Announced they will be rolling out facial recognition cameras across the country. Chief Constable Jo Farrell says it would be an Abdication Of Her Duties not to be using it we were given a very good demonstration of it through the summer. Retrospective facial recognition tracked down many of those taking part in the riots. Even those who were masked but is it becoming too intrusive . Even those who were masked. But is it becoming too intrusive . The software that identifies ourfaces is now being developed for authentication, we use it on our phones instead of a code. British telecom are currently trialling that same technology to improve cybersecurity, so that only authorised workers have access to critical systems and data. There is a lot to consider. With me in the studio, as ever, the font of all Ai Knowledge priya lakhani, ceo at century tech. She has come back from New York this morning, so bear with her. Im going to put you on the spot nonetheless. Give us a quick explainer of how this technology works. Technology works. Firstly, lets say technology works. Firstly, lets say we technology works. Firstly, lets say we want technology works. Firstly, lets say we want to technology works. Firstly, lets say we want to spot. Technology works. Firstly, i lets say we want to spot you in an image, and then, go off a bit but we want to match that to a wanted list, kristin. Put an image of kristen here on the camera, in the background you will have him in the studio and there is lots of noise in that studio, tables, monitors. First we want to do what we call classification, detect the fact that we have got a face here. The way we do that is we use Deep Learning models to be able to classify the image and find the Faith. The way that they traditionally do that is look at lots of images, Tag Lots of faces in those images, say these are pieces and these are not faces then you would build up not faces then you would build up a Idp Models and you would train them to learn where the faces are. They would look at those Feature Sets of buffets, there are some models that use unSupervised Learning, where you have that sort of tagging and labelling. The model then learns what the specific features are of the face, the texture, the edges of what a face looks like. You have got a model that can potentially spot in that very big studio over there all of the faces. We then take your face specifically and we create, and if we can play a clip, a boundary box, a Bounding Box around your face. Ive got a clip to show our viewers, this is friends. You can see, it is very short, you can see, it is very short, you can see, it is very short, you can see the sort of boxes around the Faith. You create these Bounding Boxes around the face and the model knows that it has a face, but we want to detect and whether that is your face and it does that Match Something in the database. So we normalise the image. Say you have an image of a face, you normalise the image, you want to take away lots of variables that could create inaccuracies. The lighting, you might want to turn all the images into grayscale, look at the alignment of the Faith, once you have done that, and this is where it is interesting, in those images you sought lots of spots. The Deep Learning models will extract what it thinks are the key features of the Faith. Once it extracts those key features of christians face, and this is where it is really interesting, we turn it all into mats. We vectorised the data. You end up with Christian Fraser right now, what youre used to seeing in an image, actually is a string of numbers. Just a string of numbers. Just a string of numbers. Every image of you that taken on cameras out there, modelling that face in using the sort of ai, it will have slightly different string of numbers because your position might be different, your expression. But they will be very similar because it is yourface. A mathematical representation of your Faith. Then, lets say you have wanted list. You have got a database of faces that You Are looking for. They have been through a similar process where they have been encoded into numbers. And thenit been encoded into numbers. And then it is a matching exercise. Is there a similarity between the string of numbers that represent. 50 the string of numbers that represent. The string of numbers that represent. So that is where the computing represent. So that is where the Computing Power represent. So that is where the Computing Power comes| represent. So that is where l the Computing Power comes in and where the chips coming . Because. As the chips improve, those calculations will much quicker. Will much quicker. Yes, and remember. Will much quicker. Yes, and remember, i will much quicker. Yes, and remember, i really will much quicker. Yes, and remember, i really want. Will much quicker. Yes, and remember, i really want to | will much quicker. Yes, and i remember, i really want to get the expert in this area on, but do you remember when we did an episode on semiconductors . We talked about how they are two big processes that happen. All the Training Data initially, for example, all the faces and images, then we have come up we run the model, so that is what youre talking about. Youre running the model against that facial recognition to be able to find the Faith. It facial recognition to be able to find the Faith. To find the Faith. It is extremely to find the Faith. It is| extremely impressive to find the Faith. It is extremely impressive what Clearview Ai do. Before we talk to the expert, let me show you the Promo Video that Clearview Ai puts out. In 2019, Homeland Security investigations were trying to identify an Adult Male who was in a child Abuse Video. The Adult Male was abusing a six year old girl and selling this Abuse Video on the dark web. The only clue was a photo of the Adult Male who was in the background of the abuse. Video for just a few frames, with no other clues. The investigator and the case turned to Clearview Ai. Prior to searching on Clearview Ai we must provide a reason for the search. In this example, we will choose felony sex offence. Secondly, you upload a photo of the suspect from your Desktop And Press the search button. As you can see, 25 results now match the uploaded photo, whereas in 2019, during Homeland Securitys investigation, only one result came back from Clearview Ai. This is the photo. As you can see, the suspect is in the background of the photo. Press the Locate Button on the top left to zoom into the photo, or use the compare button to see them Side By side. In this case, the investigator clicked the link to a public social Media Post online, uncovering two key pieces of information. The photo was tagged in Las Vegas and the name of the company the suspect appeared to work for. With those two clues, the investigators at Homeland Security travelled to Las Vegas to obtain the suspect� s name from the employer and, with additional corroborating evidence, secured a Search Warrant for the suspect� s computers. The Search Warrant revealed that the suspect had thousands of videos and photos of Child Abuse Material on his computer. He pled guilty and is now doing 35 years injail, and the six year old girl was rescued. Lets speak to hoan ton that, whos Chief Executive Officer of Clearview Ai. How many cases do you think your technology has solved, and what was the big leap forward for you . What was the Big Leap Forward For Ou . ~ what was the Big Leap Forward For Ou . , ~ , what was the Big Leap Forward For Ou . , ~. , for you . Thank you so much for havin for you . Thank you so much for having me for you . Thank you so much for having me on. For you . Thank you so much for having me on, it for you . Thank you so much for having me on, it is for you . Thank you so much for having me on, it is great for you . Thank you so much for having me on, it is great to for you . Thank you so much for having me on, it is great to be l having me on, it is great to be here having me on, it is great to be here~ i having me on, it is great to be here. I appreciate your interest here. I appreciate your interest in Clearview Ai. We have interest in Clearview Ai. We have done it now over 2 Million searches have done it now over 2 Million searches on behalf of Law Enforcement. That is how many searches enforcement. That is how many searches they have used on our platform searches they have used on our platform. We do not know how many platform. We do not know how many crimes they have solved, but if many crimes they have solved, but if you many crimes they have solved, but if you take even a conservative estimate, you would conservative estimate, you would be in the hundreds of thousands at least. Sometimes Cases Thousands at least. Sometimes cases you thousands at least. Sometimes cases you have to search multiple images, so on. But anecdotally, as well, most recently anecdotally, as well, most recently we worked with the international centre of missing and exploited children, and we were and exploited children, and we were in and exploited children, and we were in ecuador, and latin america. Were in ecuador, and latin america, in three days these Law Enforcement agencies, about eight Law Enforcement agencies, about eight of Law Enforcement agencies, about eight of them, went through a list of eight of them, went through a list of the eight of them, went through a list of the hardest cold cases they list of the hardest cold cases they have not solved. Missing kids. They have not solved. Missing kids. Kids they have not solved. Missing kids, kids have been abused, and kids, kids have been abused, and they kids, kids have been abused, and they find them on these internei and they find them on these Internet Forums as victims. In those Internet Forums as victims. In those three days, they made hundred those three days, they made hundred ten identifications of missing hundred ten identifications of missing and exploited children, and rescued over 50 of them. The and rescued over 50 of them. The impact is incredible. On the the impact is incredible. On ihe flip the impact is incredible. On the flip side, we know it is a very the flip side, we know it is a very powerful technology, so we have very powerful technology, so we have limited the usage of our application to Law Enforcement and governments. But application to Law Enforcement and governments. And governments. But what is it. I and governments. But what is it i said and governments. But what is it. | said that and governments. But what is it. I said that there and governments. But what is it. I said that there had it. I said that there had been. With this technology, what has been the real breakthrough for facial recognition . I breakthrough for facial recognition . Breakthrough for facial recognition . I think it is Neural Networks, recognition . I think it is Neural Networks, which| recognition . I think it is i Neural Networks, which are recognition . I think it is Neural Networks, which are part of artificial Neural Networks, which are part of artificial intelligence. Previous algorithms for facial recognitions with, try and look at the recognitions with, try and look at the distance between the eyes, at the distance between the eyes, or at the distance between the eyes, orthat at the distance between the eyes, or that i send the eyebrows, orthe eyes, or that i send the eyebrows, or the nose in the eyes, eyebrows, or the nose in the eyes, Things like that. But that eyes, Things like that. But that does not work very well if you have that does not work very well if you have an image from a different you have an image from a different angle, say a Security Camera different angle, say a Security Camera. With Neural NetworksYou Are camera. With Neural NetworksYou Are able to train, it is called You Are able to train, it is called Supervised Learning, on a lot called Supervised Learning, on a lot of called Supervised Learning, on a lot of different examples of photos a lot of different examples of photos to improve accuracy. So, the weight photos to improve accuracy. So, the weight we train our algorithm was to get a lot of publicly algorithm was to get a lot of publicly available images. Say you have publicly available images. Say you have 100 images of george ciooney, you have 100 images of george clooney, 100 photos of Brad Pitt, clooney, 100 photos of Brad Pitt, the clooney, 100 photos of Brad Pitt, the algorithm will learn that pitt, the algorithm will learn that the pitt, the algorithm will learn that the black and white photo of brad that the black and white photo of Brad Pitt with the sunglasses on is the same one of him sunglasses on is the same one of him from 20 years ago with different of him from 20 years ago with different hairand so of him from 20 years ago with different hair and so on. The algorithm different hair and so on. The algorithm learns what stays the same algorithm learns what stays the same in algorithm learns what stays the same in A Algorithm learns what stays the same in a face. And so, the more same in a face. And so, the more data same in a face. And so, the more data you have, the more accurate more data you have, the more accurate it more data you have, the more accurate it gets. There has been accurate it gets. There has been some great research out there, been some great research out there, thanks to pushing the learning there, thanks to pushing the learning and all this Data Machine Learning and all this Data Machine Learning, the Research Community has done a really good community has done a really good Joh Community has done a really good job to improve it. We huiit good job to improve it. We built upon a lot of those what we were built upon a lot of those what we were able to do was bring a lot of we were able to do was bring a lot of data we were able to do was bring a lot of data to train our algorithm. And so, now when you look algorithm. And so, now when you took at algorithm. And so, now when you took at all algorithm. And so, now when you look at all the top facial recognition algorithms, not just recognition algorithms, not just Clearview Ai, there are others, just Clearview Ai, there are others, there is a national institute others, there is a National Institute