We've covered machine learning throughout this course, but that was just one approach, custom model building. In our first challenge, you may recall you inherited a Spark ML job that your data science team created to predict housing rentals. The second challenge had you create a forecasting model with BigQuery ML from scratch, although you've saved time by coding and running it in just SQL. It's now time to zoom back out on approaches to machine learning and even look at some ways we can apply it without having to write code at all. There are three approaches to AI that you should consider. You have already seen and built custom models with BigQuery ML, and we have a separate set of courses on TensorFlow for even deeper model building. A good rule of thumb is to consider custom model building only when you have a lot of data, like 100,000 plus to millions of examples. But what if you don't? Consider using a pre-built AI which are models like the Video Intelligence and Cloud Vision APIs that you saw before. In addition, if you're looking to build a chat bot, start with Dialogflow, which is a full fledged application with ML built in. But what about if the building blocks don't work well for the specificity, you need on your data? That's when you should consider Auto ML as a good candidate. It can even work with just a little bit of data like 10-100 images per label. This lesson covers each of these approaches in detail. First step is using pre-built AI building blocks for your use case. Pre-built models are offered as services. In many cases, these building blocks can be used to create the application you want without the expense or complexity of creating your own models. Cloud-Speech-to-Text converts audio to text for data processing, Cloud Natural Language API recognizes parts of speech called entities and sentiment, Cloud Translation converts text in one language to another, Dialogflow Enterprise Edition is used to build chatbots to conduct conversations, Cloud Text-to-Speech converts text into high-quality voice audio, Cloud Vision API is for working with and recognizing content in still images, and Cloud Video Intelligence API is for recognizing motion and action in video. Good machine-learning models require lots of high-quality training data. As we mentioned before, you should aim for 100,000 plus records to train on for custom model. If you don't have that kind of data, pre-built models are a great place to start. Let's take a look back at one of the most popular ones that I use when I travel overseas, the Translation API.