Helping the elderly with Machine Learning

One of the most common issues I have seen the elderly have is in communicating. While it is not an obvious issue in small towns in America, in cities with a large immigrant populations such as NYC, Chicago, LA and San Francisco there exists a large subset of the population that cannot communicate in English. Many times, these elderly are the victims of scams that take advantage of their lack of comprehension of the English Language. A two-way translating mechanism would be a great resource to allow these individuals to translate in real time and be able to effectively communicate and examine situations rather than being in fear due to their lack of command of the English language. Deep neural networks are already at work as translators and there have been substantial results[1]. Guvenir demonstrates that English to Turkish translation can be learned and this can be extrapolated to any language. A real-time network would have to be trained. But once it is trained for a language, the machine can be consistently updated with more training examples. With the advances in computing power, something similar to Skype Translator[2] could be used. The Program already uses machine learning so it is not some far out there idea, it is just something that needs to be done better.

Another application is self-driving cars. Google already has the roads in San Francisco flooded with their cars, it is only a matter of time before these cars will be used to provide Emergency and non-emergency transport to those who cannot drive due to a disability. Self-driving cars already use machine learning(DNNs), and all that is needed is for these cars to be safe and reliable. In 20 years it is likely this technology will be sufficiently advanced to be a viable option. With further development in this area, it likely that machines can be trained to respond to anomalies and recognize needs before they occur. There is already research into many different solutions to this such as SVMs and autoencoders.

More ambitious ideas could involve using “nurse-bots” in hospitals to take care of patients and be able to recognize very basic needs and be able to tend to those. This would include bringing a blanket when a patient says they are cold. However, the skeptic in me sees this as something that will likely be done poorly because it will cut nursing staff and since machines will have trouble understanding non-trivial requests it will likely lead to poorer service. If done correctly, hospitals could use robots to perform trivial tasks and VERY preliminary diagnosis without reducing needed staff.

Unfortunately, the biggest issue the elderly face in America is not one that Machine Learning can solve. The elderly face incredibly high levels of loneliness and not robot can substitute human empathy. I believe Machine Learning can become a powerful tool to help, but it cannot provide the necessity of love.

Sources:

  1. http://www.cs.bilkent.edu.tr/~guvenir/publications/IS_23_6_353.pdf

2. http://www.techradar.com/news/software/skype-translator-starts-breaking-down-language-barriers-1277057