The biggest breakthrough here is the FDA's acceptance of basket studies-- claiming that they would approve drugs on it alone. That's a big win for patients, who based on current practices, would have found themselves in the control group.
People are missing the point that there isn't an actual sim card now, that you don't get the "apple" convenience AND the ability to pop in some random carrier's. It's convenience at the cost of flexibility.
That's actually pretty terrible. I was thinking it was a SIM card with some Apple-designed logic in it. Not a baked-in SIM card. I wasn't really considering the device but now it's out of the question. Hopefully they don't do the same thing to their phones.
We're at a slippery slope where we'll simply need more data points to better assess identification of medical predispositions based on gene analysis. Unfortunately, the negative externality that forms, and the one referenced when things like "should be" is said, pertains to patients' responses to this additional information. We don't even have enough data to full reflect on how patients are reacting to this information.
Right now, most medicine is reactionary-- confirming suspicions when patients are presenting x symptoms. These tests, however, hope to be predictive, and like many predictive aspects of life, will have associated confidence intervals. People, especially with regards to their own health, will have a hard time factoring in these odds when presented additional information.
True, but intuition and general analytical ability is often overlooked when hiring data scientists -- because the term is so broad, what people are often looking for is a mixture of statistical know-how and programming ability, but often forget to ask for a good understanding of econometric principles.
For instance, there is a difference between understanding generalization error in the statistical/machine learning sense, and generalization error in the sense of external validity. You'd be surprised at how many models, even at reputed companies, use features that do not make sense or are based on implicit assumptions about the studied population.
Obviously insight is always a good quality to have, but it is particularly critical for data scientists as the right or the wrong insight in the context of machine learning algorithms can have huge consequences on your company.