User Question: Is XAI for both data scientists and non-technical users?
The answer to this is many-fold. Our objective is to make machine learning accessible and, by that mission, we want to make sure that it's for everyone so we're putting a lot of effort into making sure that the explainability side works for both data scientists and non-data scientists. We also want to make sure that you can use it as long as you have some data.
Now, we understand that explainability is something that data scientists themselves want to incorporate into their products. Essentially, the building blocks of explainability, given the approach that we’ve taken—which the black box approach—make it so they can do this. One way this works is that if you have a model, you’re able to wrap that model around the MindsDB interface. The MindsDB interface can then run our explainability framework on top of that model. This benefits people who have already been building models without an explainability component.
The second thing is, if you are someone who builds new predictive models, we have a framework called Lightwood, which is the main building block of MindsDB for neural networks, that allows you to treat machine learning models as building blocks, or lego blocks. What we understand now is that if you treat a particular problem as having different data types as both the input and the output, one of the very first things that you can do is to understand how you can build very rich vector representations of the different data types and features that you have. What MindsDB does to this end is it finds the most state-of-the-art way to build a very rich vector representation from a specific data type and provides you with some statistical information about your data types. Once you have rich vector representations for all your input data and input features, what you’ll need to do is determine how you’ll mix all this information so that it gets to the target variable. To do this, MindsDB tries to build the most reasonable neural network topology. The mixers that it builds right now enable data scientists who haven’t tapped into probabilistic neural networks to benefit from them. MindsDB has a very interesting approach to this in that it will build a probabilistic neural network that mixes data and gets to the target so that your answers are not deterministic (because most predictive models are not), but you can also gather how certain a specific model is of the solution it came up with. You can use this as a data scientist.
To summarize, if you are not a data scientist, MindsDB is meant to provide you with the tooling for you to build a machine model and get explanations. However, if you are a data scientist then MindsDB is meant to provide you with the scaffolding to get explainability on the top of your models. Also, if you’re creating new models, it provides you with a framework to do this so you can focus on the parts that you really want and then automate the rest.
Jorge Torres is the Co-founder & CTO of MindsDB. He is also a visiting scholar at UC Berkeley researching machine learning automation and explainability. Prior to founding MindsDB, he worked for a number of data-intensive start-ups, most recently working with Aneesh Chopra (the first CTO in the US government) building data systems that analyze billions of patients records and lead to highest savings for millions of patients. He started his work on scaling solutions using machine learning in early 2008 while working as first full time engineer at Couchsurfing where he helped grow the company from a few thousand users to a few million. Jorge had degrees in electrical engineering & computer science, including a masters degree in computer systems (with a focus on applied Machine Learning) from the Australian National University.