The roots of AI go back to the 1950s, and include many disciplines & approaches. But the recent AI boom is mostly about Machine Learning.
More often than not, when people refer to Artificial Intelligence (AI) today, what they really mean is Machine Learning.
Machine Learning is essentially pattern matching, and it consists of a conceptually simple workflow.
Get a dataset representative of the domain you are working with. Curate and structure it.
Find a machine learning algorithm (aka model) that matches the structure and patterns in your data, and “train”, i.e. fine tune it, on your data.
Run the algorithm, feed it with test data, and check whether the results are valid for your data. Repeat until you get good results.
Then deploy the algorithm, feed it with new data, and get live results.
If you’ve done a good job, the algorithm should be able to give results such as what an image depicts, or what a customer is likely to buy next.
What does graph have to do with machine learning? A lot, actually. Machine learning, and its deep learning subdomain, make a great match for graphs. Machine learning on graphs is still a nascent technology, but one which is full of promise.
Amazon, Alibaba and Apple are just some of the organizations using this in production, and advancing the state of the art. Already, more than 20% of the research published in top AI conferences is graph-related.
Domain knowledge can effectively help a deep learning system bootstrap its knowledge, by encoding primitives instead of forcing the model to learn these from scratch. Nathan Benaich, author of the State of AI Report
Knowledge graphs is the best technology we have for encoding domain knowledge.