20 thoughts on “Intro to machine learning on Google Cloud Platform (Google I/O ’18)

  1. Great video, especially on AutoML :
    Is human labeling performed by real humans ? How do they know how to label, if not experts in the reviewed domain ?
    Is it possible to restrict the origin of API calls ? (Use case : don't want my competitors to use my prediction API) ?
    Is it possible to make predictions offline (something like compile model, then use it in JS in the browser) ?

  2. this video doesn't make me want to use GCP for ML anytime soon. Most if not all what she talked about can be achieved quite easily in Jupyter notebook with opensource software (fast.ai etc)

  3. How do you create ML for sifting through pubmed and analyzing medicine affects on specific genetic disorders? I know it would need 2 libraries, but how can one get it to connect missing dots overlooked by humans?

  4. I believe the example code for the wide model given at 29:48 is wrong. The model should be created like this:
    wide_model = Model(inputs=[bow_inputs, variety_inputs], outputs=predictions)

Leave a Reply

Your email address will not be published. Required fields are marked *