Surveillance cameras are considered to be digital eyes, a deterrent and a proof. At present, the biggest obstacle to video surveillance is low-resolution video, but the entry of artificial intelligence will bring unpredictable changes. Artificial intelligence monitoring starts from searchable video and will further promote the development of super-monitoring.

We usually think of surveillance cameras as digital eyes, monitoring us, or monitoring us, depending on your position. But in reality, they are more like portholes: they are only useful when someone looks through them. Sometimes this means that someone will watch the surveillance video, usually by monitoring multiple video windows at the same time. However, most surveillance cameras are passive. They act as a deterrent there, or provide evidence when problems arise.

However, this situation is changing video surveillance, and the speed of change is very fast. Artificial intelligence gives surveillance cameras the ability to match their eyes to the brain, allowing them to analyze real-time video without human intervention. For public safety, this may be good news, helping police and emergency responders find crimes and accidents more easily, and has a range of scientific and industrial applications. But this also raises serious problems for the future of privacy and brings new risks to social justice.

What happens if the government can use CCTV video surveillance to track a large number of people? If the police simply upload your face photos to the database, you can digitally track you throughout the city. What happens? Or the algorithm running on the camera in your local mall is biased. Just because you don’t like a group of teenagers, you will be alerted to the police. What happens?

Although the emergence of these scenarios will take time, we have already seen the initial results of combining monitoring with artificial intelligence. IC RealTIme is an example. The company’s flagship product launched in December last year was used by Google for CCTV video surveillance. This is an application and web platform called Ella that uses artificial intelligence to analyze the content in a video stream and make it available for instant search. Ella recognizes thousands of natural language queries, allowing users to search for content in the shots, find people who have specific animals, wear specific color clothing, and even clip images that contain a specific car brand or model.

In a web demo, IC RealTIme CEO Matt Sailor showed "The Verge" a version of Ella that connects to about 40 surveillance cameras that monitor an industrial park. He entered a variety of search content - "a man in red", "UPS truck", "police car" - all of which took out the relevant shots in a matter of seconds. He then narrows down and the time range and location range and points out how the user can slide up and down with the thumb to improve the results -- just like Netflix.

Sailor said: "If there is a robbery, you don't really know what happened." He said, "But then a Jeep Wrangler flew east. So we searched for 'Jeep Wrangler' and found it." On the screen, video clips appeared, showing different The Jeep Wrangler slipped past the camera. This will be the first major combination of artificial intelligence and CCTV video surveillance, Sailor explained: making it easy for you to find what you are looking for. He said: "Without this technology, you can know no more than your camera, and you have to filter content from hours, hours and hours of video."

Running on Google Cloud, Ella can search for footage from almost any CCTV video surveillance system. Sailor said: "It works well from a single camera system -- such as a babysitter camera or a dog camera -- to an enterprise system with thousands of cameras." Users pay a monthly fee, starting at For about $7 a month, the total price will increase based on the number of cameras.

IC RealTIme wants to target companies of all sizes, but the company also believes its technology can also appeal to individual consumers. These customers have been well served by the rapid development of the "smart" home security camera market, which is manufactured by Amazon, Logitech, Netgear and Google's Nest. But Sailor said that this technology is too simplistic compared to IC RealTIme's technology. These cameras are connected to the home Wi-Fi and provide real-time video streaming through the app, which automatically records the video as soon as they find something moving. However, Sailor said that they cannot distinguish between intruders and birds, resulting in a lot of false positives. He said: "They are very basic technologies that have existed for many years." He said: "There is no artificial intelligence and no deep learning."

This situation will not last long. While cloud-based analytics provided by IC Realtime can upgrade existing, fool-like cameras, others have added artificial intelligence directly to their hardware. Boulder AI is one such startup that uses its own independent artificial intelligence camera to market "vision as a service." One of the big advantages of integrating artificial intelligence into devices is that they don't require an internet connection to work. Boulder sells a wide range of industries and tailors machine vision systems to each customer.

The company's founder, Darren Odom, told The Verge: "The application is really comprehensive." He said: "Our platform is sold to banks, energy companies. We even have an app that can observe pizza, OK Are they the correct size and shape?"

“We are now able to identify Idaho salmon 100%.”

Odom cited an example of a customer who built a dam in Idaho. In order to comply with environmental regulations, they are monitoring the number of fish that can cross the top of the infrastructure. Odom said: "They had arranged for a person to sit in the window and look at the fish ladder and count how many squid swam." (As the name implies, the fish ladder refers to a stepped sink where fish can struggle through this channel. Upstream.) "Then they moved to video technology and someone (remote) monitored it." Finally, they contacted Boulder, who built a customized CCTV surveillance system to determine the fish upstream of the fish ladder. category. Odom proudly said: "We really use computer vision for fish species identification." Odom said: "We are now able to identify Idaho salmon 100%."

If IC Realtime represents the universal end of the market, then Boulder is demonstrating the capabilities of a boutique contractor. However, in both cases, these companies are currently only providing the tip of the iceberg. Just as machine learning is making rapid progress in the ability to recognize objects, its ability to analyze scenes, activities, and movements is expected to increase rapidly. Everything is in place, including basic research, computing power, and training data sets -- a key component in creating capable artificial intelligence. The two largest data sets for video analytics come from YouTube and Facebook, both of which have expressed the hope that artificial intelligence will help them control the content on the platform (although both companies admit that they are not ready yet). For example, YouTube's dataset contains more than 450,000 hours of tagged video, hoping to stimulate "innovation and advancement in video understanding." The breadth of organizations involved in building such data sets has given some insight into the importance of this area. Google, the Massachusetts Institute of Technology (MIT), IBM, and DeepMind all joined in and started their own similar projects.

IC Realtime is already developing advanced tools such as face recognition. After that, it wants to be able to analyze what is happening on the screen. Sailor said he has talked to potential customers in the education industry, and the other side hopes that when students are in trouble at school, the monitoring can be identified. He said: "For example, they are interested in the quick notice of the fight." All the system needs to do is pay attention to the students who gather together, and then remind someone so that he can check the video content to see what happened or in person. Go to the investigation.

Boulder is also exploring this advanced analysis. The goal of a prototype system being developed by the company is to analyze the behavior of people in the bank. Odom said: "We specifically look for bad guys and explore the difference between the behavior of a normal person and the behavior of people crossing the border." To do this, they are using the video of the old security camera to train their system. To find abnormal behavior. But this video has a lot of low quality, so they will also find some actors to shoot their training video clips. Odom did not elaborate on the details, but said the system would look for specific facial expressions and behaviors. He said: "Our actors will do something like crouching, shoving and turning back."

For experts in monitoring and artificial intelligence, the introduction of these features is full of technical and ethical potential difficulties. Moreover, as is often the case with artificial intelligence, the difficulties of these two categories are intertwined. Machines don't understand the world like humans. This is a technical issue, but when we assume that they can do this and let them make decisions for us, this becomes a moral issue.

Carnegie. Alex Hauptmann, a professor at Mellon University, specializes in computer analysis. He said that although artificial intelligence has made great progress in this field in recent years, there are still fundamental problems in letting computers understand video. The biggest one is the camera problem, which we no longer think of often: resolution.

For example, a neural network is trained to analyze human behavior in video. This work is done by subdividing the body into parts - arms, legs, shoulders, heads, etc. - and then observing the changes in these parts from one frame to another in the video. In this way, artificial intelligence can tell you if someone is running or combing hair. Hauptmann said to The Verge: "But it depends on the resolution of the video you have." Hauptmann said: "If I use a camera at the end of the parking lot, if I can tell if anyone has opened the door, Even if you are very lucky, if you are standing in front of the camera, you can track the movements of each of your fingers."

This is a big problem for CCTV surveillance systems, where the camera tends to be grainy and the angles are often weird. Hauptmann cited an example of a convenience store camera that monitors the cash register, but it also monitors the windows facing the street. If a robbery happens outside and a part of the camera's lens is blocked, the artificial intelligence may get stuck. He said: "But as human beings, we can imagine what is happening and put them together. But the computer can't do this."

Similarly, while artificial intelligence is a good way to identify relevant events in a video (for example, someone is brushing their teeth, watching a cell phone, or playing football), it still does not extract important causal relationships. Take the neural network that analyzes human behavior as an example. It might see the lens and say "this person is running", but it can't tell you why they are running because they are about to catch up with the bus, or because they stole someone's cell phone.

These questions about accuracy should allow us to seriously consider the declarations of some artificial intelligence startups. We are still far from approaching a point where computers can gain the same insights as humans by watching videos. (The researcher may tell you that it is too difficult to do this because it is basically synonymous with "solving" intellectual problems.) But things are growing very fast.

Hauptmann said that using the license plate tracking function to track the vehicle is “a practical problem that has been solved”, and the same is true for face recognition in controlled settings. (Using low-quality CCTV surveillance video for face recognition is quite another matter.) The identification of items such as cars and clothing is also very reliable, and it is achievable to automatically track a person between multiple cameras, but only if The conditions are correct. Hauptmann said: "It may be very good to track a person in a non-crowded scene, but in a crowded scene, forget it." He said that if this person is wearing an inconspicuous costume, This is especially difficult.

Some artificial intelligence monitoring tasks have been solved; others need to continue to work hard

But even these very basic tools can produce very powerful results. In Moscow, for example, a similar infrastructure is being assembled to insert facial recognition software into a centralized system with more than 100,000 high-resolution cameras covering more than 90% of the city's apartment entrances.

In this case, there may be a virtuous circle, and as the software gets better, the system collects more data to help the software get better. Hauptmann said: "I think all of this will improve." He said: "This is happening."

If these systems are already working, then we have problems like algorithm bias. This is not a hypothetical challenge. Studies have shown that machine learning systems absorb racial discrimination and gender discrimination in societies that write programs for them - from image recognition software that always places women in the kitchen to criminal justice systems that always say that blacks are more likely to commit crimes again, Bibi All are. If we use old video clips to train artificial intelligence monitoring systems, such as video from CCTV surveillance cameras or police-held cameras, the prejudice that exists in society is likely to continue.

Meredith Whittaker, co-director of New York University's (NYU) ethical "AI Now" research institute, said the process has emerged in law enforcement and will be extended to the private sector. Whittaker cited the example of Axon (formerly known as Taser), which acquired several artificial intelligence companies to help integrate video analytics into their products. Whittaker said: "The data they got came from the camera worn by the police. These data tell us a lot about who a single police officer will pay attention to, but it doesn't give us a complete description." She said: "This is a real The danger, we are generalizing pictures of prejudiced crimes and criminals."

Jay Stanley, senior policy analyst at ACLU, said that even if we can resolve prejudice in these automated systems, it will not make them benign. He said that turning a CCTV video surveillance camera from a passive observer to an active observer could have a huge negative impact on civil society.

“We want people not only to have freedom, but to feel freedom.”

Stanley said: "We want people not only to have freedom, but also to feel freedom. This means they don't have to worry about how unknown and invisible viewers will explain or misinterpret each of their actions and words." Stanley said: "To The worry is that people will constantly monitor themselves, worrying that everything they do will be misinterpreted and have negative consequences for their lives."

Stanley also said that false alarms from inaccurate artificial intelligence monitoring could also lead to more dangerous confrontations between law enforcement and the public. For example, think about Daniel Shaver's shooting incident. After seeing Shaver with a gun, a policeman was called to a hotel room in Texas. Sergeant Charles Langley shot and killed Shaver when he was kneeling on the ground at his request. The gun that Shaver was found to hold was a pellet gun, which he used to work on his pest control.

If a person can make such a mistake, what chance does the computer have? Moreover, even if the monitoring system becomes partially automated, will such errors become more common or less? Stanley said: "If technology is there, there will be some police officers who have to look after it."

Whittaker said that what we see in this area is only part of the megatrend of artificial intelligence. In this trend, we use these relatively crude tools to try to classify people according to their image. She cited a controversial study published last year as a similar example of a claim to identify sexual orientation through facial recognition. The accuracy of the results given by artificial intelligence is questionable, but critics point out that it doesn't matter whether it works or not; what matters is whether people believe it is useful and whether it still uses data to make judgments.

Whittaker said: "What makes me uneasy is that many of these systems are being injected into our core infrastructure, and there is no democratic process for us to ask about effectiveness, nor to inform everyone that they will be deployed." Whittaker said: "This is just another new example that is emerging: algorithmic systems provide classification based on pattern recognition and determine individual types, but these patterns are extracted from data that contain cultural and historical biases. ."

When we asked IC Realtime about how artificial intelligence monitoring might be abused, they gave a common answer in the tech industry: these technologies are value neutral, but how they are used and by whom they are used. They are good or bad. Sailor said: "Any new technology is in danger of falling into the hands of unscrupulous people." Sailor said: "Any technology is like this... and I think that on this issue, the benefits far outweigh the disadvantages."

Asic Antminer Machine

Asic Antminer Machine:Bitmain Antminer E9 (2.4Gh),Bitmain Antminer E3 (190Mh),Bitmain Antminer G2

The latest ETC Ethash Miner of asic antminer machine is Bitmain Antminer E9,it has 2.4gh/s hashrate,It's very profitable

ANTMINER is the world's leading digital currency mining machine manufacturer. Its brand ANTMINER has maintained a long-term technological and market dominance in the industry, with customers covering more than 100 countries and regions. The company has subsidiaries in China, the United States, Singapore, Malaysia, Kazakhstan and other places.

Bitmain Antminer has a unique computing power efficiency ratio technology to provide the global blockchain network with outstanding computing power infrastructure and solutions. Since its establishment in 2013, ANTMINER BTC mining machine single computing power has increased by three orders of magnitude, while computing power efficiency ratio has decreased by two orders of magnitude. Bitmain's vision is to make the digital world a better place for mankind.

Asic Antminer Machine,E9 Etc Miner,E9 Eth Miner,Antminer E9 Eth Miner,antminer ETC Miner

Shenzhen YLHM Technology Co., Ltd. , https://www.hkcryptominer.com