How does a robot adapt to its environment? What’s a fast way to translate English to Chinese? Machine learning specialists use data to answer questions like these every day.
We recently talked with Aitzaz Ahmad, an applied scientist at Amazon. His career has included work on robotics and natural language generation, which converts data into patterns that look like human text or speech. He’s also an instructor for the UW Certificate in Machine Learning. He teaches about neural networks and deep learning, two kinds of artificial intelligence that imitate how connections are made in the human brain.
“Everything that you’ve seen for facial recognition, machine translation, language generation, recommendation systems — it’s less than a decade old,” he said.
Here’s what Ahmad had to say about the blazing-fast world of machine learning and how to prepare for the jobs of the future.
In simple terms, what do we mean by machine learning?
Traditional programming is writing rules. Machine learning is when we learn from data.
We’re collecting massive data from online platforms, social media, e-commerce websites; you name it. The goal is to identify patterns that would be too complicated for us to write as a math equation.
For example, object recognition. The most success in machine learning has so far been in supervised learning. That means somebody has provided a data set — say cat pictures, all sorts of different scenarios where cats are sitting, standing, running, different colors of cats — and annotated it. Instead of some expert writing all those algorithms, we can just show a machine lots of this labeled data, and it generally learns the patterns of what a cat looks like.
What kinds of modern problems does machine learning help solve through patterns and predictions?
People are trying to train systems where bots are intelligent enough to at least answer some basic questions. If an automated system can do it, that would reduce the resources needed in customer service.
Machine translation has become very commonplace. If you don’t know a language, you can just speak English, and apps can translate it so you can communicate.
And we have voice assistants — Amazon has Alexa, Apple has Siri, Microsoft has Cortana and Google has Google Voice. What’s happening in the background is the collection of data. They’re learning from the data to help make the service better for you.
What advancements do you see as potential game-changers in machine learning?
One thing that I’m really excited about is artificial intelligence or machine learning in health care. Also, machine learning or deep learning for robots. Robots have already made their way to factory floors; I have worked on that. There used to be a cartoon called The Jetsons, where they had a robot in their house. That is something that excites me a lot about the future because the robot will be a mix of a lot of technologies.
Mechanical engineering, to get the robot up and running, then programming to control its movement. The brain of the robot is going to be powered by AI. The robot would need to have human-like vision and maybe some language capabilities so that they can communicate.
What kind of education or experience is helpful if you want a career in machine learning?
The ideal would be someone who has mathematical background, statistics courses and coding experience. If it’s in Python, that’s even better. All good machine learning practitioners have one thing in common: the desire to learn something new. If you don’t keep up with the latest technology, you become a dinosaur pretty quickly.
Professionals work with a range of tools, from algorithm libraries such as scikit-learn, Keras and TensorFlow, to techniques for natural language processing, such as BERT. What are the benefits of learning these skills through the Certificate in Machine Learning?
The certificate is designed by people working in the industry who have seen both the theoretical side of things as well as the practical aspects of launching large systems. These practitioners organized the courses into a logical sequence, and they impart their ideas to students, which can give them a head start.
The program starts with an introduction and ends with recent advancements in computer vision and natural language processing. Assignments and quizzes hold you accountable every week and get you experimenting.
Certificate graduates often go on to jobs in data science and machine learning. What’s driving Seattle-area demand for these specialists?
The fact that Seattle is home to two trillion-dollar companies — Microsoft and Amazon — drives a lot of demand for machine learning engineers. These companies are opening new avenues of work every day, and they need more engineers and machine learning practitioners.
Tech companies headquartered in the Bay Area are setting up big shops here. Google has a significant presence in Seattle, so does Facebook. We also have eBay, Expedia, Boeing, Nordstrom and Starbucks here. They’re all using machine learning and data science.
What’s the most important step for aspiring machine-learning practitioners?
Start writing code. Get your hands dirty with data. You can learn everything in the book, but real-world problems have their own set of challenges. Nothing beats practice.