Keeping Buildings Cool, Terrible Drivers training Autonomous Vehicles and the end for text passwords
March 30
This week we discover a new way to keep buildings cool. We examine how terrible drivers can train the AI used in Autonomous Vehicles more quickly. We revisit an AI that is able to monitor our health as we age in our homes and we investigate the possible end for text based passwords.
Keeping Buildings Cool
A team of researchers at Cambridge University have developed a plant based film that gets cooler when exposed to sunlight. The film can be produced in a variety of textures and bright iridescent colors.
For a material to remain cooler than the air around them during the day the material needs to reflect a lot of solar light. If the material absorbs the light and only reflects some wavelengths the energy from the light is turned into heat. There are few materials that have the ability to reflect such an amount of light.
Passive daytime radiative cooling (PDRC) is the ability of a surface to emit its own heat into the air round it without the heat being absorbed by the air or atmosphere. This ability can allow the surface to become several degrees cooler than the surrounding air without requiring any type of electrical based cooling.
There are some paints in development that can achieve PDRC however they all have white or mirrored finishes. The team was able to derive a cellulose nanocrystal from plants that is able to be made into iridescent colors without added pigments. If pigments are added to a material some wavelengths of light are absorbed and only the color of the pigment to be reflected. This absorption of some wavelengths causes heat.
The team was able to color the cellulose based film by using phenomenon called structural color. A soap bubble shows a prism of colors when it interacts with light due to the differing thickness of the different parts of the bubble. This is structural color. The team was able to use the structural color of the plant cellulose to create different colors of film (structural color is why plants are different colors in nature).
The team created layered colorful cellulose nanocrystals on a white colored material made from ethyl cellulose and produced colorful bi-layered PDRC film. They made films with vibrant red, blue and green that were cooler than the surrounding air.
The hope is that someday this material will keep buildings, cars and other structures cool without requiring external power.
Terrible Drivers help to develop Self Driving Cars
One of the issues with training the AI that drives autonomous vehicles is that they need to be able to deal with the rare edge cases when driving. Because these types of problems are rare in the real world, millions of miles of driving and testing gives very little training in rare incidents.
A team from the University of Michigan believes that their artificial intelligence can reduce the testing miles required for autonomous vehicles by 99.99%. They do this by simulating a driving environment full of terrible drivers undertaking the types of maneuvers rarely seen on the road. This concentration of rare events forces the AI to make decisions over and over which allows for better training of the AI.
The automated vehicles are real but the mixed reality testing environment uses background vehicles that are virtual. The team can simulate drivers running red lights, driving the wrong way on one way roads, parking really badly so as to cause additional obstacles, skidding around corners, tyres rolling down the street and other rare but dangerous driving situations.
The Waymo (Google’s self driving car division) test vehicles have driven 20 million miles however far more data is required, possibly up to 25 times more data. Not much is learnt from driving the same safe route every day. However by using rare safety critical events to train the AI the training time can be reduced significantly.
The End for Text based Passwords
There are a number of systems being developed that pro port to enable us to get rid of text based passwords. Many of us use biometric identification with our phones. Zero knowledge proofs will soon provide an alternative identification system (more on that another day but a ZKproof is a cryptographic way for your computer to prove that it is not lying without giving away any information about you).
A team at the University of Surrey has now developed an image based authentication system called Transparent Image Moving (TIM). It has been designed to be used on mobile phones to prevent shoulder surfing attacks. Shoulder surfing attacks are more common in crowded public places where sensitive data is entered manually by the user. TIM requires users to select and move predefined images to a designated position for passing authentication checks.
A proof of concept study found that users were comfortable that the system would prevent shoulder surfing and password guessing. TIM is more useful than other image based systems on the market. One day we may all be trusting TIM with our mobile phone security.
At Home AI to measure Health
We have spoken previously about the use of AI and movement monitoring devices in nursing homes to monitor patients without disturbing them. These systems would detect movement and falls and notify staff immediately of a problem. The next generation of this type of system is more comprehensive and able to be used in home.
A team at the University of Waterloo has further developed AI to unobtrusively monitor people in their living spaces and provide early detection of emerging health problems. The new system will follow an individual’s activities accurately and continuously. It gathers vital information without the need for a wearable device and alters medical experts in the need of help.
The AI uses very low power, millimeter wave radio systems to conduct the monitoring. The system can detect falls in addition to monitoring factors such as change in gait (a significant early detection indicator for many issues). A wireless transmitter mounted on a wall, sends low-power waveforms across an interior space. As the wave bounces off different objects and the people being monitored they are captured and processed by a receiver. The AI engine deciphers the waves for detection and monitoring.
Paying it Forward
If you have a start-up or know of a start-up that has a product ready for market please let me know. I would be happy to have a look and feature the startup in this newsletter. Also if any startups need introductions please get in touch and I will help where I can.
If you have any questions or comments please comment below.
I would also appreciate it if you could forward this newsletter to anyone that you think might be interested.
Till next week.