A day in the life of an algorithm

- [Narrator] Applied machine learning. Just taking a step back and looking at the big picture. In this course responsible AI, will talk about the creation of models, how they're implemented and different steps to actually check if they've been used responsibly. But I thought I'd just start off with a high level view of exactly what is the point of model development and deployment. And how it looks end to end. So, this deployment comes under the banner of applied machine learning. So, applied machine learning, starts with an amount of research and development into creating an algorithm or machine learning model. Implementing that model, monitoring that model and then ensuring that it's been used responsibly. This process usually starts with finding an organization problem. So, finding some type of use case or business case, within an organization that needs to be, for lack of a better term automated. There needs to be some type of enhancement to this process. And that enhancement will be delivered by a machine learning algorithm. Once that problem has been found, the challenge then, is to convert that problem into code. And there are two sides to this; there's one code to actually estimate what should be done. And then there's code that takes that estimation and allows it to be decided upon. So, allows it to actually be put into practice. So, one example of this maybe a production line, looking to estimate how many problems arise with created goods. And it does this by taking a photo of the good and estimating the amount of damage that that good has taken. Another example may be a marketing company, looking to understand their customers. And only send certain marketing collateral to customers that have a high likelihood of performing some type of action. So, that problem to be able to both find defects to reduce the cost of creating products. And to increase the amount of sales for marketing company, needs to be defined at the start of the project. To ensure that any automated decision process is able to deliver on that goal or to solve that problem. Once that problem has been converted into code, there needs to be a number of checks that that code is robust from a technical level. So, not only that it can serve the number of required instances, but also that it's able to actually improve the decision making process above what's already been done. Now, the last area here is where responsible AI comes in. So, at a high level again, you need to ensure that the code can operate with people so that, the decisions that have been made automatically, they're able to help serve people at a lower level and at a higher level is able to work with the different laws and governance in the areas that the models are operating in. So, this end to end process of applied machine learning will be revisited over this course. And we'll give some extra information about how responsible AI can help with it.