Artificial Intelligence: A Snapshot
“I know I’ve made some very poor decisions recently, but I can give you my complete assurance that my work will be back to normal. I’ve still got the greatest enthusiasm and confidence in the mission…and I want to help you.”
You may recognize this
ominous reassuring statement by HAL 9000, the fictional onboard computer that controlled all onboard operations and interacted with the crew of the Discovery One in the book and film “2001: A Space Odyssey,” released more than a half century ago now. One can go back even further to find imagined sentient machines in works such as “The Twilight Zone,” “Star Trek,” “The Jetsons,” and the film “Forbidden Planet” in 1956 - the same year that the term “artificial intelligence” (AI) was first used at a Dartmouth College workshop.i More recently, “The Terminator” film series painted a dark picture of potential uses of AI. In all of these science fiction classics, humans imagined machines of the future that could not only process information quickly, but could think.
As we know, part of that projection of the future - complete with AI - is here, it’s all around us, and it’s growing quickly. That’s why I call this brief discussion a “snapshot”: it would have been different a year ago, and it will be different a year from now.
AI is here in autonomous vehicles. The self-driving truck company Embark already has a fleet that regularly covers a route between Los Angeles and El Paso, Texas. Although most commercial autonomous vehicles continue to have a human in the driver’s seat to monitor systems and keep things safe (also called “Level 2” automation), many of these vehicles are capable of driving long distances without human intervention.
AI is here predicting and suggesting our purchases. Most of us who have purchased items online know that “Most people who bought a Thingamabob also bought a Whatchamacallit.”
AI can predict voter behaviors, as we know from watching the news.
AI is being used more and more in facial recognition systems, for good or for bad.
AI is used by “intelligent assistants” such as Siri, Alexa, and Cortana. Amazon (which markets Alexa units) has disclosed that its personnel monitors some conversations that are heard in homes, businesses, and who-knows-where by its devices.ii
AI can write music for a single listener based on his or her preferences.
AI can predict the location of rare animals in the wild, and it can do thousands of other things.
That includes things in the health care industry.
Artificial Intelligence (AI) in the Health Care Industry
On November 14 of this year, the Wall Street Journal first reported about “Project Nightingale,” in which Google and health care system Ascension partnered in what “appears to be the biggest effort yet by a Silicon Valley giant to gain a toehold in the health care industry” by obtaining and processing large quantities of data, some of which would be analyzed through AI algorithms.iii
That disclosure emphasized that AI now has a real and growing presence in the health care industry. In this industry, AI is used, or has the potential to be used, for many processes across the GxP realm, including planning of clinical and nonclinical studies, data analysis, early detection of safety signals, and many more.
With this quickly growing and important technology, it’s important that personnel have a good understanding of what AI is, including actual and potential uses and common terminology. Recent surveys indicate there’s a wide variation between use and understanding of AI.
AI is characterized by the programming of computers in such a way that they mimic human thinking processes. The increasing use of AI in recent years is primarily due to advances in computer processor speed and capacity, speed of data transmission, and use of the cloud to store and retrieve data. According to one survey, “Some 96 percent of companies expect to see an explosion of machine learning projects in production by 2020."iv Therefore, to use an analogy from hockey’s Wayne Gretzky, future success of health care companies may not involve having a working model based on where technology is today so much as where it’s going to be.
AI can be a great option in solutions that require repetitive actions. Renée Arnold, CEO at Arnold Consultancy & Technology LLC and Expert at YourEncore, says that health care process owners can ask themselves questions such as: “Could there be a way to automate this? Here’s my next patient; how do I treat them? There must be a way of addressing this process in an automated manner.”
As Renée points out, the use of AI requires large amounts of data, which need to be processed in a fast, powerful, automated manner. A logical source of such data is existing health care records. Accordingly, Google-Ascension’s Project Nightingale, and similar efforts that have been underway for years, use such records. Large amounts of data not only lead to more accurate conclusions when AI is used, but are actually used to teach deep learning to machines.
This is a good point to look at some of the terminology used in AI.
As with most fields, AI has its own terminology. It’s helpful to have a common understanding of these terms. There is some variation in the specific verbiage used by various associations, but the Canadian Association of Radiology offers the following that are representative, concise, and easy to understand. This is a very brief subset of all AI terminology, but these are some basic, commonly used terms:v
- Artificial intelligence: Capability of a machine to imitate intelligent human behavior
- Deep learning: Use of artificial neural networks with multiple processing layers to learn representations of data with multiple levels of abstraction
- Machine learning: AI subfield that provides machines the ability to learn from data without being explicitly programmed
- Neural network: A model composed of layers consisting of connected nodes inspired by neurons in a biological nervous system
Why Use AI?
Two big reasons to use AI are:
- Better, faster, and more accurate analysis of data, and better prediction of results
- Reduction of cost, as AI is useful to save repetitive steps that a human would otherwise perform
Potential Concerns With AI
Why isn’t everyone in the health care industry using AI already? The following reasons have been mentioned at conferences and in surveys:
- A lack of knowledge or understanding about AI. This may be especially true in smaller companies in which there are fewer resources to research potential uses of AI. As was the case for years after the publishing of 21 CFR Part 11 (the FDA rule on electronic records and electronic signatures), there is a wide range of understanding about AI.
- Bias: Analysis of data for any study is only as good as the degree to which potential sources of bias have been eliminated. This includes intentional and unintentional bias, alongside end-to-end study execution: study planning (design), study conduct (data collection), and study reporting (analysis and evaluation).
- Lack of volume or need for AI and other automated processes
- Cost of implementation and validation: This is an interesting reason, given that cost savings can be a big reason to implement AI, thereby saving repetitive actions by a human. In addition, I would argue that there is frequently a cost savings (as well as good business rationale) for validating technology that you want to be sure works as it’s supposed to.
In this brief discussion, we’ve looked at the growth of AI, including its use in the health care industry. We’ve shared some AI terminology, and described some advantages and concerns with using AI.
In upcoming articles, we’ll discuss more specific uses of AI in the health care industry, including planning and analysis of clinical trials, pharmacovigilance, genomics, and uses in the early stages of research and development. We’ll also look at the relationship between automation and artificial intelligence.
A 2017 Deloitte survey of 1,500 senior executives throughout multiple business sectors found that only 17 percent of executives “were familiar with both the concepts and their application in their companies.”vi Surveys of health care decision makers have generally found a similar opportunity for increased knowledge. However, that same Deloitte survey indicated that more than 80 percent of respondents have “already achieved either moderate (53 percent) or substantial (30 percent) benefits” from these technologies.
So, there’s confidence in the existence of those opportunities.
I look forward to visiting with you in “AI in the Health Care Industry, Part II.”
About the Author: Tony Ettwein, Former Sr. Manager at Pfizer, and Managing Expert for YourEncore’s Quality Center of Excellence, is one of YourEncore’s premier subject matter experts in IT Quality, increasing the effectiveness and efficiency of quality auditing by managing enterprise-wide computer validation quality and compliance teams.
i. Matthew Hutson, “Artificial Intelligence, In So Many Words,” Science magazine, p. 19, https://science.sciencemag.org/content/357/6346/19.
ii. Nat Levy, “Amazon confirms that employees listen to Alexa conversations to help improve digital assistant,” GeekWire, https://www.geekwire.com/2019/amazon-confirms-employees-listen-alexa-conversations-help-improve-digital-assistant/
iii. Rob Copeland, “Google’s ‘Project Nightingale’ Gathers Personal Health Data on Millions of Americans,” Wall Street Journal, updated 11 Nov. 2019, https://www.wsj.com/articles/google-s-secret-project-nightingale-gathers-personal-health-data-on-millions-of-americans-11573496790.
iv. Macy Bayern, “Why Machine Learning Will See Explosive Growth Over the Next Two Years,” TechRepublic, https://www.techrepublic.com/article/why-machine-learning-will-see-explosive-growth-over-the-next-2-years/
v. An Tang, MD, MSc, et al, Canadian Association of Radiologists White Paper on Artificial Intelligence in Radiology, Journal 69 (2018), p. 123. https://www.sciencedirect.com/science/article/pii/S0846537118300305
vi. Thomas H. Davenport, Jeff Loucks, David Schatsky, “Bullish on the business value of cognitive: The 2017 Deloitte State of Cognitive Survey,” p.3, https://www2.deloitte.com/content/dam/Deloitte/us/Documents/deloitte-analytics/us-da-2017-deloitte-state-of-cognitive-survey.pdf