“We built it and we just let it run. We’re a few dudes in an office and our goal is to keep it running. It does everything we could do, except it’s significantly more powerful and it has completely automated how our work is being done,” casually said the hedge fund manager as he described the process by which nearly $1billion was being managed within his fund.
The ‘it’ is an artificial intelligence (AI) based algorithm that uses complex statistics to analyze variables that went into successful decisions and uses advanced computer programs to keep replicating those decisions. All this, while it continuously learns from – and improves upon – its mistakes as it encounters new variables.
These machine intelligent systems are applying the many different forms of AI and fundamentally changing the financial industry. From applying Natural Language Processing in detecting Anti-Money Laundering and fraudulent financial activity to applying Cognitive Computing to analyze wide varieties of variables in building better trading algorithms and to leveraging Deep Learning to looking at consumer decision patterns and providing personalized ‘chatbots,’ AI is transforming the financial sector.
One of the most noticeable areas where this disruption is taking place is within hedge funds: hedge funds that are transitioning their trading desks to AI backed systems, are already beginning to outperform hedge-funds backed by humans alone. What’s really quite astonishing though is how, in the short span of a few years, how far reaching the results have been.
Hearing about hedgies working with AI researchers to make even more money doesn’t inspire the rest of us to greatness. However, it may be valuable to look a brief historical overview of how the financial industry reached this juncture.
It was in the 1970s that the National Association of Securities Dealers Automated Quotations (NASDAQ) first began digitally displaying bids and offers on an electronic board. Over the next decade, there were options to place financial orders electronically, as opposed to over the phone/broker-mediated as they were done before. However the real change happened in the 1990s, with internationally connected networks (aka the internet) began bringing together global markets and computing power was much easier to access. This brought together a huge collection of information and there became an easily accessible pool of information which directly and indirectly shaped financial decisions. Applying statistical methods to analyze these large data pools, and then automating future trading decisions based upon this analyzed information – i.e. implementing Artificial Intelligence – is now emerging as a powerhouse that’ll shape the future direction of finance.
The real question though is, why am I writing about the financial industry on The Healthcare Blog?
Although healthcare and finance are two seemingly dissimilar industries, there are many similarities, noticeably how variable rich both industries can be and how parallel the process of innovation implementation is in both industries. Examining the historical overview of how AI is being implemented in finance can share some valuable lessons for us as we being to implement advanced AI based computing in healthcare.
Just as global financial markets had a metamorphosis into the electronic and digital versions that we see today, we are also seeing a similar technological shift in the medical industry today. In the last several years, each and every one of us directly and indirectly involved in healthcare, has seen the influx of Electronic Medical Records, PACS, data from implantables, wearables and so much more. This has created large sets of data. And, although healthcare is on the same parallel trajectory as the financial industry, one difference is that we are several steps behind. We are now at the very beginning stages of implementation of AI in clinical care.
Looking at some of the successes of AI in financial services will help us succeed in the application of AI in healthcare. Three big thoughts come to mind:
- Truly personalized medicine: When physicians look at pathology, we classically look at the symptoms, develop a differential and come up with a treatment plan that will address the problem (or top three problems) on the list. We may even stratify their healthcare risk and tell them to take a few generic preventative measure. What are unable to do is analyze the patient’s genome, proteome, risk of developing malignancy or chronic disease from current pathology/lifestyle decisions and many more unanswered questions. Just like traders are using deep learning and cognitive computing to better understand large volumes of data and act on stock trading patterns, we can analyze significantly more data to better understand what’s relevant, really addressing the patient’s problems and possibly leading to better outcomes. Perhaps this is most applicable in my speciality of radiology, where instead of just interpreting the medical image, we can ‘see what can’t be seen’ by applying advanced AI in interpreting pathology in context of the patient’s entire medical record
- Comprehensive communication: As video visits, tele-medicine, outpatient imaging, patient centered documentation and other facets of convenient care become the new norm for healthcare delivery, patients – rightly so – want access to their information at all times. Just like Natural Language Processing is being applied in finance to use ‘chatbots’ to educate people on better investment decisions and tracking their expenditures, we can apply to it to educate our patients on their disease condition and personalize recommendations on treatment plans. This could be used to bridge education between appointments and also track patients for possible complications – monitoring oncology patients for depression or tracking heart failure patients/COPD patients for acute cardiac or pulmonary exacerbations. Essentially, we can use developments in AI to augment our reach in between the times we directly get to see them.
- AI is an ally: Perhaps the biggest take-away is that developments in artificial intelligence, be it through Deep Learning, Computer Vision, Cognitive Computing, Natural Language Processing and many more will really help extend our reach to truly make an impact for our patients.
Unlike financial traders who just ‘built the system and let it run’ I truly believe we have the power to ‘build it and let it heal.’ As a young physician, I chose radiology because I remain captivated by power of technology to make a difference in people’s health. Personal computing, internet, mobile technology and other technological revolutions are all dwarfed by the promise of AI. It is up to us, as physicians, to apply the best of technology to those who are suffering through life’s toughest times.
About the author:
Ajay Kohli is a radiology resident at the Drexel School of Medicine. He’s a keen entrepreneur, and was named one of “40 under 40” healthcare innovators.
Thanks for sharing this informative information.
I got the same question as you’ve mentioned in your blog post that was why he has written about the financial industry on the healthcare blog for the topic “Artificial Intelligence (AI) in Healthcare” until I read the first half of your post. But, after reading the second half of your blog post I got a clear picture about the similarities between healthcare and financial industries even though both are dissimilar industries. Very helpful as this blog post really helped me to learn some interesting things about AI in healthcare by understanding some of the successes of AI in financial services. Thanks much for sharing such a great blog post. Here is an event you wouldn’t want to miss! Catch our webinar on “Imagining Future of Healthcare – An AIkidos Experience”. Thank you!
The frequency domain, testing processes that could be applied to evaluating the operating characteristics of homeostatic control systems are hampered by the non-linear attributes of these biologic systems. This problem was long ago idealized by an aphorism of sociologist Ashley Montagu who said: “Man is the only 150 pound non-linear servo-mechanism that can be wholly reproduced by unskilled labor.” All things considered, it is certainly a very narrowly conceived observation about the human condition. The scientific evolution of testing has so far remained largely in the time domain. We probably have much to learn about the effects of disruptive events and entropy on Stable HEALTH from the frequency domain for understanding homeostasis, especially as it maintains our memory processes.