Machine learning (ML) can be a boon for software developers as they will be able to train computers through algorithms and statistical models for a specific task to be performed effectively without the use of instructions. The computer system just relies on patterns and inference. ML algorithms can be used in a variety of applications such as email filtering, computer vision, etc. According to Gabe Hollombe, Senior Technical Evangelist, AWS, the innovation took real shape only when software companies realized that a simple hardware device such as a graphics card that lets computers draw better 3D models for better gaming experience can do a great job of training machine learning models.
“Machine learning basically means adding and multiplying numbers very quickly. It has been there for quite some time. However, this was only the domain of really big companies as they were the only ones that could afford as training a computer requires a lot of compute power,” says Hollombe.
However, with the advent of cloud computing things really started changing. Cloud computing provided the kind of compute power required on demand, that is, the compute can be made available only for the time somebody needs to train a computer on a specific model.
“It used to require so much capital expense that small companies couldn’t do it earlier. But now they can. Innovation in hardware, coupled with the realization that we can use different types of hardware, is responsible for the upsurge in ML tools. We at AWS are providing that capability,” says Hollombe.
But, it is not just about the hardware. Companies can offer ML capabilities through software as well.. A case in point is Amazon SageMaker, which sits in the AI and ML stack of AWS. According to Hollombe, the tool provides every developer and data scientist the ability to build, train as well as deploy ML models quickly.
“Amazon SageMaker service covers the entire ML workflow to label and prepare your data, choose an algorithm, train the model, tune and optimize it for deployment, make predictions, and take action. Your models get to production faster with much less effort and lower cost. Through it one can teach the computer and then host the model somewhere else so that it works in a secure and scalable manner.
You Don’t Need a PhD in Data Science to Leverage ML
As per the technical evangelist, one of the biggest advantages with Amazon SageMaker is that one doesn’t need to know how to teach the computer to train on ML models. It already has a number of built-in algorithms. These have been formulated after years of research. The tool guides you about which algorithms to apply to different types of problems to make these models and that too in very short time.
“I tried this out myself. I am a bird watcher. I trained my computer to identify birds. I took 350 images of Bird A and Bird B and I said learn. Within 3 minutes, it came back with a model that was 98% accurate. It blew me away. In fact, the system gets better in figuring out the differences than I can. And, do you know how much it cost me to train that model? Just 20 cents,” concludes Hollombe.