 All right, thank you all so much So you've been hearing a lot about the Navy steel and I just want to say that that is not me Our team is interesting when peaks unfortunately couldn't be with us here tonight But we're a small team although the image here might mislead you We are small in part because we're working on a pretty controversial topic and that is the topic of algorithmic warfare That's what brought us together and in particular when people hear that term They think of a lot of different things. They may think of the sky the terminator Skynet Thinking about robots and here in Cambridge Boston Dynamics But we actually wanted to challenge that paradigm a little bit when we began this project We were looking into what is the role of AI in warfare? What is algorithmic warfare and we came to the conclusion that there's kind of this interesting misunderstanding, right? So there's a really big effort underway to address the dangers of These kinds of robotics and it's real right that's a valid effort But there's also this very real threat that a lot of people kind of overlook Which is that at the end of the day, it doesn't actually necessarily Rely on a robot being the delivery mechanism of violence if you're still being targeted and so that's really what inspired this project We wanted to look at how surveillance how these algorithmic decision-making systems and surveillance systems Feed into this kind of targeting decision-making and in particular what we're going to talk about today is the role of the AI research community How that research ends up in the real world being used with real-world? Consequences and then talk a little bit about what we found in our investigation of the space and what we invite you all to join Us in doing as we move forward so to take a step back There's a pretty diverse crowd here So I just want to Contextualize a little bit about what we're talking about when we're talking about this kind of algorithmic surveillance here You're seeing depicted a video surveillance system that has image recognition technology So it's identifying it's putting boxes around various people and vehicles and so on So we're looking at a couple of different types of systems here computer vision systems systems that can listen to voice or audio As well as systems that may look at social networks or things that we post on social media to come to conclusions about whether or not We fall into a certain group now in this project in particular We're talking about an incredibly rich and interrelated system And I'll talk a little bit about this data visualization in a second But I want to first just contextualize that when we're talking about this research community we're talking about a lot of different players in a global ecosystem and You know as researchers we want to share we want to collaborate and that's a beautiful thing But what can be challenging about this space is that these threads of connection could lead to places that we didn't Originally intend and we may as researchers not have even been aware And so our work in particular is exploring how some of those outcomes how some of the end uses of Research that may be done with the thought of being theoretical or benign can actually have real harm in our world So with that I'm going to turn it over to Carl I'm Carl. I'm the one they've been talking about And I'm gonna teach you how to hunt people So Many practitioners in this space will often say that the technologies it just isn't mature enough to be a true threat You couldn't use facial recognition alone to build this target deck and I promise you that grossly undermines the threat How you hunt people is actually with a series of overlays You don't use one individual technology you use them to hone in and find the corollaries to build a smaller List until you have your approved target deck so if I wanted to identify all of the Protesters in Hong Kong yesterday. I wouldn't just use facial recognition of CCTV camera footage I would take that and put that on an overlay of Metro cards used in the subway system in Hong Kong right and that might give me a more accurate deck to begin our interrogations So jumping off of that short example a quick deep dive So I'm not exhaustive, but if you're not familiar with the Uighur crisis in Xinjiang, China I'll give you a just a couple of sound bites. So since 2014 the Chinese people's War on terror has been a systemic or systematic effort to oppress the Uighur population Which is an ethnic minority in China that practices Islam roughly 10 million individuals So 1.5 percent of the Chinese population, but also does account for 20 percent of all arrests in China Consequently there are over 1 million Uighurs currently in Assessed to be interned in Xinjiang province in Re-education camps. These are the camps of you know, Holocaust lore where there's killing torture Primarily this method of assimilation via submission It's a heavy workload. Genocide is tough. So the People's Liberation Army has leveraged the tech industry So this region this small province and cross-section of society accounts for about 7.5 billion dollars of the security industrial complex in China And they leverage both facial recognition voice recognition and other forms of AI to apply Lists of people who are either demonstrating Islamic practice or they demonstrate phenotypic features of this ethnic minority Now Where you stand depends on where you sit and in the interest of positionality. We're going to bring this home to MIT Bunch of institutions in America here serve as key nodes in the surveillance Supply chain, but here at MIT Through our research we found that there's about 15 projects ongoing in some kind in some way shape or fashion for surveillance type of technology so this The nature of this ecosystem lends itself to this really complex interdependence. So The research ongoing here at MIT is done in affiliation with or with funding from both private and public institutions So private institution would be NEC is a Japanese based security firm That's basic of the product is predictive policing. We use it here in the States It's used here in the UK other Companies that have contributed funds to this type of research at MIT include sense time and I fly tech Those two companies are both implicated in the provision of services in Xinjiang province Also one fantastic supplier if you can use that term Funding for this type of research is the US government So the same researchers that are linked to these companies are also receiving funding from DARPA office-enabled research Marine Corps Warfighting Lab and Army Research Lab Occasionally they actually appear on the same publications in the acknowledgments section So we wanted to present this and for some folks in the room It's probably new information But we are not the first people to talk about this issue the issue of surveillance tech is Really important and it's been an emerging and burgeoning conversation here in the United States as well as around the world There are actually folks in the room here who are experts in the topic But really there's this common theme here, which is that researchers and product teams have an important role to play in Actually helping to think about how this technology is used or prevent it from being used Irresponsibly and in particular I wanted to talk a little bit about what we've covered so far So we've looked at in reality thousands of different projects and contracts from the US government or journal Publishings and we honed that into a couple hundredth list of Surveillance tech specific and then we ultimately came down to about 65 research institutions here in the United States with 49 funders as Carl said ranging from Chinese surveillance tech companies to US surveillance tech companies to our own US government and they covered a broad kind of breadth of different types of surveillance research including facial recognition social network analysis and Person re-identification which for me was a new phrase which is a little bit like what Carl was talking about at the beginning of the presentation here So what we've been doing is putting together all of this information and really trying to literally visualize it So I know this kind of looks like scribbles on the screen here But this is a data visualization we've been putting together Representing all of those different connected nodes and ultimately what we would like to do is to be able to move forward on a Project that has a better understanding of the reality of what's on the ground We want to know is this data reliable? Can we get access to more information for instance today? We were only using public sources and we also want to really involve the voices of the people who are developing these tools one of the common themes in the work that we were doing to research this project is that we found that a lot of the People involved actually didn't know How their work was being used in the world and that is both concerning But also a really exciting opportunity for us to educate and do better But ultimately there are also a couple of things that even if you're not a machine learning research Or even if you're not in this kind of surveillance tech space There are a couple of key points that we would really like you to push on as you're having Conversations with your representatives about what we can do about surveillance tech at least here in the United States So this is where we go from fact to opinion. So the opinions here in are not representative of the US government But they are ours. So call to action two points that through our research We feel like we can apply some leverage to this problem set and ideally make a difference rather than admire this problem first is Department of Commerce That's the office that holds the export administration regulations So these are essentially regulations that govern the export of commodities to include dual use technologies and software So in adding facial recognition and surveillance technologies to that list We would be a forcing function for anybody that's in this complex ecosystem to have to apply some level of risk mitigation and threat modeling to their decision calculus and then finally We should probably call for a moratorium on US government funding for the research of these technologies Not saying that the requirements are not going to be fulfilled by the US government but the acquisition of this technology should probably happen in a highly regulated commercial market as Opposed to being directly funded because the complexity and the low density of these skill sets Just don't lend themselves to ethically Sound and transparent funding profiles. So let's take that basic research funding and put it somewhere else Thank you