It’s easy to write a post about the general efficacy of AI in the enterprise, this post seeks to delve into how an enterprise can quickly and effectively start seeing ROI from machine learning.
Let’s face it, machine learning and AI are here and they are a force to be reckoned with. Why is that? Because machine learning represents the biggest return on investment in a generation.
For a primer on what AI, machine learning, neural net and other terms mean, read this post.
Ok, lets talk business. The reason why you’re reading a lot about machine learning and AI today is because of a somewhat silent revolution in computing. For a while, Moore’s Law seemed to have broken down as major advancement in CPU power plateaued. Interestingly, there was a different type of processing technology called a GPU (graphical processing unit) that ended up being really good at handling the same kinds of calculations that neural networks require. GPU processing power has exploded over the course of the last few years, and as such, the reality of using neural networks in a practical manner is upon us.
I emphasize the word practical because neural networks and machine learning have been around for a very long time, but the time and resources required to process a face recognition system (for example) was extremely cost prohibitive.
The relative low cost of GPUs has suddenly made powerful machine learning a viable technology for business.
But Machine Learning is still hard
In spite of the growth of excitement around machine learning and AI, it turns out that its actually still very difficult to implement. You need specially-trained machine learning engineers, PhD data scientists mathematicians, and managers who understand the complexities and limits of machine learning. For enterprises, this is still an expensive proposition.
Having trained machine learning models my self, I can tell you that the hardest part of building something useful is gathering the training data. In some cases, you need thousands upon thousands of examples of the different categories of data you want to train the model with. You have to make sure the data is in the right format, and doesn’t have too much variation. Once you’re happy with your data set, you have to make one of hundreds of decisions about what kind of processing you’re going to use to get the results you want. And you’re usually wrong the first time you try it — so be prepared for lots and lots of tests.
6 months later, you’ve got a model you’re comfortable with but now you need to figure out how to deploy and scale it in production. That’s a whole other discipline outside of machine learning that you need to have experts on in order to have something viable.
Enter the cloud companies
Computing in the cloud has its benefits, namely you only pay for what you use. Theoretically, you no longer have to pay for air-conditioning, equipment replacement, and physical space in your data center to power your service. That should be a huge savings. Its the op-ex revolution!
But it turns out there are still issues with the cost of moving content back and forth, the incredible growth in data, vendor lock-in and complexities with security. When the cloud companies started offering machine-learning-as-a-service (MLaas), enterprises ran into a similar barrier. I personally have heard customers complain about the same two things when it comes to integrating cloud-based machine learning; 1) Its too expensive at scale because they charge per API call and 2) I don’t want or can’t have my data leaving my premises.
On premises machine learning
Large enterprises already have numerous applications and data repositories in place today. However, with a broad business focus on digital transformation, the time is right to pursue new data applications and incorporate machine learning and artificial intelligence.
Given the resources available in the cloud from Amazon, Google, Microsoft, and others, it often makes sense to start there first. But companies should keep in mind that the ability to deploy solutions on any cloud or in their own data centers may be an important prerequisite to optimally service their companies long term.
Enterprises need to be masters of their own machine learning, in affect, building an internal MLaaS solution to suit their own needs and drive customer value.
Let me give you an example.
One customer of my company Machine Box works at an enterprise in the healthcare sector. He explored building a machine learning team and discovered it was too expensive. He then looked into IBM Watson, but told us that not only was their text analysis capability less accurate than Machine Box’s, but they were going to be charged a fortune to use it.
What he needs to do is free up his team from manually pouring over thousands of pages of text every day so they can focus on their primary jobs (a perfect use case for machine learning by the way).
What changes the game for enterprise is being able to integrate powerful machine learning tools to save time, create new products, generate revenue, and overall derive tremendous value from the endless data being created, managed, and processed, in a cost-effective way. Remember, this is about ROI above all else.
Enterprise-scale, simple APIs, and unlimited use
The enterprise is going to need all three. Fortunately, that is exactly what Machine Box does.
What is Machine Box?
Machine Box puts state of the art machine learning capabilities into Docker containers so developers like you can easily incorporate natural language processing, facial detection, object recognition, etc. into your own apps very quickly.
The boxes are built for scale, so when your app really takes off just add more boxes horizontally, to infinity and beyond. Oh, and it’s way cheaper than any of the cloud services (and they might be better)… and your data doesn’t leave your infrastructure.
Have a play and let us know what you think.