A couple of weeks ago, I had the pleasure of attending the Artificial Intelligence (AI) Summit in San Francisco. There I had the opportunity to learn more about AI and to get a better understanding, given the unabated buzz between AI hype and reality.
It was definitely interesting to see organizations with true AI capabilities, as well as those wannabes who are all of a sudden AI experts. Organizations such as Amazon, Facebook, Google, and Microsoft, all have a pretty good grip on the AI front given their respective platforms (e.g. TensorFlow, Cognitive Services, etc.) and more importantly having access to vast amounts of data sets. From my perspective, organizations who have access to large meaningful data sets and intelligent AI algorithms are going to be the leaders and game changers in this space. Whereas organizations who have access to just one of these ingredients will have a hard time being successful.
This brings me back to the fundamental question – what’s in your data? As written in one of my previous blogs – Artificial Intelligence Euphoria in Healthcare, it is clear that AI will have a major role in our society. From a healthcare standpoint, this is very crucial as historically the industry has had access to vast amounts of patient data (structured and unstructured). The right AI algorithms could provide incredibly powerful, actionable insights on everything from chronic disease management to providing doctors more sophisticated tools in battling cancer prevention, detection, and personalized treatments. AI can provide better efficiency by processing more data, identifying patterns, and in many cases, discovering solutions that are not immediately apparent through traditional human-led data analysis.
Let’s take a scenario where a hospital is trying to identify patients who have the highest risk exposure of not being able to make their bill payments. By applying AI algorithms, the hospital can predict the high-risk patient pool and take appropriate measures. The accuracy of the inference model would improve dramatically if the data source contains multiple valid inputs, such as patient income, work history, demographics, zip code, prior payment history, etc. If sophisticated AI algorithms are applied to insufficient set of meaningful data, then the output is going to be less viable. The point being, one has to have the right kind of data in order to build an accurate predictive model.
Another use case is in the personalized medicine front, where AI can be leveraged to enable providers and patients to choose one medication over the other. Longitudinal machine learning (ML) cluster analysis could reveal drug exposure and efficacy by analyzing patient data sets based on areas such as demographics, co-morbidities, medication profile, genetic disposition, etc. Again, in this scenario, it comes down to have the right data sets to apply ML techniques to determine an outcome with significantly high levels of confidence.
While I do anticipate bumps and detours along the way, I do feel very optimistic that AI will be a game changer in multiple industries, particularly in the healthcare space. According to Accenture’s Digital Health Technology Vision 2017 report, 84% of healthcare executives believe AI will revolutionize the way they will gain information from and interact with customers. I am a firm believer that AI technology will fundamentally change the healthcare ecosystem by:
- Empowering patients to manage their own health.
- Improving the healthcare organization’s productivity.
- Driving exceptional science and clinical innovation.
- Enabling advanced analytics to power actionable insights at scale.
In order to continue down the AI path with a high level of accuracy, we will need to find and apply right AI techniques to meaningful data sets. I would love to continue this conversation and get your thoughts on this topic.