Description
We can't expect Machine Learning experts to be both domain experts across various fields and also true experts in ML, it is crucial to rethink how we build tools so that we can provide Machine Learning capabilities to experts in various disciplines that are not necessarily savvy in machine learning, but are experts in their own domains, so that they can too become participants of the data science community.
This talk is divided into three sections:
The first one, dives into the importance of democratizing Machine Learning, its objective is to demonstrate why now this is a crucial issue to be solved and to show the risks and problems that present themselves when Machine Learning Engineers take in the responsibility of building predictive technologies in domains that are not necessarily where they are the most experts on.
The second part of this talk focuses on some solutions and approaches to challenges of democratizing Machine Learning, as well as the journey and results we have seen at MindsDB while at this endeavor as well as what would be new ways that the ML community can think of the next generation of tools being built.
The third and most extensive section of this talk, focuses on the new issues that are born once machine Learning capabilities are on the hands of not ML experts, more importantly, the implications of delegating the Machine Learning machinery to a system, and what is important in order to trust those systems. Which leads to the importance in further developing our understanding of explainability, interpretability, and robustness of ML.