that is my tool beats your tool even if it's like one percent by one percent that it's not clear if it's statistically significant or not and i may be king for only 30 seconds
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
because another tool comes in, right? And then there's like the scientists on the other end that just open the toolbox and say, okay, well, what is good for whatever, you know, whatever prediction task I want to do. And then they pick a tool out of that.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And so a lot of this like benchmark hacking or state of the art hacking happens on the engineering, on the AI machine learning side, the computer science side, because you want your tool in that finite toolbox.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
It is a very big problem. I mean, there are multiple angles to this. So one is, for example, because of all the hype, oftentimes people on the engineering side don't talk about the assumptions that they have made or the technical limitations of their system. Because of that, we have this reproducibility problem.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
So not even a replicability problem, but a reproducibility problem, which is just a code. Can I just reproduce your code as you have it? Right. And even with your training data, even with like how you broke it up with these different like folds or whatever, you know, and so which is like very, very, very low bar to pass.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
But that doesn't happen because there are lots of assumptions that are being made, etc. Then there's this notion of we are living through this era of big models. I want a model that has many, many, many parameters, even if I don't need all those many parameters. Or for example, maybe I do care about interpretability. That is, I want to know what the model is actually doing.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
But because, again, for that one or two percentage point on the prediction side, you let go of it and you go with the bigger models. But yes, it's a big, big problem. For me, the lowest bar would be that we require, at least with federal funding, and in some of the service that I do for the federal government, I've been pushing this.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
I'm not going to be a very popular person, but that if you get taxpayer dollars from in your reports to the government, you have to have a section on assumptions and technical limitations. Because the problem is the way the peer review culture goes is that if I have a technical limitation section in my paper, the reviewer will just copy and paste it and say reject, right?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
But the federal government isn't going to do that, right? NSF isn't going to do that. NSF has already given you the money and you're doing the annual report. And so it has to be, come on, just be honest, right? Like I did not test this method on biological networks and they're very different than social networks. So like caution,
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah, I love that problem. I've thought about that problem a lot. So the issue there is similarity is an eye of the beholder, right? And it depends on the task itself. So similarity is an ill-defined problem. And so you can say, okay, well, I can go with something like an edit distance. Like, okay, how many new nodes do I have to add to graph number two?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And how many new edges do I have to add or remove to make it look like the other graph? And then try to solve the computationally hard problem of isomorphism. In fact, alignment, right? And in many cases, you don't need alignment, right? So, for example, you can think about two networks and you have started a process of information diffusion on it, like you started a rumor, let's say, right?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And you would just measure, like, how similar does this rumor, the same rumor, travel through network one versus network two? And if like, you know, it travels similarly, let's say, you know, I'm going to throw some jargon, like the stationary distribution of a random walker that is spreading this rumor becomes the same at the end. You would say the networks are similar enough. Right.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And so you don't need to have like the sizes exactly be the same. So it could be, for example, you have a social network of France and a social network of Luxembourg and you start a rumor in France and in Luxembourg. And they are processing the same way. And you would say the networks are similar, even though one is much, much bigger than the other.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah, yeah, now the problem with grouping nodes, this is a very important problem and it's been studied by lots of people. Within graphs, it's called community detection. Basically you want to group similar nodes together. Now you can have different functions that you define about what similarity there means. It could mean that these people just talk to each other more, right?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
So there's more connections between them than what you would expect in a random world, right? or just more connections between them than other folks. Now, this kind of community detection, Aaron Closet, who's a professor at Colorado, showed that there's no free lunch theorem there. And actually, it was Aaron Closet and others. And I think actually Aaron was the last author.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
So I think the first author is Leto Peel. But you know how it is. You usually just name your friend.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
My apologies to the other authors. But they showed it in no free lunch theorem, which basically means that it is not the case that there is like one particular group of or one particular collection of nodes that you're grouping that would give you the best or the best. true communities. You see what I mean?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
So because when you are doing these grouping of nodes, you have some objective function that you're trying to maximize. And basically the idea is that there is no one peak there. So there's not like one particular community that you can put Tina on and say, okay, Tina belongs here. That's where she has to sit. And so some of that becomes an issue.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
But this notion of what does it mean for one network to be similar to another network has its tentacles to community detection, to clustering of nodes, and all of those are ill-defined. So it really is driven by the task at hand.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah, and that becomes what we call the small world problem, right? Or the Kevin Bacon or the Erdős number, right? You don't have to go that far out. to be connected to famous people.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
I mean, for downstream tasks that you can like have some, let's say, confusion matrix where you can draw like true positives, false positives, true negatives, false negatives. We're actually very good at it. But if it's about like, OK, I found these communities and do these communities make sense?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
It kind of breaks down into whether they're like hard clustering where you put Tina into just one community or you put Tina into multiple communities. And then there's a little bit of just like eyeballing it in a way. If you do not have this downstream task that you can say, okay, here are the true positives, here are the false positives, and so on and so forth.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
But in many cases, it's difficult to place a person in a social network only in one community because people are multifaceted.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
I think they're in part, they just want your attention. And so the objective function is such that, you know, they just want to hold your attention. And so they will show you whatever necessary that will keep your attention.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And so if they believe that like my tie to Brandon is very strong, that we have a strong relationship and Brandon found these things interesting, then they will show it to me as well to just test it, to see whether, you know, they can capture my attention. And then through that, they can show me more ads.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Exactly. Exactly. And so they kind of go hand in hand. And in fact, this touches on this issue that we have written a couple of times about. There was a Nature Perspective piece a while back and more recently an AI Journal piece on this. in a way like human AI co-evolution.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
So if you think about it, when you're using Amazon, when you're using YouTube, when you're using Google, you're providing data for them. We talked about this, right? And they take that data into account and they make recommendations. Those recommendations then affect what you do in the real life. And then you go back and you provide them more training data.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And so there's this kind of feedback loop that goes on and on. And it's oftentimes not captured in terms of who's influencing who most. And one example that I like here is like think about dating apps.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
There was a story recently from Stanford that like most people are meeting on online dating apps these days instead of like through college or through their friends, family, et cetera, or at the local bar. Now, those dating apps have recommendation systems, right? And based on those recommendation systems, perhaps you meet somebody, you partner up, and you have babies.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And so over time, these recommendation systems actually have an impact on our gene pool going forward.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah, but it's like, I suppose, and because these recommendation systems are all about exploitation and not exploration, but maybe you would say like my aunt or my grandmother or my college were also all based on exploitation and not exploration, right? But there is this notion that there are these algorithms that we can't understand what they're doing.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And perhaps 100 years from now, they may influence how our genome is evolving.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah, I don't know, like, how far do we think certain things are going? And society has to decide. Like, for example, New York Times had this article a while back about how there's a person who's trying to set up a company, an online dating company, where, like, on the first or second dates, which are usually...
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
you know, not very good, my avatar and your avatar will go on the date and then they will report back. And only if, you know, both avatars are happy, then on the third date, we actually go out on the date. And so like how much of actually our human behavior are these things going to take over?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
I think like I'm an introvert, so I'm like, oh, and also I'm a computer scientist. I'm like, oh, this is great. Let somebody else do the dirty work. And then maybe, you know, if it's a good day, I'll get out of my cave and I'll like go and talk to. But, you know, I for extroverts, they don't like it at all. So my husband was an extrovert. Like, what is what are you talking about?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Am I just a brain in a vat now? Like what's happening now? So I think it depends on where you are in this extrovert, introvert scale.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Indeed, indeed. For 30 plus years, it's been fantastic.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah, and I think in particular with, for example, generative AI as it's generating content, whether it's text or video or images, there's this notion in the late Dan Dennett, who you had on your podcast, very famous cognitive scientist, called these generative AI models counterfeit people. He had an Atlantic article a few years back about it.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And also because people treat these generative AI systems, these counterfeit people, as if they're more objective somehow. They know more than me. People tend to give their agency to them. And also these AI systems evolve faster than us. And so it's not quite clear, not that it's a race, But it's that they're evolving a lot quicker.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Their objective functions are different, like attention, money, et cetera, than perhaps the objective function of people, like maybe the good of the society or public good or something else than just like money or some like GDP or some measure like that.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
I think these days it's too much in flux. But I think, for example, there are certain things that can be done to improve it. Whenever you or another human being asks me a question, perhaps I would come back with another question. I'm like, did you mean this, Sean? Or did you mean that, right?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
But for example, with child GPT or these large language models, they never come back and say, like, did you mean this? The reason is that it reduces their utility, right? Me as a human being, when I ask the question, I want an answer and I want it now. Yeah. Right. Or like it never comes back and says, I don't know or I'm not sure of it.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And maybe you would accept that from a human being, but you don't accept it from a large language model. You're like, oh, you're a tool. You need to tell me. Like I asked you about this and I want the answer now. And, you know, and so there's some of that going on. But like the big tech companies could add those features. to make it more equal in terms of this conversation that is going on.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
But at this point, utility is winning over all these other things.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yes, there are actually other generative AI systems, especially for programming, that I've heard where, like, it tells you, like, okay, if you want to code X, this is how you code it. And then you code it and you're like, oh, it didn't work. You're stupid to the generative AI. Like, the human says you're stupid. And then the generative AI says to the human, you're not a good programmer. Yeah.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
You know, so then there's some kind of a, you know, then they get at it. Gets in a loop. But that's only like for, you know, specific ones. You're absolutely right. With chat GPT, it's not going to be that kind of antagonistic.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
I mean, you can try to find biases. I mean, there's a lot of work in that, like these large language models are sexist, misogynist. We wrote a report for UNESCO for last year's International Women's Day about how sexist and misogynist these large language models are. Um, the problem was that is whenever like I get, uh, somebody asks me that question that, oh, well look, humans are biased too.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
The problem is that I can hold a human accountable. I can sue a human being. Who am I going to sue? You know what I mean? And especially in America, we're very litigious. And so then this gets into accountability. And in fact, there's a lot of work in the government.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
For example, our government is putting a lot of our tax dollars into like trustworthy machine learning, trustworthy AI, et cetera, et cetera. And to me, it rings a little hollow because there's no accountability. Like, how can I trust you if there's no accountability? I feel like they go hand to hand. And so there's some of that going on, which is like, You know, who am I going to sue?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Am I going to sue OpenAI because it's sexist and misogynist? Like one of its products is sexist and misogynist. You know, that's not the case right now.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Perhaps, right? The thing is, at this point, what it gives out is what's the most probable and what it believes you will like, right? So it's a two-place function, what's probable and what you will like. But yes, you could definitely do that.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And there's this comedian, unfortunately, I forget his name now, but he was saying the secret to a long marriage is to never say what comes to your mind first or second. Always say the The third thing that comes to your mind, right? And this goes back to what you were just saying. Maybe you should just say this third thing, the third most probable thing.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And in fact, along those lines, usually the students who use these generative AI tools for like math problems, math homeworks, the first answer is usually wrong because a lot of the answers that have been uploaded into like Course Hero, et cetera, et cetera, they're wrong. Usually it's the second answer that's the correct answer.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
These are just anecdotal, right? Like I haven't had anybody do like a systemic study of this, but that like usually the first answer is not quite there, right?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah. So in the true computer science, AI, machine learning sense, we're very good at coming up with names for our system. So we called it Life2Vec. So we're just putting your life into a vector space, whether you like it or not.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
But you're just a vector in this vector space. Now, basically, the idea is that if you look at these large language models, right, so they're analyzing sequences. And so as human beings, we also have a life story. That's a sequence. Right. And so I was lucky enough to work with a group of scientists in Denmark.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
So if America has a surveillance capitalism, in Denmark they have surveillance socialism. So there is a department there, Department of Statistics, they call it, like Ministry of Statistics that collects information about people. And so we had information for about 6 million people who have lived in Denmark from 2008 to 2020.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And we were like, well, can we write stories for these people in a way and then feed it to what is the heart of these large language models, a transformer model, which is basically just the architecture of a neural network that learns association weights for within some context window.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
um and that's what we did so but instead of so for example chat gpt goes online and gobbles up all this bad data that that or that people have put in all the misogynistic sexist data we didn't do that so we had very good data from this department of statistics and we created our own artificial symbolic language
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And then we fit that artificial symbolic language for these six million people into a transformer model. And then we were able to predict life events. And so one of them that caught the media's eye was, will somebody between the age of 35 and 65 pass away in the next four years? And we picked that age range because that's a harder age range to predict for.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Like if you're over 65, then it's easier to predict whether you're going to pass away in the next four years. And if you're younger than 35, it's also easy. The other, right, you're unlikely to pass away. And so that's one of the things. The other prediction task was like, will you leave Denmark? You know, so then you can predict for that.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
But it had this similar technology as these large language models, which is like you have this one, what they call like predefined, where you just learn based on the data that you have what's likely to happen next. And then you fine tune it for whatever prediction task that you have.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
It's a logical encoding because the data that the Department of Statistics has in Denmark is all tables. So it is not like this kind of sequence. So then you could say, like, Tina was born in Copenhagen in December, blah, blah, blah, right? And we could generate a natural language, but that's difficult. Why would we do that?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
So then we generated a vocabulary for this artificial symbolic language, and then we And that was actually a lot of the intellectual property of the work is like, okay, well, how do you take these tables and then create this artificial symbolic language that then you can give to a transformer model?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Well, the thing that we found, which was very interesting, I think, so like the accuracy in terms of the model was about like 78%, et cetera. And I think that's why people were showing a lot of interest in it. But to me, that wasn't really the takeaway.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
The takeaway actually was that labor data is a very good indication of whether somebody in that age range is going to pass away in the next four years or not, because health data is very noisy and inconsistent. So even in Denmark, where they have universal health care, it's not like everybody goes to the doctor all the time and you have good data for them.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And then the other stuff was basically just which sector you were working in. Right. So if you're like an electrician. It's a bad thing. It's not a very good thing. Right. As opposed to like an office worker. So the labor data was actually very, very helpful than the health data.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah, maybe. Yeah, we didn't do any kind of causal stuff, right? Like a lot of the work, a lot of the hype that's happening now in AI and machine learning, they're all on the correlation side, not on the causation side. So we didn't look at that at all about what causes what. That's very difficult. And I haven't touched the field of causation in part because I'm married to a philosopher.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Because every time I try to approach the topic, I just heard nightmares. And so I haven't gone that way yet.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah, I think that there's some of that. I think the best way of using this is perhaps government policy. Right. When government issues a policy and then like maybe 20 years from that, you have if you have good data, you could see, OK, what has been some of the correlations that have come about based on this policy?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And then maybe, you know, the actual social scientists and political scientists can then draw some causal diagrams from what we find. Because the one thing is, Usually like from the computer science, AI, machine learning, we treat causation and correlation as if binary, right? As it's like a coin this way or that way. But that is really not the case, right? It's more of a spectrum.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And so if you have a model that is producing robust predictions, there is some underlying causal model. You just don't know it. And then maybe that could steer you into the right direction. for that kind of work. But we didn't look at that for this particular work.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah, I am very interested in the feedback that we were talking about and how do we capture that feedback between, for example, when I go and I'm using Amazon and Amazon is making me these recommendations and then I buy things, I tell my friends and then all of that data goes back into Amazon and how much does my contributions or my friends' contributions amplifying what Amazon is doing?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And so there's some of that going on. And then there's also in terms of like society is a complex system and the place of these tools in these systems. So the tools that help us spread misinformation and disinformation make our society unstable in that then you're not quite sure. what you are reading is true or not, right?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Thank you. Thank you for having me.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
So right now with the fires in LA, there's a lot of misinformation and disinformation going on. And it's like, Who do I believe? And maybe like you believe LA Times and you believe, you know, what you read in CA.gov and so on and so forth, but not what you're seeing on Instagram.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And so there's this notion of the place of these AI tools within our society and whether they're making our society better or worse. And by better or worse here, I mean stable versus not stable, more chaotic. And I think we can all agree that we would like to live in societies that are more stable than not, right? So there's some of that that is going on.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And I have a new project along those lines, which actually touches on philosophy, which is called epistemic instability, which is what are some stability conditions of what you know? So if you genuinely know that whales are mammals, no matter what I show you, perhaps I won't be able to convince you that a whale laid an egg. You're like a whale is a mammal and mammals do not lay eggs. Right.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And you're very sure about it. Right. But then you start talking to me and to chat GPT. And maybe if you don't know something, then you're like, as, as well as you thought, right. Then I, then you're malleable. Right. Then I can like change your mind. And then now you have groups of people who are talking to these within themselves and with, These generative AI tools.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And then basically you go from like individual to groups to this hypergraph notion. And what I'm interested in is when our phase transitions in this hypergraph in terms of what the society believe, like maybe the society believe that vaccines are good. Right. And now all of a sudden the society doesn't believe the vaccines are good.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And what are the leading indicators of those kinds of phase transitions in our society as it's being modeled by conversations formally represented as these hypergraphs?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
But now, even if you're on the fringe, because of the information technology that we have, you can connect to other people who are on the fringe and then you believe, oh, no, we're bigger than the fringe. We're actually in the middle. Right. And then that kind of thing spreads. Right. So that is one of the things I'm interested in.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Regarding gay marriage, one of the things that was interesting is I was talking to a philosopher who I just taught for a very long time at the Ohio State University, and he was teaching ethics and issues related to gay marriage and abortion, et cetera.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And he was saying that with gay marriage, similar to what you were saying, he saw a shift in terms of opinions for or against gay marriage, mostly for, but he didn't see any change when it came to abortion. And I think that had to do with the vagueness of when is, let's call the thing a baby, right? When is the actual fetus a baby or whatever, you know?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And so, and that vagueness, because like we could all agree that maybe like the day before you're about to give birth, obviously you're not going to do anything. We all believe it's a baby. But that vagueness is something that doesn't shift the opinion on abortion so much for or against. And I like that vagueness aspect of it.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
So there are certain things that are vague and maybe you will never have that kind of phase transition. And then there are certain things like the vaccine where like there are people on the fringe that our information technology allows them to connect to each other. And so it feels like a bigger thing.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And then maybe there are other aspects of information that really do make people change their mind just based on talking to other people. And so they're not as sure or as stable in their knowledge.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
So it's a work in progress right now for us on this. I'm trying to stay away from making it a psychology or a social science problem because then you get all these confounding factors. And that's what I said. It has more tentacles to philosophy. So in terms of what people ought to do in terms of their knowledge and how sure they are of their knowledge.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And so right now, the way that we're representing psychology, the knowledge or like what, you know, these things as vectors. Cause I'm a computer scientist.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Basically how much does this vector space move in one direction versus another? So as you talk with others, so you can build these like kind of simulations, right? Not kind of, you can build these simulations in terms of, in terms of conversations and see how much the vector space shifts.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah. So there's a book by Ladyman and Wiesner. And I know that you had James Ladyman on your podcast as well. He's a philosopher at Bristol. And Caroline Wiesner is a mathematician at Potsdam now.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
uh about what is a complex system and their book uh that came out i think in 2020 talked about complex systems in terms of features and how there are certain like necessary features and there are certain like emergent features and then there's some functional features where like for example our human brain is a complex system and as you were saying like if it has a shock it adapts and it still perhaps can function unless the shock is like catastrophic
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Well, when you're trying to understand the phenomena, usually you have multiple entities, like multiple people, and they have relationships with each other, right?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And so what we are not seeing, if we tie this to, for example, the AI models and how they are operating within this system, is we don't know even the role of this AI system, like how much instability is it causing in the system, right? How much feedback is it causing in the system? How much memory does it have? Right. Because they're evolving so quickly that it's not it's not quite clear.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
So this is like an open area of study of like going through these different features of a complex system and trying to see, OK, well, how do I measure it for, let's say, a chat GPT? Right.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
In fact, a lot of people say, oh, well, you know, it doesn't have a good memory. based on like what I told it yesterday kind of a thing, right? So memory is one of those features that a complex system has.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And so when we're looking at graph, like machine learning with graphs or graph mining, we're trying to find those, what we're calling relational dependencies, that like the probability of you and me being friends, given that we both like Apple products, is greater than the probability of you and me just being friends.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah, I think where it comes in, in fact, this is how it links to my new project on epistemic instability, is that it introduces epistemic instability, right? Like when my dad was getting his PhD in America back in the 60s, the most trusted man in America was Walter Cronkite, right? If he said something, you believed him. Now we don't have such a thing, right?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
We don't have a person or an institution where you say, okay, I read it here and I believe it. And then there's also like, depending on where you are on the left or the right, you're like, maybe you believe New York Times, you believe Fox News. And so because of that, I feel like one of the things that we need to do if we value our democracy is teach our kids critical thinking, right?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
It's just like, Don't believe what you read or what you hear. Question it, right? Does it make sense? Talk to different people and make your own decision and don't give up your agency. But that's a hard task, right? Thinking is not easy and people don't want to think in the age of TikTok.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
I think there was less distraction for sure, right, than it is now. I think the dopamine hit that we get by just scrolling through Instagram, TikTok, et cetera, is something that has been studied. And, you know, I'm not a psychologist or a cognitive scientist, but that people, it's just like you let your brain go to mush and you just like spend hours on it instead of,
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
maybe actually sitting quietly and thinking about a problem, you know, it's boring, you know?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah, in fact, that's such a perfect thing. I always say to my students, what is your objective function? Because we all have an objective function, and that objective function changes over time. And perhaps if all of us just think, okay, did my objective function change from yesterday or from last month or whatever? You know, it would be helpful for society.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
So as a computer scientist, as a machine learning person, I always think about objective functions. And in fact, I cannot look at a mountain range now and not think, OK, if you drop me there, will I find the peak or not? The global peak? Probably not. But, you know, like, please drop me at a nice place.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
So the gradient is with me.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Or the probability of me liking Apple products, given that we're friends, is more than the probability the prior probability of each of us liking an Apple product. So the second one that is, we are friends, you influence me. And so I like Apple products and I buy Apple products or I buy this headphone, right? Headset. And the first one is that because we like similar things, we become friends.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
I mean, as an introvert, I'm fine with it. But yeah, no, I think we see this in society now where like people aren't,
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
um as good as interacting with other people or they're not as not as courteous to other people perhaps as before i don't know maybe i'm out of an age now where i'm like oh yeah people are not as courteous as they were before um but you know the more you interact with people the better you get at them unless you interact with them the worse you get at them and so if we don't put a premium on like oh look like tina can't actually pick up the
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
As opposed to just sending a zillion emails or text messages. I think there's a value to that. And I think there is this notion of trust. Even the most introvert among us, there are a few people that we do trust. And so if it comes to a point where you trust an AI system that we don't know how it works and that it's vulnerable to attacks, then that is a problem, right?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And so, in fact, this gets us to this phrase called the red teaming that we hear all the time now that, oh, well, don't worry about it. They will red team it. And so the phrase red teaming came from the Cold War era, right? So the Soviet Union, the red team, America, the blue team, right? So, and there was a lot of this red team, blue teaming, for example, for cybersecurity, right?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
But this phrase red teaming is not well defined when it comes to these generative AI systems. And my friend and colleague, Professor Hoda Hedari at Carnegie Mellon has written extensively about this because there's no guarantee, right? So you cannot guarantee that somebody cannot jailbreak chat GPT. And jailbreaking is basically that chat GPT has put in some information kind of guardrails, right?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Like you shouldn't, it shouldn't tell you how to like rob a bank, but you can jailbreak that and it will tell you how to rob a bank, right? But there's no guarantees. It's not like, oh, here's a theorem, the proof, QED, go home. You cannot jailbreak this.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And so if you're getting all of your information from these AI systems that we know can be manipulated and we don't know how they exactly work, then you may not have a shared reality. with other citizens. And that's, I think, the worst for democracy. We really do need a shared reality to be able to withstand our democracy, to hold it and not lose it.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Well, I guess as a professor, to me, it's education. I think actually educating The public, and I spend a lot of my time educating the general public and not just the students at my university, but educating the public about how these tools work, what they're good at, what they're not good at, not giving their agency to these tools and critical thinking skills. I think that that's the way forward.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah. Perhaps we should think about how we educate people and maybe they'll see the value of education, right? And that education is about enlightenment. Education is about empowering yourself, right? So education isn't like a teacher just pouring water knowledge into your head. It's about you learning about the world and so you could do better in the world.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
As a teacher, I'm already 11 on my guitar, right? I just want you to do better. And if you do better, then I will also do better. The society will do better and we will all do better, right? And so I think part of that is maybe we should rethink about how we sell education.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
This notion of homophily or like birds of a feather flock together. But in a nutshell, like people who work on, Machine learning on graphs, network scientists who are interested in understanding phenomena, network sciences and interdisciplinary discipline. It is about these relational dependencies and like, what can we find? What are the patterns?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah, I think so. I mean, there are certain things that I have I have heard. So, for example, now there's some privacy aspects to this.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
But if you are a college and you are tracking how students are doing on their homework, et cetera, and let's say Tina took calculus and she didn't do very well on differential equations and now she's taking machine learning and they're going to talk about differential equations.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
that you could tell Tina, oh, you know, maybe you should go brush up on differential equations because they're going to talk about differential equations. Yeah, okay. You know, so there's some of that kind of a thing to like to help you. And then there's also like basically like personalized tutoring that I think AI can be helpful there.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
I use it for fun. Give me a bio of Sean Carroll in the King James style. I don't use it. I haven't used it for any real work stuff or anything. I don't trust it. That's the problem.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Oh, that's great. Yeah, I asked it some stuff about linear algebra and matrix norms, and it was really bad at it. And I was like, wait, what? Like, there's so much about linear algebra. In the world, you should know about matrix norms.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
yeah yeah so so i guess maybe the the point is let's not teach linear algebra to kids and then no no no because the whole of machine learning is basically it's all linear algebra and like quantum mechanics also so yeah linear algebra kids that's that's your lesson for today from mindscape learn more linear algebra i think it's the key to everything yeah exactly exactly but it's very good at like um
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Basically, admin stuff. So if you show it some picture of Google Scholar, put it into BibTeX, put these references into BibTeX, it does it for you. So some of those kind of admin stuff it's good at.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Indeed. Indeed.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Thank you. Thank you for having me on, Sean.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
What are the anomalies in the relationships that get formed?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah, there's some of that. I would say that, so I have this thing I call the paradox of big data, which is like there's a lot of data, but to predict specifically for what Tina wants, it's difficult, right? You don't have maybe as much information about Tina.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Now, if Tina belongs into some majority group, then maybe you can aggregate from the majority and say, well, Tina is part of this flock, and so Tina will like whatever this flock likes, right? Um, but really I feel like the problem these days is more about, uh, exploitation and going with things that are popular, um, than, um, exploration, right?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Like in the past we would go to the library or the bookstore and you're looking for a book and you would find other things. And those were, you know, they basically did that. The cherry on top of the cake. Right. The cream is like, oh, yeah, I found this. Right. And now we're really not getting that. Right.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
So when you use all these recommendation systems, whether it's Google or any other Amazon, et cetera, they oftentimes show you what is popular or what they believe you would like. Right. So in a past life, I worked at Lawrence Livermore National Laboratory, which is a physics laboratory.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And like when I would do searches there, and this is many years ago, I would get more like physics books than like when I lived elsewhere. They would sell me they wouldn't show me as much physics books, right, just based on the location, the zip code. And so there's some of that that's going on. And I feel like that is more of the problem of like not really serving the individual or exploring.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
as much as possible.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah, so, you know, it depends on what kind of network it is, right? So in social networks, for example, we know that there are two dominant processes that form social networks. One is closing of what we're calling wedges. So if I am friends with you and you are friends with Jennifer, then I will become friends with Jennifer, right? We close that triangle.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And in fact, if you and I have, for example, many common friends, or let's say me and Jennifer in my example, we have many common friends and we are not friends, then there is something going on, that there was lots of opportunities that we could become friends, but we chose not to become friends, right?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Now, there's also, of course, partial observability in that, like, maybe I didn't observe it, right? However big your data is, you're not omniscient, you don't see things, right? But we do expect that friend of a friend is also a friend. That's one. The other one is this notion of preferential attachment, right? That everybody wants to connect to a star.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And so you're interested in like, basically those are the two big patterns. And then you look at deviations from that. So a work that was done by John Kleinberg at Cornell is, He's a very well-known computer science professor. This is a while back, was think Facebook, for example. Who is your romantic partner on Facebook?
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And he and his colleagues showed that basically you are the center of a flower and you have petals around you. These petals could be your high school buddies or college buddies, etc. They have just more triangles in them. And people who fall outside of these petals and have a lot of connections to these petals are either your sibling or your romantic partner.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
That is, you are introducing them to other facets of your life. And they show that when that connections stopped, establishment of those connections stopped, it's a leading indicator that you will break up.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah. So you were talking about which connections to pay attention to, right? It's like, so those are some of the things that are fun when you look at social networks. I mean, biological networks are totally different. So in biological networks, it's a whole other ball of wax. There's not like, you're not looking for common friends.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
You're looking more for like complementarity between different proteins that serve some function.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
You're trying to understand what is going on, what is the underlying process that is happening in this network and why these links exist. Now, the one thing that makes studying of graphs and networks really interesting is that it is not a closed world. So just because you didn't see a link between me and Jennifer doesn't mean that we're not friends.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And so for machine learning where you need both positive examples and both negative examples, which negative examples do you pick becomes difficult because the edges or the links or the friendships that don't exist may because like they don't want to be friends or for other reasons. And so this what are the negative examples becomes an important aspect of things.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Indeed, indeed. So there are lots of assumptions being made, obviously, in terms of like how the network is being observed. And in fact, this is one of the big differences between computer scientists and
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
that study graphs and network scientists that are typically physicists or social scientists, where, for example, they're like, well, there's a distribution and this graph fell from it versus like the machine learning graph mining folks typically don't question where the graph came from. They're like, oh, here's data and they run with it. Right.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
And it's just it boggles the mind that like you should think about where this data came from, how it was collected, What were maybe the errors in collecting it? And in fact, this touches on a sore point for me because what happens is they don't question the data, right? They just like feed it into their machine learning AI models. And then on the other end, they don't measure any uncertainty.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
So like if you have something like, let's say, a social network that you've observed, there's all this stuff about like representation learning, right? Where basically I take Tina in the social network and I represent her as a vector in a Euclidean space, right? Like maybe with 60,000, a vector with 16,000 elements in it. So the cardinality is 16,000 and there's no uncertainty.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
They're like, no, Tina falls exactly here and it just doesn't make sense at all, right? And so then those kinds of models, given that, You didn't start with, okay, well, my data could have some noise in it, some uncertainty in it. And then you don't even capture the uncertainty of the model at the end.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
It just, there are lots of problems that can occur, including, for example, adversarial attacks or like your model is not just going to be, your model is not going to be robust. Let's just put it that way.
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability
Yeah, I think in part, one of the reasons that folks, at least in the CS side, the computer science and the machine learning side, aren't too bothered by it these days is because we are going through this era where prediction is everything. Prediction and accuracy is everything. And so, you know, there are these benchmarks and it's basically benchmark hacking or state of the art hacking, right?