“AI (artificial intelligence) is easy but AV (audio-visual) is hard” was a joke shared among speakers and attendees during AIMed North America took place in Dana Point, California between 13 and 15 December 2018.

The remark contradicts two forms of technology: while AI put a heavier demand on users to acquire its related knowledge, it is far less of a hassle as compared to AV. Since it exempted users the embarrassment of not being able to smoothly maneuver between their presentation slides or to ensure their mics work as they speak. 

Regardless of which, AIMed North America is a magnanimous event. It welcome everyone who is interested in AI and its allied topics. Although majority of the attendees come with medical background, their expertise in data or AI is seldom question. Thus, it can be confusing at times. Especially when everyday vocabulary like “data”, “build”, “train”, “system”, and “machines” are now coated with brand new meaning. 

“What do I need to know as a doctor?” asked a delegate during one of the pre-conference workshops on AI and machine learning (ML). It was a spot on query which most medical professionals have. 

Know the terminology, trends and benefits 

“On a personal note, it depends on you to learn what are the possibilities and pitfalls… for you to form an opinion, at least have an understanding of what is good and bad” answered Crystal Valentine, chief data strategy officer of Eventbrite.

As of most expertise, step one often includes knowing the common terms. Understanding the meaning of phrases like “training an AI model” or “robustness of data” etc. and differentiating between machine learning and deep learning. All these provide tremendous help in keeping an eye on what is out there and keeping up with a AI-centered discussion. 

artificial intelligence medicine AIMed North America 2018 data machine learning deep learning conference workshop literacy technology
Crystal Valentine and Dr. Christina Chen at the pre-conference workshop on 12 December 2018

“It’s good to get a sense of the types of model out there and what can you do with each of them” Dr. Christina Chen, nephrologist of Beth Israel Deaconess Medical Center, Harvard medical school added. 

Like medicine, AI is an umbrella term. It encompasses anything between a chatbot which programmed to answer specific questions to a fully autonomous machine which can make its own decisions. While knowing the mechanism and programming language may be helpful, the more important part is to understand how AI can be integrated into one’s practice. 

Most medical professionals are positive with the possibilities brought about by AI and new technology. It’s important to distinguish between hypes and trends. Keep in mind that new technology is like a new kind of treatment option, not everyone will benefit the most out of it. 

“When you train a (AI) model with a kind of population, will it be able to translate to another population?” Dr. Christina Chen said. 

One can’t be a know-it-all, so do machines 

In another pre-conference workshop on basic AI for clinicians, Sam King, industry fellow, Center for digital transformation of University of California, Irvine cited a rather bizarre experience. 

A 6-year-old shouted dollhouse and cookies to Alexa Echo and the next day, a US$270 worth of dollhouse and cookies arrived at the child’s doorstep. Something which surprised the child’s parents turned out to be the virtual assistant misunderstood the shouting as a form of online order. 

artificial intelligence medical medicine data AIMed North America 2018 data machine learning deep learning conference workshop
Sam King at the pre-conference workshop on 12 December 2018

The fact that AI is designed by human mean biased may be involved. AI is still on its way to perfection and it’s something which may not be able to achieve. During the process, AI will continue to make mistake like the one cited above. In medical field where mistake equals to life and death, presently, there is still a lack of capacity and infrastructure to deal with AI. 

Surely AI will not create the kind of awkwardness AV gives. However, concerns derive from the use of patient data to train AI and the possibility of a provider/ patient power struggle in near future, all these, are well beyond AI literacy. 

Author Bio
synthetic gene empathy chinese artificial intelligence data medicine healthcare ai

Hazel Tang

A science writer with data background and an interest in current affair, culture and arts; a no-med from an (almost) all-med family. Follow on Twitter.