Building AI course project.
Note this is merly an idea for a possibly future implementation.
Virtual Pal (VP) is a software application for assisting people at home. The application will communicate with you and learn your habits and needs. VP will assist you with information you need, reminds you of things and check if you are okay when your life pattern is unnormal.
The trend seems to be that more people live alone. Too many also lack close friends coming to visit them. Those living alone may feel insequre at home because of their health condition or anxiety for being victim of crime. Accidents like when a person fall or faint could end up being stuck on the floor for hours or days without any possibility to contact rescue service.
A monitoring system with videolink and regular visits from a security company may be too expensive and violate peoples integrity. Instead a monitoring system with self learning technology working round the clock where the data is kept locally at the user for integrity would be an alternativ.
VP is supposed to reside on some device at the users home. Cameras with motion detection, tracking and face recognition covering the home area and various IoT devices available are the main input for the system to learn the users life patterns. VP will ask, by voice messages, the user for basic information at first and periodically ask the user if the user routins are normal. If VP detects anomalities in the user routins it will ask the user for confirmation or otherwise contact help, like rescue service. The system should also be able to detect intruders or fire. VP can remind about eg. take medicine and the user can ask VP to search for information or connect with user contacts.
Datasources are images from video cameras, IoT devices and voice interaction technology. The user data and images is the property of the user. AI methods are Neural networks and Deep learning.
The idea in this project is challanging. Skilled developers from different areas of technonlogies are needed.
The model is supposed to train continuosly on live data and from the answers it get from the user by asking questions to the user. The big challange is to tune the model to detect true anomalities from the complex data. This is crusial for the user to build a trust in this technology.
Another challange will be to balance sufficiently powerful devices for the system against the costs.
Hopefully this project will generate new ideas in organisations working in this field of technologies.
-
Thanks to Reaktor at University Of Helsinki for this interesting and inspiring course.
-
Thanks to the movie 2001: A Space Odyssey by Stanley Kubric for inspiration (even though HAL did not work as intended...)