Rekognition 

For some of us this topic can be really new but some may have already heard about it. Face recognition. Unfortunately the face recognition that we are going to talk about today is not the one used only by individuals. It is actually the opposite. It is used against them. Recently the world learned that FBI is using “Rekognition” for their own use with the help of Amazon’s big data base of faces. But for now lets get back to the actual topic of face recognition and other possibilities that Rekognition is giving us.

“Rekognition” is not only face recognition, it is object and scene detection, facial analysis, face comparison, celebrity recognition, image moderation and much more. To be honest this system already sounds like something that can be useful for the police.

Object, scene and activity detection 

Object detection algorithm detects objects such as people, animals, cars and things like backpacks, glasses and much more. Object detection is being used for example in autonomous cars where all rectangles are detected as pedestrians. In rekognition detection system is much more complex. When you want to detect objects on the photo, the system will give you answers with percentage of compatibility to the given object. For example :

Facial analysis 

This tool is mostly used in surveillance systems. It can detect things such as gender, emotions, if you are wearing glasses or not, if your eyes are open and if you are smiling or not. It can work not only with photos but also with videos and create a timeline with changes in individual’s behaviour. Facial analysis is much more complex then facial recognition as it gives you extra data.

Real-time analysis of streaming video

This feature is probably mostly used by police departments as it is able to detect, analyse and recognise faces. It not only do that but also is able to detect inappropriate behaviour and detect certain objects such as backpacks or bags in the live video. This feature is for sure step forward when it comes to preventing terrorist attacks but lets not forget that while police is going through millions of footage from surveillance systems, they are also detecting faces of individuals not related with those crimes. This topic is starting to be a concern because those police departments (Orlando and Oregon’s Washington County) that are currently using Rekognition are paying between 6$ to 12$ a month for this software. The police is able to get into every city camera in town and detect faces. Lets think about the data that they are gathering through this process. Some may say that those actions represent invasion of privacy and public is really concern but on the other hand what if this is the future of our world ?

Resources :

https://aws.amazon.com/rekognition/video-features/

https://www.theverge.com/2018/5/22/17379968/amazon-rekognition-facial-recognition-surveillance-aclu

Diabetes.

We have all heard about it and probably briefly know what it is. Maybe someone in our family is affected by it or you have it. This disease is spreading in unimaginable way and we can not stop it. But what if the is a way to make diseased lives easier and not worry about checking the insulin level all the time.

It has been brought up at the Cleveland Clinic Innovations list as Number 1 medical innovation of 2018. Artificial pancreas is a system of devices that mimics the behaviour of normal pancreas in order to deliver specific hormones that are needed for the patient with diabetes at the time. Patients that this devices has been created for are type 1 diabetes patients.

The pancreas contains of a glucose monitoring system (CGM) and an insulin infusion pump. An algorithm connects the two devices to deliver continuous communication in order to monitor glucose level and adjust needed level of insulin. The pump is placed on the abdomen and measures glucose many times and does not need to be checked among the day, not like the normal finger prick, which is less efficient then the pump but also way more complicated and time-consuming.

“Patients still need to be aware of what their blood sugars are and what the system is doing, but we are hoping that more work and research in the field will keep moving toward wider use.”

A CGM is specifically responsible for measuring the level of insulin and then delivering the data to the controller which contains of control algorithm that performs a series of mathematic calculation in order to send information about dosing the insulin to the insulin pump.

Why only patients with type 1 diabetes?

First of all, unfortunately this device can only help minority of people living with diabetes. Approximately 5% of diseased have type 1 diabetes. What is different with type 1 and type 2 diabetes is that with type 1 diabetes, the body is not producing insulin at all. Every carbohydrate that person with type 1 diabetes eat is being broken down into blood sugar (glucose), not into insulin which is a hormone that is being produced from the glucose.

When it comes to type 2 diabetes which is the most common one, patients can treat it with pills, insulin pump or even with a good diet and exercise, so there is no use for the artificial pancreas since their body is not making enough glucose, not insulin.

Sources:

https://www.fda.gov/medicaldevices/productsandmedicalprocedures/homehealthandconsumer/consumerproducts/artificialpancreas/ucm259548.htm

https://www.healthline.com/health-news/tech-innovations-healthcare-2018#8

http://www.diabetes.org

Tanmay Bakshi, 14 years old self-programmer who recently partnered with IBM Watson, came up with the idea of combining deep learning technology with neural networks and using it to save lives. Specifically lives of those who are crying out for help. 

Neural network

Suicide and AI? Why he wanted to do make something like that? Simply because second cause of death among teenagers is suicide. These days, when all we see is perfect lives, perfect bodies, perfect relationships, suicide and mental health is a big problem for everyone, not only teenagers. But still, lets not forget about the fact that usually teenager will not be looking for a help of others.

That is why Tanmay wanted to help his peers and worked for over 3 years with his team to create an app that will track and analyse your browser history, social media accounts, exercise level, sleep, public interaction and look for changes in your lifestyle, trying to find patterns of abnormal behaviour. He used deep learning technology that allows computers to act in sort of natural way and artificial neural networks that is inspirited by our behaviour, not mimicking it. ANN in particular fits this job perfectly because it can detect patterns of irregular behaviour better then our own brain because of the ability mentioned earlier.

When the app caches the irregularity, it immediately warns parents. Then it helps you to find solutions such as online meditation or gives you advices about good clinic in your neighbourhood.

Even tho Tanmay has been working on this app for over 3 years, the app itself is not available yet and the data of its release has not been mentioned anywhere.

Sources:

Neural Networks vs Deep Learning – Useful Comparisons To Learn

The Trolley Problem. 

Loved by ones and hated by the others. Questions with no right and wrong answers. Decisions that no one would like to make. What if there is a solution to the trolley problem and our hurting morals in the world of Artificial Intelligence and Self-driving cars.

“Moral Machine”. Creation of students from MIT, who wants you to be god for a few minutes and decide who lives and who dies. This platform is a simulation of a self-driving car seconds before the unavoidable accident.There is also a choice of creating your own design. In every scenario there is someone dying. The question is who and how to decide on that. .

Even though the trolley problem may seem old and not realistic, the driverless car for sure do seems realistic enough. With the technology that we established, we are able to make a car that is going to decide for us on what to do in every situation possible. The biggest advantage of it is that the car is actually able to process the situation and make the best decision, when with people we can not predict anything. But what about the situation when the driverless car is faced with unsolvable problem and need to chose who lives and who dies. How can we overcome those ethical issues that seems to be a no-win situation. Those issue has to be solved so that self-driving cars can enter the market but for now we are facing not only problems of morality but also problems with ethnicity and people’s beliefs. It turns out that people from different countries have different reactions to the situations presented in the “Moral Machine”. For example, someone will save a pedestrian who is crossing the street on a red light and someone else will think of it as an exception to the basic moral rule and not save this person. But when we think about an accident, it is a split second before we hit someone. No-one is thinking about the fact if the person crossing the street on the red light is poor, rich, sick or healthy. It is still human life, no matter what.No one will judge you for intentionally killing a rich person instead of poor one. Unfortunately when it comes to driverless-cars, those decisions are being proceeded differently and we need to face them. The only question is how and who is going to set the rules.

http://moralmachine.mit.edu

References:

https://www.theguardian.com/commentisfree/2018/nov/14/cars-drivers-ethical-dilemmas-machines

http://moralmachine.mit.edu

https://whatis.techtarget.com/definition/Moral-Machine

https://www.nature.com/articles/d41586-018-07135-0