These articles were written whilst employed with NetworthPick, the website and social media have been deleted, so I have preserved my articles here.
Machine Learning (ML) seems to be everywhere these days. Weekly we are seeing interesting applications of it shared wildly online, from everyday applications like Netflix learning about your taste in films to suggest what to watch next or Siri using ML to improve its understanding of your voice to better assist you better. Looking further afield we are witnessing the global boom of technology startups harnessing the power of ML to grow their business.
Machine Learning is a subset of Artificial Intelligence focusing on optimising algorithms to develop efficient and accurate solutions. These applications can be applied to nearly every industry. This is achieved by building a mathematical model of the problem and getting an algorithm to attempt to correctly perform the outcome, learning more about the problem from its feedback of being correct or incorrect. This can be as simple as training an algorithm to detect dog breeds by testing itself guessing against a dataset of pictures of dogs; over time it will get more accurate to the point it can be reaching its conclusions with accuracy over 90%.
Many sectors have utilised the technology behind ML to expand into new areas, the most striking of such has been the Medical Industry. Whether it be through the rapid advancements in digital imaging techniques for diagnosis purposes, allowing for the early detection of cancer or other complications from scans, pushing the boundaries of what can be achieved, or the post-pandemic boom industry of Virtual Personal Trainers (VPT’s), working to improve general fitness and health within the population. These advancements are taking us a step towards a cooperative future between doctors and machines.
Looking in more detail now at the business applications for the Personal Training industry, there has been a small growth in the use of virtual personal trainers. These work by using the camera on a phone or computer to guide people around tailored workouts at home and give real-time feedback at the convenience of the user. This would allow companies to create and market their own personalised training regimes through a program like this to reach a much larger audience for their services.
Chatbots are also another similar technology that has seen a similar rise to ML in the business sphere, and most websites these days have some form of chatbot assistant to help you navigate through their website and answer some questions. A chatbot is an AI that you can message and it replies by imitating a human interaction. They are used generally for many different purposes, one of them being for the first line of technical support by having the Chatbot be trained on key issues to help handle basic issues faced by users of the system.
The benefit of this type of technology is twofold. Firstly, it opens the business up to be more connected to their potential clients to increase lead conversion. The second main benefit from this technology is the cost-saving measures; not only does it free up the time of any receptionist staff you may have from smaller menial questions ringing in, but it also allows those questions to be answered 24/7, offering a benefit to your potential clients at a relatively small cost.
The medical industry has seen a growth in both the use of chatbots and the use of ML features to grow the chatbot to be able to deliver more niche services. An example of one way this technology has been used in this way is through diagnosis programs, whilst serious medical issues can only be diagnosed by a doctor, an app can help people get advice on cuts and bruises, rashes, and other small illnesses to gauge if its worth taking to the hospital. These programs use a collection of the text from a large collection of medical publications and text books (called a dataset) to gather an understanding of the links between symptoms to be able to give estimated predictions to assist in diagnosis , when presented with a selection of symptoms and patient information. The relevancy of this technology to those in the further field of Computer Science can adapt the same process to develop advisory solutions into other areas like first line support or even to generate interactions within a video game. Expanding from the area of computing now, this technology could also be used by GP’s surgeries in developing nations to help expand healthcare availability and advise training medical professionals on options for treatment or medication. With the rate of these evolutions of this technology it might not be too long now before we are starting to thank our Dr. Roboto.
A successful commercial example of this type of technology is Infermedica.
Infermedica is a medical software company with a few products under their belt now, the most prominent being the symptom checker. This software is focused on the detection and suggested treatment and signposts them to the relevant areas to receive more support from a medical professional if needed.
The system works for a trained model on medical journals and triage documents to learn how best to work out what possible issues might be up with a patient and asking more probing questions to better narrow down the possible conditions they are suffering with. Once the AI has narrowed down the issues, the expansion of this technology which Infermedica uses remotely connects you to a team of their own medical professionals to allow a seamless transition through the levels of care which they offer.
They are a good example of companies starting up and growing in these emerging areas and building a complete service out of it. As meteoric as Infrmedica’s success is, the technology behind the chatbot is able to be partially replicated utilising free and open-source software and tools.
The way these chatbots work is you start off by training an ML model off of a collection of information about, for example, symptoms and treatments for medical issues. When this model is trained it is then tested to improve itself on the data it has. This testing and training are what allows the model to predict with the accuracy that it does.
One of the open-source technologies for achieving this is the BioBERT model by DMIS, this is “designed for biomedical text mining tasks such as biomedical named entity recognition, relation extraction, question answering, etc.” This model has been trained on over one million PubMed articles and has a question-answer model ideal for the situations discussed above.
Taking a step back, the system of training and modelling which BioBERT uses could be applied to any dataset, be it articles on Yoga for an interactive yoga assistant, or the health and safety information of your product to offer an intuitive first line of customer support.
This could be adopted by any business looking to go into the Medical, or Health and Well-being industry for a question and answer technology about many things, and the opportunities don’t stop there.
The same basis for this technology is used for Virtual Personal Trainers (VPTs); you could also utilise Machine Learning through the personalisation and tailored workouts. Using similar technology to the one mentioned earlier you can train a model to offer training on the areas in which the user is weaker to help them improve the areas which need it. This not only would add a lot of value to the service offered, but I personally feel would offer a much better level of training.
A commercial example of the use of this technology is infiGro, an AI fitness app by Infivolve. This app records and directs your workouts giving feedback on the exercises you have just done, and it also aids in nutrition tracking and guided meditation. This commercial use of this technology is interesting due to its use of combining many small technological features and turning it into a complete app.
An open-source implementation of this kind of technology is using the premade PoseNet model to begin. This is a trained model which can detect the poses the body makes through a camera to aid in the detection of where the limbs are in relation to the body. This has been achieved through training the model off of many images of a selection of poses, to understand firstly, which limb it is looking at, and secondly, in which pose/ position that it is in.
This technology can be leveraged to accomplish the technology listed above. There is an open-source solution with an explanation available on Github. This explains how to build a pose detection system for a yoga mobile application with integrated payments to give inspiration for how this technology can be adapted to your business.
The world of ML is growing and expanding into so many industries month after month with new avenues to take your business down and be part of the cutting edge of offerings to your customers, and with most of the technology being free and open-source, it’s a great way to start learning and experimenting with new ideas to move your business in.
As I hope you have seen that with this technology, the successes are out there showing the profitability of expanding your businesses into these areas. Although the learning curve can be a step to get past, the free information and open-source contributions found online these days can be a great stepping stone to building a better business.
More generally, this technology has also shown the societal benefits of the expansion of this technology in the health and medical sectors, allowing greater access to these technologies. With the ever expanding open-source implementations of this, cost is becoming less and less of a factor in passing on the good from this.
Hopefully this article has given you a level of appreciation of the less publicised work getting done around the world. Every new development, product or proof of concept to build on this idea is leading us ever closer to a more healthy and positive world for us all, so instead of being afraid of technology, maybe it’s time we start saying thank you to robots.