AI Med Day One: Key Takeaways
Posted: 12 December 2017 | By Charlie Moloney
Dr. Anthony Chang just shared these key takeaways from yesterday’s workshops at AI Med, the conference on AI in Healthcare hosted in Orange County this week:
– AI in medical images having impact
– Deep learning is not the only solution
– Statistics is not knowledge
– Beware of automation bias
Let me share with you my key takeaways from day one before the keynotes get started today:
– Wearable devices which can collect data from children in the important developmental years of 2-4 could help lead to an earlier diagnosis of lifelong conditions, improving overall care outcomes. For example, a Google glass style pair of glasses which can track eye movement could reduce the age at which autism is currently diagnosed (4 and a 1/2 years old) to 2 years old max.
– According to Gary Marcus, a renowned expert in the AI world, we may never crack the problem of getting machines to converse/understand with natural language. People have been trying to understand linguistics and the logic of language for decades, and haven’t made much progress. Systems don’t understand anything that we’re saying, they just look for keywords and hope for the best. That may be fine in certain fields, but in a domain where you need a high degree of accuracy, like medicine, will a statistical solution like that provide any value? Prof. Marcus thinks not.
– Spyro Mousses, Chief Scientist at the Children’s Hospital of Orange County (CHOC), shared the prediction that by 2030 a single cognitive system will be able to generate more new biomedical knowledge in one hour than in the entire history of human scientific endeavour.
– The quality of care at the moment is poorer than we generally realise: the average US patient can expect to be harmed by a diagnostic error at some point in their lifetime.
– A big challenge is creating practical, reproducible frameworks that can be used by clinicians, because unless new technology can be implemented it is innovation theatre.
– Another challenge which persists is finding a mediator between medical professionals and technology professionals. Without co-operation at every stage, medical professionals will not accept the introduction of new technology and technologists will not create solutions which will solve real-world medical problems.