Chapter 5 - Opening Case - Elementary, Watson. Not Quite so Fast, IBM Flashcards
What happened in 2015 with IBM?
IBM launched Watson Health, which was an artificial intelligence (AI) platform with the purpose of helping health care providers revolutionize caner treatment.
What are the hypothetical advantages of AI in the healthcare setting
They provide deep insights and suggestions, by consuming vasts amounts of data, analyzing and interpreting it, and providing insights. They can use AI to summarize data about cancer as a whole when dealing with individual patients rather than just from the data they have been able to review.
What happened to Watson Health by 2021? What happened by 2022?
1) It did not deliver the progress it was hoping for.
2) it only made revenue of 1 billion when the investment cost was 4 billion.
3) Many potential clients cancelled their orders.
In 2022, they sold Watson to another health company.
What problems came to be that caused the failure of Watson?
1) Sheer complexity of health care data - While Watson could analyze it, it required human capital to create points so that Watson could analyze it.
2) Combining multiple data records into a single story is impossible.
3) The data entered into Watson was not actual patient data, providing insights that many healthcare professionals found as sceptical.
4) Watson fails to consider that location specific needs are related to socioeconomic status.
What is AI bias?
The artificial intelligence is biased against individuals and does not consider cancer patients from all walks of life. For example, having a hospital with very rich patients and using this data would not accurately reflect society as a whole.
What was IBM’s response to these failures? How much info did each company give/ what type of info?
They spent billions of dollars of acquisitions with Truven, Phytel, Explorys, and Merge.
Turven - 300 million customer datapoints
Explorys - Clinical database of 50 million patients
Phytel - Physician data
Merge - Huge imaging database.
What is the overall message of the possibility of AI in healthcare?
While healthcare is data rich it is analysis deficient meaning that AI has no role in this place. Until there can be a unified data approach where items can be shared, accessed, and analyzed can only AI truly be used.
What is Fast Healthcare Interoperability Resources (FHIR)
A standard that defines how healthcare information can be exchanged electronically.
What is Health Infoaway?
Federally funded, not for profit organization that has the mandate of speeding up the adoption of digital healthcare solutions across Canada.
What is a positive that healthcare is capable of doing?
Image classification and natural language processing (understanding spoken commands and carrying out a basic conversation, but health care is much more than just this.
