 This is not about generally taking over, although perhaps I've talked a little bit about events. But rather to welcome Cashier Hill, to which I'll ask his first client, to talk about her new book. There are copies around that are yours to use. So this book belongs to you. Thank you Cashier for that generosity that you've been offering her. Thank you Random House. And I've known with Cashier's work for a very long time, and been consistently blown away by it. I remember when Cashier did a series, I think he wrote a series of kind of facts. I figured out what we like to unplug from the various letters of like the fame it had for him. Yeah, it was sort of a Google, Amazon, Facebook, like it's often happened. And all at the same time, but also these videos. Yes. Yeah, so that's itself a story worth telling at some point. But we have a story like behind us and in front of you, and for at least 150 of them, asked themselves a lot of those in the sky as a word. To talk about a phenomenon that ran under the radar for quite a while, it's entirely possible that I first found out about it in the movie. And when I found out about it, I somewhat pride myself, I think, on not swinging at every inch, not being excitable about every possible, this changes everything a moment. And what amazed me about Airbnb AI, the company first switched water, was the way in which what it did was an automation of holding together just a few basic things that anyone could have done. In fact, anyone did. I'm looking at us and I'm going to catch here the founder of your views. Reputation is well captured by his previous big app was a Trump-like hair. Any photo app in the app store, which probably is so cool. I couldn't have seen the photo. Yeah. And then thinking about the bottom line, called if you've got nothing, you've got nothing to lose. He didn't want Facebook or Microsoft or others out of here, even we not have touched this kind of a poll, which was to spray billions of their chips, get whatever nearby tagging or negative data or other information about them might be available, especially if you're getting no files on LinkedIn or Facebook or Insta or whatever, do that a billion or more times to use some basic indirect recognition to say if I'm offered a photo of someone, a live stream, although somebody walking down the street, maybe it's never been on the net before, he's an accuracy and we identify who that person is. Implications have been able to arbitrarily identify people. I'm pretty staggering. When I first found out about it again, I was just like, why isn't this a 501 fire? I'm not generally excitable, but I'm feeling very exorcised right now. And isn't this what privacy regulation is for? Isn't this what... I co-wrote a short op-end in the Washington Post that proposed that clearly may I be metaphorically burnt to the ground when he burnt the salt and lure which resulted in the CEO reaching out to me as part of a kind of yard campaign that's been good and cashmere will tell us more about. And here we are in 2023 and it seems like the cap is pretty much out of the bag. As you hear about the twists and turns of that, discussed together and maybe even think about what, if anything, ought to be done next and how would it be done? And if nothing can be done and you think something should be done, how is there any hope for privacy more generally that this classroom hypothetical comes to light and cannot be duplicated? This is a good question. So cashmere, over to you. It would be great to get into the topic of happy life and maybe a little bit about your sort of disposition towards your recording what you bring to the act of being an identified as professional tech reporter and how you got into this trail specifically. So nice to see everybody. See, I identify as a tech reporter I also identify as a privacy fragmented and part of the reason why is that I've been recording on privacy and technology for more than 10 years now and it's been to have my lens on technology and started as a local law in 2009 called the Not So Rather Parts and it was just this religion of at that point basically the internet with our notions of privacy and running about, you know, Facebook kind of convincing all of us, for example we put our photos on the internet against to our faces which at the time people were troubled by and it didn't lead eventually, or 10 years later to something that we forgot that we put those two things together and rather than looking for the of my name and finding out what they look like you can open up my face and find out what they look like. So yeah, I first heard about Azure AI in the fall of 2019 a public records researcher who stumbled upon it asking this department about what kind of technology they were using and at that point please I think that means in case you are thinking about technology for 20 years back in 2000, 2001 really got a bump from September 11 I didn't really work that well as far as we knew the algorithms were pretty clunky for bias, you know, like did not work as well on some of these other groups and things were limited to essentially criminal mug shopping bases and in some states too, driver's license photos and design is actually just a big report about how the things were mixed when it came to major organization technology at least for sure how useful a tool it was but this public records researcher got this PDF from the Atlantic Space Department and it has advertising materials for through AI describing themselves as Google interfaces stop searching start solving with their tagline and there was a legal memo in there written by Paul Clement a very high profile lawyer former U.S. sister general one of the many, the high profile lawyers in the orbit of what turned out to be a pretty scrappy startup people right now are like a possible job and you're talking about it all the time and can you describe what your UAI had done that they have scraped billions of photos from the public internet you know about anyone who's since at that point in time if you have a million faces in their database today they have 30 million faces who said that they have this app that works on the like 9.8% 6% accuracy Clement said you know my lawyers have tried it within our firm it returns fast and accurate results and he had an entire pipeline interview with the LIGIS demo to reassure police officers that they could use their UAIs without breaking the law that they state privacy laws or federal laws and just one quick question there as we go you said I think considerably police officers rather than police department their view was just approaching police officers directly they were offering free trials on various inserts one called Climdex for example for running a little on crime investigators and they would say hey try this out it's a free trial and it's allowed I heard that NYPD was using it and I went to NYPD as a digital spokesperson that are using clear UAI and they said they weren't which later turned out to be not the truth and what they would later say is it wasn't coming to our firm in office this is what I think I need to report is that officers don't load any technology they want and start using it in investigations without having it bedded without knowing the background of the app without knowing how well the algorithm work or who made it and I was shocked by how that and many other things about the algorithm as soon as I saw the memo the thing I flashed back to was when I said first-party coverage privacy there had been this federal workshop in Washington D.C and it wasn't funny about it and it's like all the big tech companies like Google was there Facebook was there Apple was there privacy advocates were there the e-sale of you was there for democracy and technology Facebook was rolling it out to tag your friends and photos and suggest their names Google tried to use it to unlock the android so it didn't work that well and it happened and lost the android leaving Google to tell its employees internally don't use this it's not secure even though they were offering it to their customers we all gathered there and talked about what do we do about this and what do we do when it actually gets really good and those kinds of people in one room don't often agree much but they all agree the one thing that no one should do is build an app that you can just take a picture of a stranger and so it was very shocking to me to see this little startup that I hadn't heard of that no one seemed to have heard of when I talked to experts that they had gone ahead and done it they had accomplished this and I just immediately I read this this e-mail I got in public life I need to find out if this is real if it is please do find this and that was the beginning of yeah I did this x-ray in a different time so you probably read that and then I was obsessed with how we got here and how the technology got and how powerful and why we had never built any hardware to use on insurance this isn't much of a story about people it is about kind of common forces or institutions but just to tell the rest of the kind of stage setting there is a little bit more of these institutional pieces so we mentioned that this database started with 3 billion not 30 billion this was all scraping in terms of service we know this for instance is academics that if we are going to try to do even just baby steps research on something and evolve scraping a website to get data they are probably violating in terms of service of that website scraping is to say to send along a bot not in the script read what's on the page what was the reaction of the likes of facebook which true to its then name is pretty clear epitomously titled what it is name when it gets up facebook but it's meant for those individual searching as we were talking about and we're not supposed to just grab everything and you would think it wouldn't go into the facts of its users by closing that door what it might have been legally how did the likes of those who ran the platforms of the scrape how did they react so the company is we're not happy when they found out about things as a clear view and they express that by sending so facebook sent such a letter then though which is actually one of the first big sources of faces that one has had in the book but it was one of the first big hits he got google linkedin they all say clear view letters and hey you're violating in terms of service you know we see this as a cfa a computer fraud abuse computer fraud violation stop doing this delete the information from our site and then that was it so they invested the cost of a stamp in fact they sent the letter by a regular mail and do you know what it was I don't know I don't know I can't say for sure whether they replied or not but they didn't they didn't send the letter so the next step if you're a company that sends a cease and desist letter I know there's a lot of lawyers or law students in here so you know you would tend to sue but none of these companies have sued clear view AI and I think the reason why is in part because I don't think they wanted to bring more attention to the fact that they were elected from their sites but also because the law around scraping is is pretty fraught and there has been at least one big federal court decision that said it's you know it's not illegal to scrape and in many ways I think people that are advocates for digital rights want us to be able to scrape as a journalist I've run into this before I did a story I did a series one year about Facebook's that suggests to you people that you might want to friend because perhaps you know them in real life and we are trying to understand what was the data going into that how are they making these creepy connections and you know linking psychiatrist patients for example to each other suggesting that they might know each other and so my colleague and I he designed a scraper that would scrape all the suggestions that somebody got so technically something like that is in violation of the terms of service but there's many ways in which we want to be able to scrape and so as a kind of digital rights community I think we're torn on that and now it's coming up again and again not just with Clearview AI but with these generative AI companies that have gone out and amassed these databases just by scraping the public internet collecting your comments on Reddit blog posts, New York Times articles artists work and I think as a society right now we are feeling a lot of anxiety about this that the public comments is kind of getting used in ways that we did not expect when we put our information out there in some first year property classes taught in this very room sometimes on the syllabus will be eBay versus Bitter's Edge a now venerable case in which one website called Bitter's Edge which nobody's heard of you can probably know how the case turned out with scraping eBay in order to do a meta auction site so you could search for toy truck and see what Yahoo auctions remember Yahoo auctions and eBay and others were doing and eBay did more than a cease and desist there another reminder that markets eat norms for breakfast and in that case it was seen as a threat to eBay's network effects term of art meaning the more people who use it the more buyers will show up the more buyers there are the more tempting it is to sell there and something like Bitter's Edge would have been a way of breaking that cycle because you could go to Bitter's Edge and see toy trucks from lots of auction sites but that required scraping and as you say you could imagine then very salutary uses of scraping for which this is why law school is three years long to learn the doctrinal tools to be able to say well this scraping is good and this scraping is bad but at the moment as you say seems at all scraping except by academics trying to do the right thing seems to be disfavored is there any other stage setting you think is good to do about how before we get to the personalities and the people who are truly larger than life and for all I know Ho-yan is on the webcast right now I wouldn't be surprised if you weren't tuned in although he's probably seen your book talk elsewhere and so I'm curious is there any other stage setting we should do structurally including maybe what would you say was the apogee of the market and use of clear view which is to say was there a time when anybody could just sign up or a friend of a friend could just start using it at the dinner table to see who was around the table if they didn't quite know them and of course by identifying them you're also being able to see where that photo came from so you're knowing what kinds of websites they frequent and establish a presence on should tell you a lot more than just their name so I'm curious about that the shape of who has been able to access it over time and what the company's disposition has been to people who want to go to the individual trouble of saying please take me out of this yeah so clear view in its early days it really got started kind of around 2017 2018 and initially the founders will talk about but they built this tool that could find photos on the internet if you kind of fed a face into it and they didn't really know what to do with it they had a product in search of a customer and they were based in New York City so originally they thought let's sell it to New York businesses which is mainly real estate firms and hotels and grocery stores and so they were just kind of like going around trying to pitch the tool and they're also trying to raise money one of their first investors ended up being Peter Thiel he gave them $200,000 before they were clear view AI so they were a sort of a corporate checker should observe at this point Peter Thiel, Facebook board member fiduciary duty to Facebook as a board member and investing in the company that's scraping Facebook against the terms of service and got a letter from Facebook yeah so interesting though Peter Thiel has kind of a history of being companies that are making money off of data whether it's Facebook or a company that will then use Facebook's data and so they were trying to raise money and at that point they were kind of offering clear view AI pretty freely to people to just to demonstrate its power so they would give it to investors say hey, you know why don't you test this out and it was a fundraising mechanism and so a lot of investors went out using it as a party trick they would use it on dates I heard from one person who knew an investor and he said you know I invested this in part to make sure that my face isn't in here and the faces of my friends he was basically trying to protect his own privacy while investing in this thing that would intrude on others and they gave it to a New York billionaire who owned a grocery store chain because he was thinking about using it somehow to prevent shoplifters he didn't end up installing it in his stores but he had it on his phone his name is John Katz Matides he's run for mayor of New York before and so I was talking to him about it and I said well was it useful to you on your own phone and he says yeah one time I was having dinner at an Italian restaurant and my daughter walked in and she was with a man I didn't recognize clearly on a date and I wanted to know who this guy was and I asked my daughter to go take a photo of them and then I ran his face through Clearview AI and I found out he wasn't a charlatan he was a San Francisco venture capitalist I said are you sure about that see what you did there and so they were just kind of like handing it out to anybody they were just looking for somebody who wanted to pay for this or give them money to work on it more and when they were pitching it to a commercial real estate building the person who was betting it was their security director and he used to work for the NYPD and he said you know who would really love this my former colleagues at the police department can I introduce you and so he sent an email to them introducing them to the financial crimes department that he used to work with they had a meeting and the NYPD started using the app and they were like wow this is great and they were telling other officers about it and just started spreading through the NYPD and there's this whole whisper network in law enforcement that I kind of wasn't aware of before but where they talk about useful tools to them and so it just started going from the NYPD to other departments around the country that is how the FBI heard about it Department of Homeland Security heard about it and law enforcement officers around the world and all of a sudden Clearview went out these free trials all over the place and then eventually departments start signing up and one thing that I found really interesting when I first started looking at Clearview is that historically facial recognition vendors were charging a lot of money for algorithms like hundreds of thousands of dollars sometimes millions of dollars per year Clearview was offering it for about $2,000 a year per subscription and so it was cheap and it was powerful which I found out once I started investing in Clearview AI they were trying to keep all of this a secret they did not want to talk to a reporter they didn't return my calls I was knocking on doors it was kind of a fruitless exercise at first I think they were hoping I would just go away and I started talking to police officers and they said yes this is incredible it works like nothing I've used before one detective in Gainesville said I had this stack of unsolved financial fraudster crimes on my desk and I'd run it through the state facial recognition system and got nothing I ran it through Clearview AI and I'm getting hit after hit after hit this might be somebody walking into a western union office to receive stolen funds or something exactly somebody standing in a bank counter who got ill-begotten gains or somebody in an ATM machine and he said what was great about it is before I was restricted to looking through kind of Florida people and now I could look I was getting hits on people outside of the country hits in other states and it just worked better than anything I used before and I said wow that sounds really great I'd love to see what that looks like and he said well just send me a photo and I'll run you through it and I'll send you your Clearview results and then he stopped talking to me and he wouldn't return any of my phone calls and this happened with another officer he actually just used it in a sexual assault case this week somebody who had taken a photo of her attacker at the bar earlier in the night it's great it works really well as long as somebody's not an online ghost if they have like things on the web it'll come up and he offered to run my photo he ran my photo there were no results and he said that's really strange because I googled you and you have a lot of photos on the internet and I said yeah like I would expect to have results here and I looked at their servers down he stopped talking to me eventually I found out Clearview has put an alert on my face and every time an officer is running me they are being notified and they're reaching out to the officers and telling them not to talk to me and they blocked my face from having results and so this was pretty chilling to me it shows the extraordinary efforts you'll go to to protect your own privacy and so it shows me in control whether you can be found and then it just showed me the power of a technology like this they were just using it to track me but when you think about this technology out in the real world this is a way to track investigative journalists or government officials or your political opposition you can put an alert on someone's face and they have no idea unless they act on it in some way and so that is what I find so compelling about facial recognition technology and why I wanted to write a book about it is all of this tracking that we have all been dealing with for the last two, three decades in terms of the internet that we've built that can track you everywhere you go that can compile this information about you where there's cookies on your computers that say you're interested in law or you're a gambling whale or you have some kind of addiction all that can now be just attached to our face in the real world as we're moving through real environments where we kind of assume we have anonymity if I might I'd like to ask for a vox pop interlude where especially given what we're talking about even asking people to raise their hands and answer to a question feels possibly intrusive so we'll do it the way the internet engineering taskforce does I'm going to call for a hum if you agree at the count of three and then we'll see what kind of consensus is in the room so a few rapid fire questions from what you've heard or know so far about this tech I'm curious how much you think it is a problem on a scale of one to ten if you think it is a two or greater let us know with a hum one, two, three all right if you think it is a six or greater let me know one, two, three weirdly it got louder people had a chance to kind of think about the mechanism here how many people would put this at a ten one, two, three all right well it's a somewhat self-selected group that arrived I'm not sure if the webcast would agree a couple more questions now there were a bunch of fellow Americans present at the United States Capitol on January 6th 2021 they streamed out at the end of a shall we say busy day without any arrests being made because it was just enough to try to secure the place and there's been a years-long effort by law enforcement to identify the people who rioted at the Capitol that day let me know with a hum if you would favor the use of Clearview AI to identify what we will suppose are people who otherwise would not be identifiable who were clearly at the Capitol rioting that day fair enough one, two, three okay did you hear it was a disappointed hum but a hum nonetheless and last question unless you want to add any on I might have been your piece if not another who said that there was some sense that the Ukrainian armed forces had been using Clearview AI which is constantly looking for different markets to identify Russian soldiers who had been killed on Ukrainian territory and left unclaimed by the Russians for the purpose of getting their names through whatever Russian social media say had been scraped with their photos and to then be able to reach out to families to let them know that their family member had been killed fighting for Russia in Ukraine how many people think that is a salutary good use of Clearview AI one, two, three that was a thoughtful hum that was like I need a little more time I don't know, Kendra are you surprised at the reaction so far yeah I'm surprised that it sounds like the same people are humming for all of the differences well that's the magic of the hum but yeah one of the reasons I asked the way I did was I confess and this is, I already put my cards on the table as somebody who wrote the Washington Post op-ed wanting the company to be done with you in the January 6th example it's like well it's there why not use it and it was used on January 6th yeah I mean should that evidence be excluded at trial we I think doctrinally no big deal about the use of it for these purposes it's the kind of thing that does complicate the story the way you were saying the story of scraping is complicated I don't know if you have a quick intervention there may even be I don't know if they're turned on a microphone in front of you yeah although the people on the webcast probably can't but bellow and we'll repeat just yeah of course it's a biased way of asking the question sorry the question was given that I loaded into the question that people wouldn't be found any other way so I I think the reason I asked it that way was to give as much momentum as possible to an affirmative answer given what appeared to be a pretty strong consensus in the room that the thing shouldn't exist at all and it might mean then in you know more careful review that there might be plenty of other ways in which you can identify people for which it's like why are we doing proof of work where you have to spend more money to get to the same place also opens the question is the problem with this that it doesn't work that it's misidentifying folks and putting them in the crosshairs when it's not underwriter laboratories tested as a technology it's this weird proprietary thing or is the problem that it does work often when other things don't and at least some of your reporting seems to say there are plenty of instances at the moment where it works when virtually nothing else will yeah I mean facial recognition technology and it's part of the story of the book a long time it really did not work very well even as we were were actively deploying it and I find that very troubling it has gotten very powerful in recent years because of all of these advances in machine learning and pattern recognition the kind of neural net technology that's driving other kinds of advances why we're seeing magical seeming things like chat GPT on that question so yes on January 6 it was used you know the FBI essentially did screenshots from all of the social media that resulted from the intrusion the invasion of the the capital and put the photos out there and said help us identify these people and at that point FBI didn't seem to be using clear view but all of these local police departments were and so they would take the FBI photos and then run them through clear view and if they got a match that they felt was credible where you know they found other photos on social media of that person wearing the same outfits they would send it to the FBI as a tip as you said there are probably other ways to identify those people even you know eyewitnesses maybe the most passionate users I've talked to of clear view I have been child crime investigators who often will come across abusive images of children like on the dark web or in some kind of random account where they cannot figure out you know who that account belongs to or whether that's the person who originated the image and so clear view they say is really an unprecedented tool for them because so one of the cases that basically convinced the Department of Homeland Security to get a subscription was they had this image it was found in a Syrian user's yahoo account they could tell it was from the US because of the power outlet in the photo you could tell it was somewhere in America and that's all they knew and so they run the photo through clear view I sometimes sometimes this is helpful for identifying the victim clear view as one of the first I think at the time was the only database that these investigators had access to that had children's faces in it so you could possibly identify the child but for the abuser they found this photo of him in the background of someone else's Instagram photo and he was standing behind this this booth that linked him to a particular company based on Florida the investigator followed the breadcrumbs and eventually located this man in Las Vegas he was arrested he's in jail now you know the child was removed from the abusive situation so these are kind of scenarios in which clear view I might be the only tool that helps them solve the crime and so that is a powerful use case and this has been powerful for clear view I you know as I told you they weren't originally designing this tool for law enforcement but once it proved popular law enforcement it is the most effective argument that they have when people say hey this rat raises huge privacy issues you know none of us consented to be in your database why do you have the right to put us in there they say well we're just using it to help police we're only using it to help criminal crimes and this is I mean it's gotten them out of trouble several times now and just this month so there's a very different approach to privacy here and in Europe and in Europe you the privacy laws there say if you want to collect sensitive information from citizens like your biometrics you need to get their permission first so when I exposed clear view AI privacy regulators outside the US as illegal you can't do this kicked clear view AI out of their countries and some of them find clear view and so the UK privacy regulator the British privacy regulator was one of those people who issued a fine it was for around 9 million dollars and clear view appealed and they just won that appeal and the reason was jurisdictional the appeals court there the tribunal court said essentially that UK privacy regulator doesn't have jurisdiction over how foreign governments use data about their citizens so specifically because clear view is says they're only working with law enforcement allowed them to get out of this fine which is fascinating and the UK privacy regulator can still appeal that and they might but it really it raises this problem of how I'm like how do we regulate a technology like this that is so global that you have companies around the world that are doing this they're scraping information from all over the place it really starts to look like this international problem that we have to address so there's so many directions we could go right now so in the spirit of a choose your own adventure book is it best that we open the door and talk about some of the personalities and the people for which you devote a lot of time to that I know there's an interesting Trump-Maga world connection should we attempt a demo on the fly of PIMIs for which there's now some competition to clear view or should we take questions and you're welcome to try to multitask too I don't know what maybe one one news you can use in terms of regulating clear view there there are privacy laws here and we don't have anything basic at the federal level that addresses what clear views done but there are state privacy laws that are very relevant and several states now have passed the right to access information that a company holds on you and the right to delete it and so if any of you are from California, Connecticut, Colorado or Virginia you can go to clear view AI and remember domicile is intent to reside sure you can go there if you have any kind of address that links you to one of those places you can go to clear view AI and say I want to see my report and you can ask to be deleted from the database I don't think that opt out laws work that well I looked at California also requires companies to say how often they get these deletion requests and California has 34 million residents and in the last two years something like 700 of them have deleted themselves from clear views database about 20 are probably in the room it doesn't work that well but yeah should I try to do the demo just so you can see what this looks like yeah why don't we give it a quick shot and I mean technology demos never go wrong especially when they're not planned but let's take a look so one of this I mean good okay and there's a picture on the desktop of us okay let's see if we get and you're logged in already go to upload photo I'm going to go and do we need to share this screen with the zoom should we do that real quick so this is not clear view AI clear view AI is restricted to law enforcement use but as Jay Z was saying when he was kind of doing the introduction here what clear view did was was made possible by tools that are increasingly open source and off the shelf and so clear view kind of I say that they made a ethical breakthrough not a technological one they were willing to do what other companies like Google and Facebook hadn't been willing to do and now they've had copycat and so oh it's not that was a long distance shot of us let's try you also put on the one with your sunglasses should we try that and so now we do have these other companies that have done the same thing that clear view did this is a now there is a gate here you have to check these boxes or it's not going to do the scan so it's supposed to be only the search for your own face and you do have to check a box that says I'm only this is my face and I have permission to search it but I have a subscription that gives me 25 you can see daily searches available to me so I don't think I'm going to be searching my own face 25 times do we want a safe search and or a deep search safe search for sure we do not want we don't want to put porn up on the how about a deep search search you have to pay $300 for and I'm not sure oh wow all right great but yes safe search is important and so now so PIM eyes has kind of done what clear view did they have a smaller database but you can get a sense from the search how well it can work so now in this case I'm actually trying to get it not to take all the cookies and of course I don't know how many is the purposes and I understand so there we go search like this when you do a facial recognition search it's not usually it just gives you one result it will give you a bunch of results ranked in the confidence that it's me and so you see here it only found photos of me it didn't even have any doppelgangers these are all me it did when we uploaded that photo it analyzed my face came up with a biometric identifier for my face and then looked through it's database of I think they have about 2.8 billion faces in their database and then found these other photos of me with links to where they can be found you do have to have a subscription to click on the face and see where it can be found but but you can see it like it works pretty well and this is a company that is run by a guy who lives in the country of Georgia the site is headquartered in the UAE and their legal services are offered by somewhere in the Caribbean so Clearview AI has gotten a lot of attention PMIs hasn't gotten as much attention even though it's out there and like anyone can use it and you can pay you know $25 a month and or $30 a month and you know you could use it on people in this room right now and it's part of why I wrote this book because I am worried that this is just getting out there and we're not doing enough to choose you know the world that we want to live in we're letting the technology dictate it I did write about threats to children from AI including you know deep baking their voice or you know taking a photo of a kid on a playground and doing something like this to figure out who they are where they live after I after I published that article PMIs actually blocked searches of minor spaces so public scrutiny is kind of working with me but yeah it's a little bit of a wild west I'm worried that we are not doing enough about it to put it mildly yep and it's been interesting that you have in publishing both the book and the sequence of articles in its run up you've covered all the angles including the angles that may be ones that are challenging to a view that is salty earth such as the child abuse imagery identification stuff yeah I mean one story I tell in this book is about a guy who came to me knowing I'm interested in facial recognition technology and he wanted to confess how he was using this tool PMIs because he wanted policymakers to know, regulators to know because he thought that what he was doing was so wrong and that he shouldn't have access to a tool like this and what he described to me is essentially he had a porn addiction and a privacy kink where he would watch videos of women and then he would want to know usually because that work is so stigmatized people use pseudonyms he would search their faces and find out who they really were he described it you know finding one woman's high school photos on flicker from like her spring formal and he just liked to compile this this file of their vanilla life and he would just keep these files he says he didn't act on it he was just a peeping Tom but that eventually got sick of this and so he started going through his Facebook friends and he would search for them and see if they had any risque photos and he did find them, he would find them on revenge porn sites where you know their name wasn't there but their photos were and now they were findable because we are seeing the internet kind of reorganized around biometric information around our faces and I just think if we don't do anything about this you know we'll see that happen in other realms we'll have a voice search where you know you upload somebody talking and then you find everything they've ever said that's on the internet a privacy activist talked a lot about if we don't do something about Clearview AI we'll have startups that do this in the DNA space that they'll go to hairstylist and collect clippings from their customers sequence it and create this DNA database which is just a public or not a public DNA base their proprietary DNA database that they could sell to police or to companies whoever might be interested in that information I see people are kind of like absorbing all this just any quick thought on the personalities involved and maybe even the question I gather of war and peace which is how much of this history was as you hinted at earlier an inevitability this was just out there waiting for the first person in the pool to jump versus a somewhat zig and zag of particular personalities doing this and being in the right place at the right time yeah I mean to the extent that what Clearview AI did was ethical arbitrage I did want to with this book understand their motivations you know this is they're not your typical tech founders a very rag tag crew who has corresponded with Jay-Z apparently is you know he was kind of a tech obsessive just he grew up in Australia loved computers dropped out of college at 19, moved to San Francisco to kind of chase the tech dream was building Facebook quizzes and iPhone games in that Trump Hair app that we mentioned and when he first moved to San Francisco he was pretty kind of liberal like grew his hair long hung out with artists played guitar but then he moved to New York in 2015 and as he described it to me he said he was radicalized by the internet and he really fell in with the MAGA crowd and you know started retweeting Breitbart people and wrote a song about Milo Yiannopoulos and how he was unfairly banned from Twitter and as one does and went to the Republican National Convention with this guy Charles Johnson or Chuck Johnson who's kind of conservative provocateur and internet troll and their idea originally was it would be great to have an app on your smartphone that you could just point at someone else at the RNC and just get information about them and yeah I mean originally it wasn't just face recognition they are actually Charles Johnson as a big believer in physiognomy or technology the kind of this idea that you can derive from people's facial features who they really are how intelligent they are whether they have criminal tendencies so at one point Juan Tantat was talking about after the Ashley Madison hack that happened a few years ago that exposed all the users he said wow now we know all these people's names we can look them up on Facebook download their faces you know have AI crunch that and figure out what a cheater face looks like and so that was their original idea he later swore off those beliefs and they moved more in the direction of just pairing people's faces but yeah it really came out of this conservative kind of mindset and so I think one of the questions of the book is was it particular to these people that they came up with this tool that they were willing to kind of push the boundaries of what was considered acceptable at that point by Silicon Valley and they did struggle to raise money a lot of investors loved using clear view AI on their own phones but didn't want to put money into the company because they were either troubled by the background of the founders the other guy was once co-founder was Richard Schwartz who was like a New York Politico who worked with Rudy Giuliani when he was mayor he's much older it was a really strange pairing but they were either troubled by the backgrounds or just were worried about the legal risk like was this legal to do was this data ethically sourced it's our intention to go to 130 if that's okay is that I'm waiting for anybody in authority to tell me that's problematic nope I got thumbs up that's great there were a number of enthusiastic hands up and I was thinking it might be great to collect some comments seriatim before landing back on you and we can just kind of do a scan of the room as it were and we have the next here somebody up front Patrick feel free to tell us who you are or don't and then someone can look it up hi my name is Yana I have a background in biotech and genetics so I'm not a law student here I'm an alum of the college the DNA comments reminded me of something called environmental DNA which is where researchers who are doing ecological surveys trying to figure out what species are in a certain area have found that by taking a sample from the stream sand the air they can find human DNA recognizes as such and potentially even link it to people so that you'll need to be a hairdresser you'll need to work at a nail salon to get people's DNA you can just be at a place collected sequencing is becoming extremely cheap like prospecting with a metal detector at the beach but now it's just a human detector collecting in a stream might not be very effective but you could go on trash day and just pull a little trash out of every house and then very quickly know what DNA belongs to which house or which address I would love to know your perspective on this and also if you think something should have been done we should have done something a decade ago for facial recognition what can we do now for DNA great question Patrick do you want to route to another upraised hand two folks here I have a question over this side I see we're going to alternate that's fine earlier you mentioned the term like the police was searched and if they found something credible what does that mean and who determines credibility and what is the threat now that there's so much like deep fake technology out yeah great question about whether deep fakes could frustrate the quote descriptively effective use of this too just building off I the toothpaste is out of the tube and I feel like that's kind of a common argument with tech in general but then there's this fang meeting meeting of the minds early on saying we're not going to do it where's the jump in point for either political regulation or industry self-regulation or is it just that the toothpaste is out of the tube and technology is running and we all need to get on the board where's the line let's take two more is the mic over here somewhere as we're alternating I was at the airport a number of months ago about to travel abroad and I was in a very long line in NSA I noticed somewhere there was a sign and some technology that NSA you might maybe they're overlapping just wondering and it said we are testing out a new facial recognition technology or I don't even know if it's a facial recognition but and I just thought nobody was I'm sure there were a couple of people who were paying attention but I just felt helpless in a way because if there was nothing I mean I had to go through the line to travel and I just wonder I just wonder about that you know they're testing out a new technology we don't have a choice in the matter if we want to go to our destination and what is that technology and so I just started thinking a lot about that one more and then we'll hear from catch me again and we'll try to synthesize as we go back to the faculty associate at the Brooklyn line center and I was just wondering if you came across in your research anything around emotion AI and looking at a face and reading what the person is experiencing inside their companies like affectiva that have done things like look at the face but also like give people beer sports event to wear a fitness tracker so they can track their heart rate as well and sort of line up all that data and do research with it so I was wondering if there are concerns there like what should we be thinking about or doing about this in your view so questions about DNA questions about effectiveness and how to know what things like it's enough to proceed to further police investigation or even prosecution and then a number of questions about what now both retrospectively if the cat is out of the bag is there any way to put it back in and return to some status quo anti and also prospectively we don't expect that this will stand still what lies around the corner and should we be anticipating and regulating about that seem like a fair characterization I guess I'll I'll start with the credibility question so so you saw is a face recognition technology can work pretty well like there it just paired me to me but facial recognition technology is not perfect and you know sometimes it will pair you with a doppelganger and so every kind of police officer I talked to about facial recognition technology will say you know it is just a lead we would never arrest somebody based on a face match alone that said there has been a handful of mistaken arrest based on facial recognition technology six that I know of and one of those did involve clear view AI a guy named Randall Coran read who was arrested in Atlanta Georgia day after Thanksgiving driving to his mom's house gets pulled over pulled out of his car and put under arrest for larceny in Jefferson parish and he goes where is Jefferson parish they said it's in Louisiana he said I've never I've never been to Louisiana in my life and you know basically there been somebody buying purses with stolen credit cards there was surveillance camera footage they ran a match it there's a human being who looks at the results and they thought that Randall Coran read was the best match they looked him up on Facebook he had a lot of friends in New Orleans and based on that basically they sent a warrant for his arrest in to Georgia along with a request for extradition so he actually spent a week in jail waiting for the Louisiana authorities to come pick him up had to hire lawyers the lawyers are like why is he tied to this crime they're lucky enough to figure out its facial recognition technology and so they get photos of his face a video of his face send it to the authorities in Louisiana and they notice he has like a little mole that their suspect didn't have and so they dropped the case but that is you know that is crazy and so credibility is something else that ties this person to the crime beyond that they look like the person because just looking like somebody like no one should be arrested for the crime of just looking like someone else and so some but what is that what is the other thing and I think that's one of the questions I have when we talk about guardrails if we are going to use facial recognition and policing do we want to use a database like Clearview AI that's looking through 30 billion faces looking through probably all of us in this room to find a match to a shoplifter in New Orleans do we use it for shoplifting do we only use it for more serious crimes in what are the other things that you need to do a lot of the mistake and arrest have been because they do facial recognition technology get a match and then put that person into a lineup like a six they called a six pack where it's a photos of six different people and they show it to an eyewitness and they say do you see the criminal here and then they will often agree with the computer yeah this person looks the most like the person and then they arrest that person and so that's a very dangerous practice and yeah I do I do worry about a little bit about the facial recognition technology where we have deep fakes matched with facial recognition technology I can just imagine a way of smearing somebody's personality by seeding these like you know pornographic images of them on the internet and it comes out later in a face search and then creates a whole news cycle around that so I do think it's a concern on toothpaste is already out of the tube because the technology exists and is capable of doing this that we just have to accept it like that is why we have a world of laws and regulations and policies and we have had moments like this in our history I've been listening I've been reading The Listeners by Brian Hawkman which is about the age of wiretapping and when we first saw in the United States kind of the development of small listening devices and bugs and we were like oh my gosh we won't be able to have a private conversation anymore conversational privacy is dead the technology is there we can't resist it and then we pass laws that made it illegal to secretly listen to people and Jay Stanley at the ACLU says this is the reason why all of the surveillance cameras all over the United States are only recording our images and not also our conversations we can create zones of privacy but we do have to pass laws that force them and so I think one example of that is in the EU they say you can't put people into these databases without their consent or we can do it the way we like to do it here or we say you can get out of the database or you can have a law like Illinois passed one of the rare laws that was moved faster than the technology passed in 2008 Biometric Information Privacy Act and this kind of addresses the DNA question a little bit but we can't use somebody's biometric information without getting their consent or you face a very stiff financial penalty if they're just settled isn't it they there's two different versions of it the class actual one against I think I just saw I don't know I have to look it up but it's they've been saying it's going to settle for a while but they haven't at least last I checked hadn't publicized how much they're going to pay but yeah it says you can't use some of these biometrics without their consent my favorite example is Madison Square Garden which is started using not Clearview AI but facial recognition technology installed a few years ago to address security threats they installed it around the time of the Emmys so I assume it was kind of the Taylor Swift security thread of known stalkers that are trying to get into our concerts and using facial recognition to keep them out but there's been surveillance creep function creep the owner of Madison Square Garden said well this would be a really good way to get keep my enemies out and use this to ban lawyers who work at firms that have suits against MSG from personal injury slip and fall suits to big shareholder litigation who thought he spent way too much money building the sphere in Las Vegas and when those people tried to get into Madison Square Garden actually went with an attorney who was on the ban list you know as soon as you walk through the doors a security guard comes over and says you're not welcome here until your lawsuit is dropped and the person I went with wasn't even working on the lawsuit it was one of her colleagues she had her daughter with her is that oh I didn't go with her there's a girl Scouts troop the mom got turned away because she worked at a law firm she didn't work on the litigation but everybody at the firm was banned MSG also has a theater in Chicago and they can't use their technology that way they're not allowed to use lawyers biometrics without their consent and the way MSG did this is actually scrape the lawyers photos from their websites so yeah so I think laws work and we just would need to actually pass them I don't know how soon that's going to happen at the federal level here in the United States where we have a bit of gridlock there but the states have been passing relevant laws on DNA I think the same thing like we do have the one DNA law Gina passed actually around the same time as FIFA but it only regulates how employers use genetic information and insurance companies and I just do wonderful see an expansion of that particularly as we are seeing law enforcement now start to they're starting to access DNA that has been shared by genealogists people like investigating their family history and then putting it online to try to find ancestors and police have started basically going into those databases to try to match criminal suspects and sometimes victims and find out who they are well the cat is at least somewhat out of the bag but maybe best not to think dichotomously it's Schrodinger's cat there's still ways to try to keep at least the tail inside et cetera et cetera and what an interesting sort of call for further study or action not just the law students and lawyers among us but to everybody where there's just so many dimensions to this story and Catherine thank you so much for taking the time and effort in so many ways to limit it to tell the public about it and to cover it in such a nuanced and thoughtful way thank you so much for having me here we appreciate it thank you nobody seems to have barged in to make use of the room so if you have a couple minutes maybe we can just mingle a bit and we will turn off the webcast and the cameras