5 Matching Annotations
  1. Jul 2021
    1. Short interview that covers the findings of a systematic review that aimed to identify the number of studies trying to replicate the outcomes of clinical decision support systems.

      Article being discussed in this piece: Coiera, E., & Tong, H. L. (2021). Replication studies in the clinical decision support literature-frequency, fidelity, and impact. Journal of the American Medical Informatics Association: JAMIA, ocab049. https://doi.org/10/gmb35n

    1. The researchers started with 140,000 hours of YouTube videos of people talking in diverse situations. Then, they designed a program that created clips a few seconds long with the mouth movement for each phoneme, or word sound, annotated. The program filtered out non-English speech, nonspeaking faces, low-quality video, and video that wasn’t shot straight ahead. Then, they cropped the videos around the mouth. That yielded nearly 4000 hours of footage, including more than 127,000 English words.

      The time and effort required to put together this dataset is significant in itself. So much of the data we need to train algorithms simply doesn't exist in a useful format. However, the more we need to manipulate the raw information, the more likely we are to insert our own biases.

    2. They fed their system thousands of hours of videos along with transcripts, and had the computer solve the task for itself.

      Seriously, this is going to be how we move forward. We don't need to understand how it works; only that it really does work. Yes, it'll make mistakes but apparently it'll make fewer mistakes than the best human interpreters. Why would you be against this?

  2. Jun 2021