Grandma, machine learning and the internet of things
- Machine learning systems wouldn’t just make life easier for healthcare professionals or the elderly.
- Having a machine learning system that would help elderly people lead healthy, independent lifestyles in a supportive and appropriate way could change their lives.
- Machine learning systems would allow healthcare workers to use their time more efficiently, focusing on proactive and needs-based treatment.
- A leader in this field of context aware machine learning systems is Sentiance who has built an advanced machine learning system that can be used to monitor health and activity in our normal lives.
- Healthcare is a public good, and machine learning makes it possible to improve outcomes not just for our Grandmas, but for the aging population in general and the institutions, people and governments who support them.
What role does AI and machine learning have to play in a world with an ageing population? This article explores the issues.
@ipfconline1: Grandma, #AI, #machinelearning and #IoT via @graphcoreai
People are getting older. The aging populations of developed nations are growing rapidly, and it is projected that the combined senior and geriatric population in the world overall will reach 2 billion by 2050.
Many of us have elderly relatives, and we’d like them to remain self-sufficient and able to lead independent, fulfilling lives. Regardless of whether we’re separated by distance or live around the corner, it’s not possible to drop by as often as we might like, and we worry about them.
But a mobile phone, a few sensors and some help from machine learning can give Grandma a much better level of help and support – and improve outcomes for healthcare workers and broader society. It’s possible to make healthcare more personal and efficient for our aging population, giving them a higher quality of life and the ability (in many cases) to avoid institutional care.
Our smartphones have a range of sensors that allow us to track our daily activities. Adding a few additional sensors to monitor temperature, blood flow and heart rates gives us key data that can track our health. If we could get our Grandmas to carry their phones with them and keep them turned on, we could take regular measurements and log these to the cloud. We could also add a few internet-connected sensors in the home that serve a dual purpose of
It’s easy to collect this kind of data and log it to the cloud. The challenge is what we do with it. One option is to set up a private website that displays this information in a dashboard, and then monitor this from time to time to check that Grandma is doing well. But while this might be useful, it is a little intrusive and may not be the best solution.
Predictive, context-sensitive healthcare
Fortunately, there are other options. A few specialist machine learning companies have been developing systems that can learn and predict from smartphone and IoT sensors to build up a picture of where people are and what they are doing. A leader in this field of context aware machine learning systems is Sentiance who has built an advanced machine learning system that can be used to monitor health and activity in our normal lives. More importantly, the system can use sensor data to understand the context of a user: if they’re in their car, on a bike, hurrying for a bus or at home asleep.
Being context aware allows for more accurate predictions to be made from ‘regular’ health data. For example, we may find that Grandma’s temperature and heart rate are high: if we know that she went out for a bike ride, we could infer that the data is normal. We wouldn’t be able to make this inference without context. With contextual data it is possible to see situations as they emerge, and it should be possible for the system to know when to get a care worker involved so that preventative steps can be taken. These types of systems have huge potential to provide holistic healthcare that can really improve people’s lives, while respecting their privacy and supporting independence.
With up to two billion elderly people around the planet needing these services, machine learning hardware accelerators will be required. The key challenge in these context-aware machine learning systems is to not just identify features but also the connections between them, based on changes over time – what we call temporal machine learning models. Features might come from changes that happen quickly or changes that happen over a much longer time period. Current accelerators are not efficient at processing on these high dimensional temporal machine learning models and new solutions are needed.
Artificial intelligence, real independence
Having a machine learning system that would help elderly people lead healthy, independent lifestyles in a supportive and appropriate way could change their lives. But this system would need to be extremely intelligent so that it can advise and support in a way that is thoughtful and non-intrusive, respecting the privacy of individuals and only intervening when necessary.
And the elderly aren’t the only people who would benefit from a system like this: healthcare worker could use it to provide better outcomes for their patients. Machine learning systems would allow healthcare workers to use their time more efficiently, focusing on proactive and needs-based treatment. Enabling people to lead enjoyable, independent lives is much more rewarding than having to deal with chronic health problems that are destroying quality of life.
Lastly, machine learning systems wouldn’t just make life easier for healthcare professionals or the elderly. Health insurance companies, governments and taxpayers would benefit too. As the population ages, public health systems and institutions will bear the cost – but machine learning systems could improve and support healthcare, helping to avoid the rising cost of institutional care and adding value to the economy.
Healthcare is a public good, and machine learning makes it possible to improve outcomes not just for our Grandmas, but for the aging population in general and the institutions, people and governments who support them. Having the right data is the first step, but what we do with it – and how – is what matters most.