Please feel free to correct any misunderstandings I have of the evaluation in the tweets that follow.
To begin, the Babylon engine is a Bayesian reasoner. That's cool. Not sure if it qualifies as AI.
Jan 25, 2018 • 10 tweets • 3 min read
“We’re spending our days doing the wrong work” - the despair of clinicians forced to use the wrong solution to clinical documentation is palpable nejm.org/doi/full/10.10…
The reason the modern EHR causes so much pain is that it is a conceptual mistake. The EHR sees the clinical encounter as a process of documentation with recording data primary, rather than as a process of sense making and deciding, where documents arise as side-effects.
Jan 21, 2018 • 11 tweets • 5 min read
A few weeks ago I tweeted a set of simple rules to help interpret papers reporting new #AI results, as well as media coverage of those papers. Here they are again collected into a single thread. #AImedia1/ If humans helped train the AI eg telling it which examples are from one class or another, or by selecting the training data, then it didn’t “beat” humans, it had help. So report that human plus AI outperformed human alone. #AImedia
