Design as a tool to ML
Lately I have been searching for ways to expand my creativity and augment my productivity. In searching for it I've found some early explanations of when the computer was introduced in our society and in this explanations they view the computer as a co-creator, a tool to enable the creation of new things. You can read more about this here.
This links to me to a relevant information of where ML will take us? If you see we have a big trend of people moving to the area with the standardization of tools like Tensorflow and Pytorch (Tools used to create machine learning models). If we look at the start of the ML rise, we will see that we evolved from more raw tools like Theano and Caffe, then it came Tensorflow and Pytorch first versions and now we are seeing more high level tools like Keras and FastAi. The big difference here, in my view, came from the fact that each new library was built in top of older libraries knowledge and then the API was made easier. In this way new users can create machine learning models more easily. Enabling more people going into the field, but with that in mind I ask myself where can we evolve?
Also we aren't seeing another trend, major ML developers are talking that by using the models that we already have and giving them more hardware/data we can achieve better results. In the book AI Superpowers from Kai-Fu Lee he talks about the fact that when an enterprise has a lot of access to data and processing power, the difference between a good Ml practitioner developing machine learning models and a not so good practitioner will be small, because what matters is the amount of data that one have. It seems to me that to grow now, the major benefit comes from a better designed data pipeline or a better data curation flow.
With the question in mind and this related trend, I began to feel that we needed another view, so instead of arguing how difficult is to grow in ML models. I now think that we should analyze how ML is being used. One usage that I became interested in was the one from Patrick Hebron in his talk: Integrating Design and Technical Innovation in AI-First Products. From his speak I came to realize that ML will come to us like the computer once did, enabling us to better communicate between the creation of new things and the computer. Better translate from the thinking space to the software space. This trend should come to make more people work in creative tasks, but we still need to make the bridge between how the user views machine learning in a simple way, without realizing that underneath the hood ML is being used.