Host John Markoff chats with artist and AI ethics advocate Şerife (Sherry) Wong about “Fluxus Landscape,” her interactive map - created in partnership with CASBS - of private and public organizations engaged in work on the ethics and governance of artificial intelligence.
CASBS AI Ethics & Governance Map Launches
Center for Advanced Study in the Behavioral Sciences
@casbsstanford on twitter
Narrator: From the Center for Advanced Study in the Behavioral Sciences at Stanford University, this is Human Centered. In June of 2019, CASBS, along with Stanford's new Human Centered AI Lab, partnered with artist Sherry Wong to release Fluxus Landscape, an interactive map displaying the various research, ethics, religious, and governance stakeholders that are engaged in the broader discussion around artificial intelligence. Host Jon Markoff spoke with Sherry Wong to learn more about the origin of Fluxus Landscape. They discuss the problem of what we mean when we say AI, the concepts of AI justice and AI nationalism, and the emerging use of AI in art.
John Markoff: So your project for CASBS was this visual map of sort of the world of ethical AI, is the way I saw it. Maybe describe the project as you, as you visualized it?
Sherry Wong: [Speaker] Sure, so I started on the project by making a spreadsheet, right? So I collected what I saw as stakeholders in AI ethics and governance. And I wanted to be as wide of a net as possible to have a large variety of stakeholders. So mostly when you see people do lists or look at collections in AI governance, they look at just a small subset of people. The people who are in power, governments, large research centers. And I wanted to compare things that were as different from each other as possible. So I wanted to compare grassroots organizations to an important center of AI research, or maybe an activist and an artist to Singapore's AI strategy. So I tried to make them very, very diverse because then you start to see not where people agree, but where the differences So where are the conflicts? Where are the tensions that are gonna be caused? So to do that, I did not define AI or ethics or governance. I let everyone sort of define themselves. So when you do that, you get people you might not necessarily yourself put into that group, you know? So I had transhumanist organizations rise up, some AI religions even, lots of different types of things. And when you compare those, they start to, show you that it's a really, really diverse universe and we're gonna have a lot of problems potentially.
John Markoff: And did you approach CASBS and propose the project to them?
Sherry Wong: I did. I was thinking of it in sort of artistic terms, right? It's a research project, but it's this conceptual art piece 'cause it's also trying to be like push the boundaries of where art should be made. Maybe not in the art world. There's a lot of art there. Maybe it should be made elsewhere. So I came here to have lunch to like, pick Margaret Levy's mind. And she was like, oh, we would love a stakeholder map.
John Markoff: That's neat. Okay. And okay, I want to ask you about how you frame yourself as an artist. There's this— there are digital artists, there are analog artists. Where do you fit on that spectrum?
Sherry Wong: So I used to be a painter, but even when I did paintings, I thought of myself as a conceptual artist because I think that the idea is what comes first. So it doesn't matter how you're making it and what your medium is, you're just trying to communicate between two things, between me and the audience. So if I'm communicating through a painting, it looks like a painting but might not be about painting. It might be about like there's language or representation. So I work a lot with the meaning of what words are.
John Markoff: And let me stop and ask you, is it possible to to give us the URL or a pointer to the visual map so people can find it? Because I think it really is quite— it's worth finding.
Sherry Wong: Oh, thank you. So if you just Google Fluxus Landscape.
John Markoff: And where does the term Fluxus come from?
Sherry Wong: I'm glad you asked. It's one of my favorite words because it's my favorite art movement. It's from the mid-century Fluxus movement that started in New York, where life is art and everything's sort of art. It was a very playful movement, but it was It's also a little bit political towards the end. So I'm trying to position the art in that world of conceptual art, but it's also in flux because AI governance and policymaking and that whole world is in flux. We don't know what's happening. And it's a landscape because it's a landscape map, but it's also a painting like a landscape.
John Markoff: Did you build your tool or was that tool that you built the Plexus Landscape in around to to populate?
Sherry Wong: So I didn't want to spend time building a tool, and I think that everyone should just be able to make whatever they want. So I used an existing tool called Qumu, which is a very easy network mapping tool that you can just like plug a little node in and then connect other lines to. And you can actually even download my map and make your own map off of it. So it encourages people building off of an existing landscape.
John Markoff: Yeah. So my experience in sort of my— the time I browsed the landscape was, first of all, whoa, there's a lot going on. I mean, I was struck by how much you found. And the next is then I got lost in the detail and sort of diving into little groups that I knew nothing about. Are there other sort of things that you learned from the project about— I mean, you talked a little bit about terminology and what else sort of was emergent from sort of putting the entire, whatever this subfield is, down on paper or on display?
Sherry Wong: Yes, I think we do have an emerging subfield. I think that out of AI is coming out this like AI ethics movement, and when you look at it, it's not so easy to categorize because when people say ethics, they mean a lot of different things, and when they say fairness or privacy, they mean a lot of different things. So you have people who don't agree with each other. You have Google on there, and they have several different ethics projects, and they say they work for human rights and they support privacy. But then you also have the Google Walkout movement on there, and they're using the same language. And so they're categorized under caring about human rights and privacy, but they might not necessarily, like, see eye to eye.
John Markoff: You also use the term AI justice. I mean, is that— does that sort of frame your perspective in some sense? Is that a good way to identify sort of your intent? And where does it fit?
Sherry Wong: See, that word's interesting for me because it came as a result of doing the research for this project. So in art, when you're trying to draw something, you often look at the negative space and not at the object. So if I was going to draw a vase, I wouldn't look at the vase because I'm going to draw what I think the vase is like. I look at what's around it, draw everything but the vase. So when I was researching ethics, I would see all these principles and the word ethics over and over and over again. And when you see a word over and over again, it loses a lot of meaning. And I would look at, um, these strategies and lot of different principles that everyone's coming out with, and they didn't seem like they had a lot of meaning after a while because Did anyone change their behavior because they wrote principles? I didn't see any behavior changes. You know, and what— so it wasn't— when you remove the vase or remove ethics and the words, what did you have left? You had like agendas, right? Goals, power. So that's what I was kind of like confused about. So I started switching over into words that were coming from social justice movements. 'Cause they have a different sort of history around it and they're not used in the same way. And at the same time this happened, it was, what was trending was the co-opting of ethics by mainstream sort of large companies. And there was a big backlash, the techlash against using ethics or having ethics boards.
John Markoff: Mm-hmm, mm-hmm. No, you mean, well, yeah, Google in particular.
Sherry Wong: Mm-hmm.
John Markoff: Yeah. And so, is it possible to— I didn't try this— is it possible to sort your dataset by corporations and NGOs, or in some for-profit, nonprofit, or if you sorted that way, what would emerge?
Sherry Wong: So, the categories are color-coded, and they always sort of remain. And that was difficult, because whenever you categorize things, you have problems with categorization, because you have your own imprint and your own biases. But I categorized into corporations and businesses, nonprofits, all the think tanks and foundations went under there. I also have some individuals categorized because the artists and activists weren't able to be represented on the map as stakeholders unless they were individuals. And then there's some hybrid organizations and I also have like other types of Businesses. Gosh, I wish I had a categorization to look at.
John Markoff: Yeah, but that's— and if you— you also made links. Could you do social network analysis on this? Have people done social network analysis? Is research being done now on the map you created?
Sherry Wong: I had a friend of mine who was going through some of it to work on misinformation campaigns, where she was running things a little bit tighter. So you can actually compare closeness and in-betweenness between the nodes for like where activities are taking place. But I also purposefully like left my own biases in for what I was looking at because for example, there's a lot of things from California and Berkeley because I am from here and I tried to like gather other things as well. So my friends that were helping me out with it, Aparna Ashok, She is from India, so we brought in a lot of different organizations from India. And they talk about things a little bit different. So when you do categorize and run statistical analysis through it, you're not going to get correct quantifiable information. You're going to get biased information that's filtered through our biases.
John Markoff: How common is the phrase human-centered, and what do you make of it?
Sherry Wong: I tagged it. So if you go on the map and you look up human-centered, you will— the groups that use human-centric or human-centered will pop up because this is pretty common. It's growing.
John Markoff: As this became a controversial issue, as AIs became controversial, there was a rush by technology corporations to— well, they would say they were IA companies, which was a way of saying they were intelligence augmentation. They were not AI companies. They were, you know, using AI to augment humans. That was sort of more palatable, they felt, Well, certainly from an ethical point of view, but I was wondering whether I could trust what the company executives were saying, whether it was simply a device that they were using to make AI more marketable or more— Did you find anything like that? Did you come to distrust any of the corporate use of the ethical language?
Sherry Wong: Yes. I think that you're— What was revealed is sort of like what's really, what's behind those words, your agendas are, you're trying to make more money and sell more tools and sell products. So companies are thinking of AI as tools and products, whereas, you know, civil society is thinking of it maybe in different terms, especially if they're trying to push for policy and regulation, they think of it much more like in how they're deployed. Almost all the words are heavy and politicized. I don't distrust companies completely, right? They're just playing into what they're supposed to do, what their stakeholders want. They have to make profit. So I think that accountability lies with government for giving them the correct incentives so they can make different decisions that benefit humanity and not ones where our privacy is sold off.
John Markoff: You tag government as well? Are government organizations sort of visible?
Sherry Wong: I did, yeah.
John Markoff: And the other thing I wanted to ask is, so are you maintaining the project? Is that like a full-time job to just keep up with what's going on?
Sherry Wong: No, I placed a date on it, so it exists as sort of like a time capsule of that time. But I have been sort of following everything because I'm obsessed with the world that's changing around me. A lot of things have changed since then, which is only like 6 months ago when we published. But you have new emergent themes like AI as an ethical imperative for climate change is a sort of new thing. Even though we've always had AI for good or AI— not always, but AI for beneficial reasons has been a theme. It's now, it's imperative, an ethical imperative to do climate change. And we've seen a rise in AI feminism and more people talking about decolonization of AI. We've also seen a backlash against, um, sort of competing narratives. For example, you have the narrative of surveillance capitalism, you have the narrative of platform capitalism, and also the attention economy. And they're playing with the same ideas, but who they reach and how they talk about this, the same kind of system, is very different. And it's for their agendas are a little bit different and why they're using that language.
John Markoff: You're here today at CASBS because of the Berggruen Foundation, and CASBS sponsored this workshop on metaphors and AI. Did you come away with sort of new ideas? At the end you talked about a painting that you had in mind, and I don't know if you had that idea before you came or whether that was something that emerged out of the conversation.
Sherry Wong: Yes, so I'm going to be working with Maya Indira Ganesh to illustrate her research on how different groups are using metaphors around AI. So you have people talking about data as the new oil, that's a metaphor, or AI as war, a sort of a struggle that we're competing to the race for AI, like who gets AI first. AI is replicating how we are, right? Because it studies our data and then spits out more. And it kind of narrows our field of possibility into like, well, replicate the same injustice that you had, replicate the racism in the criminal justice system. It just replicates. Well, that's not how we want to be. Like, humans want to be something bigger than that and greater than that. Like, what would that look like? And how do we design for more open possibilities that's not fed upon the data that we've created?
John Markoff: Are you living and working in what we call Silicon Valley now? I mean, it's kind of an amorphous thing, but the Greater Bay Area, I guess, is sort of continuous almost with Silicon Valley now.
Sherry Wong: I'm reading the DAOS. I live in Berkeley.
John Markoff: And is there— I mean, are you immersed in sort of an AI culture in a sense because you're around? I mean, there are meetups, there are organizations that you can sort of travel in for ideas, activist groups?
Sherry Wong: I'm deeply immersed. You cannot escape, I think, not being part of the AI, I don't know, force almost brewing around you, which, I mean, that's how I came to it. I'm an artist, but I was working in technology at Autodesk at a place called Pier 9 where they work in the sort of the future with lots of robots.
John Markoff: Oh, you're an artist at Pier 9? Okay, well, maybe I walked by you on a private tour.
Sherry Wong: And a programmer.
John Markoff: Yeah, okay. Are you still there?
Sherry Wong: No, no, no, I've gone full-time into doing art and activism on AI. But that's kind of, you can't escape it, 'cause everyone's working with machine learning or else they're adjacent to it or they're not and they might just be using the term AI.
John Markoff: Do you consider yourself an activist or an artist or both? Where do you put yourself on the, I mean, is there a community of activists around AI that you consider yourself to be part of?
Sherry Wong: You know, the community's pretty broad and it's everywhere. There's a lot of people. When I first started researching this, I thought to myself like, "Oh no, what should we do?" And like, "Why isn't anyone doing anything? This is terrible." And the more you stay in it, you're like, "Oh, there's thousands of people doing things. You just don't know who all of them are." And I'm meeting all these wonderful people, a lot of women, a lot of people of color who are working in AI who are— I was trying to like pass laws and policies and they're informing people. I don't consider myself an activist all the time except that I am in sort of action. I am always artist first and foremost, but I think for every artist you have to sort of be an activist now because what else are you going to do? Like, the earth is boiling, you know, like injustice is right there. We're in a transition period and if you don't respond to that, you're not going to be able to do your art for much longer.
John Markoff: I mean, you were— were you at Autodesk when you sort of ran into the reality of the AI, sort of explosion of AI technologies, and that, that's sort of where you sort of began to look at this?
Sherry Wong: And yeah, I was working, um, with really wonderful creative technologists, and a lot of them work with robots and machine learning and just creative things and they were doing generative art and also generative projects and manufacturing. So, if you have a computer that's able to design a car without a human at all, like it's this creative tool, it's like, well, what are you gonna do, graphic designers and artists? They're taking our jobs. So, I started getting concerned about the future of work. And then when you start pulling on that thread, it leads to, Wait, it's not even the future. What about right now? What about the present of work? What about gig workers? What about, you know, all the Uber drivers? And it's just like, oh wait, what about like where they're using AI for interviewing people for jobs and emotional recognition systems that don't actually work? It's—
John Markoff: You mentioned generative art. Day before yesterday, I had dinner with a friend former CASBS fellow, Hector Postigo, and Hector is playing around with GPT-2, which is this interesting language model. He's done some remarkable stuff with it that raises all kinds of questions. The first project he did is he basically took GPT-2, which has now been released to the public, and he fine-tuned it by importing 700 papers written by people in the field of AI ethics. Then seeded it and asked the AI to write about AI ethics. And it came up with something remarkably interesting. I mean, if I sort of stumbled across this in an academic setting, I would think I might have been fooled. But then he tweaked it some more and he got it to write music. And it wrote fairly believable classical music. And then he gave it— I'm blanking on the name of the particular hip-hop artist— he sort of played with a hip-hop artist, and he's going to talk to the Copyright Office next week because, you know, we're deeply into this new space. Who gets credit? Have you seen people playing with it?
Sherry Wong: Oh yeah.
John Markoff: So it's out there.
Sherry Wong: Oh, it's great.
John Markoff: And what do you think will happen?
Sherry Wong: Well, so my friend Tahir Hemphill, when I was working with him at Autodesk, did a project where he taught an AI like rap music because he has every lyric possible so he can feed it. And then it was producing pretty good stuff, like it was good. And, um, and I've seen what GPT-2 can do and it's scary.
John Markoff: Do you have a sense of— so you were talking about the future of work. What about the future of art? Yeah, I mean, in that same—
Sherry Wong: I had a dinner party about the future of art. So here, here's where I fall on it. So, um, my which I'm going to steal my friend Memo Akten, who's an AI artist, his concept. Can AI be an artist? And he's like, well, as soon as I start to care about why an AI did that art piece, then yes.
Narrator: Do you use any— have you ever experimented using AI as a generative part of your art?
Sherry Wong: You know, I haven't because I don't— I like to use forms that are really accessible to all people. 'Cause I think everyone is an artist and everyone should be an artist. So I wanna unleash that kind of area. But I am looking at it and I've helped other people on their projects. I've worked with artists on making recipes for dinner parties where you feed an AI lots of, sort of neural network, a bunch of recipes and spits out dishes that are, not delicious at all. They're kind of random, but you can't— you cherry-pick through it because it was gibberish. But these days it would be probably like good recipes. But I, I don't think of, um, or like you can feed pieces of music into AI and you get some good, good results out of it. A lot of our AI art though, I, I don't believe is quite there. It's still a new tool and I think that needs to mature. There are some artists who are working with it on a conceptual basis where they have this nuanced understanding of our relationship to it. But you have to be very, very good to use AI in your art because it's pretty gimmicky still because it's new. So when you're using a new tool, even if it's VR or AR, if you notice the tool first because you're like, wow, that's so cool, then maybe it's not the best art. You know, it should just be a vehicle for something else that you're carrying.
John Markoff: The Valley as a place really swarms in terms of investments. I mean, it went from all the money going into social media to all the money going into machine learning in terms of venture capital investment. And so there's this explosion of these sort of new industries. And I wonder where we are in the arc of sort of this AI spring. You know, people talk about— we've talked about AI winter, but we're clearly in either AI winter or AI spring or AI summer. Or when does AI fall happen? Do you have a sense of how far this technology will take this region?
Sherry Wong: So people are predicting an AI autumn because of the hype, right? Because you don't have a lot of AI tools living up to the expectations. You have a lot of snake oil.
John Markoff: Well, something that Silicon Valley also does very well. Just a short story. I was around the New York Times when the Enron scandal happened. There was an investigative reporter who asked me to come in and help, because Sun Microsystems got kind of pulled into it. He showed me the press releases that Enron executives went to jail for. I was looking at them, and I'm going, "This is just a standard Silicon Valley press release. I see these things all the time." And what? These guys, poor Texas guys, are going to jail for doing the same thing. To your point about hype in the Valley, sometimes it's—
Sherry Wong: Well, I mean, it's like, it makes sense. Like, AI is a predictive tool, and we are seeing a rise in metaphors and narratives around magic and fortune-telling. And so it's sort of like a gambling thing. And I think one of the stories I like to pair with Silicon Valley is The Emperor Wears No Clothes, because you're left holding the bag. But it's so easy to point fingers too. It'd be like, oh, look at you guys being so bad. It's like, well, it's hard to disentangle how all these small parts made that. You know, I mean, and this is your area too, right? Counterculture got together with, you know, all these academics and research scientists, and it gave birth to this beast of a thing. But I think we're seeing a big narrative switch So I'm not sure what will happen. I worked at a hedge fund. I'm pretty sure that past performance is no indicator of future performance. But I did learn that the narratives matter for economics. And the narrative that I'm seeing emerge that's pretty recent is a shift to AI nationalism.
John Markoff: Interesting. As in China versus the United States?
Sherry Wong: As in, yeah, for our battle for AI dominance, it's a point of honor that we must be driving this narrative of AI at all stakes. We must be the ones to get AI. Get what? Like, is there a prize at the end? What is AI that we're trying to get? Like, I thought we were already doing AI according to the press release. But we have to change those narratives, right? We need different stories. Like, nationalism doesn't have to be sort of a nationalistic kind of thing that leads to tribalism and separation. It could, you know, a sense of pride and collaboration Like, what if it was a collaborative? Because AI is already collaborative. Like, we're pitting like China versus US all the time, and the researchers work together. They're, you know, maybe sometimes we don't want them to because it leads to like greater surveillance and better tools for surveillance, but everything's woven a little bit closer than our stories tell.
John Markoff: Yeah, it's interesting that, you know, you're, I guess, In terms of AI nationalism, I think it was probably distilled by Peter Thiel in the op-ed pages of the New York Times. He sort of wrote the canonical op-ed piece arguing that we're in this war with the Chinese and we need this dramatic separation between the Chinese and Americans. And the Chinese have made this national commitment to being the dominant— whatever that means— the dominant force in AI by 2030. So that's clearly AI nationalism. And yeah, and it gets confused with all the metaphors because then you have to ask, what is AI? Maybe it's just statistics at this point. There's nothing really.
Sherry Wong: Spreadsheets.
Narrator: But even there, we talk about China as this surveillance state, and we are a 2-minute bike ride from Palantir.
John Markoff: Yeah.
Sherry Wong: Right. When you point fingers at the other for being bad, it comes right back around, points right to you. Because if you look around, we're doing very similar things, you know.
Narrator: Earlier you said, you described the map and then you said, "We're in a lot of trouble." Oh. And we've talked here, you know, and I think I've pieced together some, but is there any kind of, you know, is there any summary of the trouble we're in or the weather you see in this regard?
Sherry Wong: Yeah, it's so different because when we say AI, we don't know what— we don't have the same language for anything. And I don't know if we need to, right? We don't necessarily— we've never had this shared language and we've gotten along through life and we've got— things have gotten a little bit better perhaps. So this— I see a lot of schisms rising though between just the same ones maybe we've seen historically. And I'd like to maybe look at the history of things a little closer for my next projects. But Differences between how are you supposed to make change? Do you work within the system or do you want to tear it down? We might have the same agenda or goal of, hey, let's make better, change things, but how we do that is a big difference. Who's funding it is another big question. I'm seeing a lot of schisms actually rise up in the AI ethics community. Which maybe they're not necessarily a community because they might not all be located in the same place, but there is a sort of communal feel because people are friends and they go to the same conferences and they talk about the same things. You know, how do we make things transparent and fair? How do we have accountability? Right? So they should be on the same team, but they're, um, it's like, well, you take money from so-and-so, I do not. And, um, I think eventually these things can crack And maybe there'll be a time when we'll have to actually work together and it'll be difficult because it's hard to take backwards. It's hard, you know, when you say I'm better than you to someone, they don't like that. And who wins then? Like, we might have to collaborate a little bit closer. We're going to have to create a sort of policy or a new reality together. And you can't collaborate if you're fighting. And it's all trade-offs anyway, right? You can't have privacy and safety both at the same time. At the same time.
John Markoff: It's a classic lesson from the computer security world, right? Yeah.
Narrator: Is there a particular work of art— could be a book, a photo, a painting, a sculpture— that really inspired you to get engaged with this topic, to think about these topics?
Sherry Wong: Oh, this is— is that a question from the blue, or did you look up— wow, because yes, deeply so. One of my favorite paintings is The Fall of Icarus by Bruegel, and I named a sort of— I'm working on an art think tank, sort of like a research— you know, think tanks are research centers, right? But what if they were art centers made up of artists? That's my idea. So I'm calling it Icarus Salon because of that painting. And the Icarus metaphor people have used before for describing what's going on in AI, because you can imagine the inventor And they've come up with this, you know, new tool, AI. And it's like, don't fly too close to the sun because, you know, you don't know what you're doing quite— you know, ambition. Don't be so ambitious and that your wax melts and you fall in the water. But then don't fly too close to the water where the wingtips get wet and then you'll drown. So there's that. But the painting is not about that. In the painting, you can barely see Icarus. His feet are just sticking out of the water. And he's invisible. The painting centers around everyday life of a man plowing the field, a shepherd looking off into the distance, a fisherman in the corner, all this other regular life, everyday people. And they should be the center. That is the story here. It's not like, oh, the inventor that got away or, you know, the tech company that's like drowning over there. That's not the story. The story is like, what impact is this having on everyday life?
John Markoff: Thank you for spending time with us. That was great.
Sherry Wong: Thanks. Wow, I can't believe it. Such an honor.
Narrator: That was artist Sherry Wong in conversation with John Markoff about her work titled Fluxus Landscape, which shows the organizations engaged in researching, building, regulating, and thinking about the future of artificial intelligence. There's a link to Fluxus Landscape in the show notes for this episode, as well as links to Sherry's website and the Icarus Salon organization, which she describes as part art production studio, part think tank. Human-Centered is a show from the Center for Advanced Study in the Behavioral Sciences at Stanford University, and it is produced by Michael Gattani and Joseph Munzel. Thanks for listening.