We are seeing the beginning of the new artificial intelligence economy. This has many parallels to the infrastructure-as-a-service wave led by Amazon Web Services (AWS), which provided the world with access to highly-scalable compute capacity. AI technologies are being exposed as core infrastructure via the cloud, enabling companies to build smarter applications and services.
If you think you aren’t already a part of the AI economy, think again. Most of us are already participating through our interaction with popular applications and services. For example, Google Maps uses AI technology to better understand Street View images to give more accurate directions; and both Siri and Google Now use a combination of speech recognition, language understanding, and predictive modeling to act as digital personal assistants.
So the big question is: why now? Historically, AI technologies have been limited by a lack of data, insufficient compute capability, and poor algorithms. We’re now witnessing the convergence of three major forces: ready access to massive data, highly scalable on-demand compute capability, and a number of core algorithmic breakthroughs that enable us to better train robust AI systems. This is a perfect storm that has resulted in significant advances in computers’ ability to understand text, images, video, and speech.
Companies increasingly rely on AI technologies to bring new products to market and to stay competitive in a rapidly evolving world. However, this stuff is hard. It takes a major capital investment in infrastructure, data, and research talent to build AI systems from scratch. We’re also witnessing a land grab as big companies acquire talent in this space. So where does this leave us? How can companies big and small participate in this new world?
Let’s look back to the late 1990s: at that time, it was extremely costly to start an online business. Companies were required to invest millions of dollars to build out a robust web infrastructure. Amazon changed the game by enabling companies to rent infrastructure by the hour, drastically reducing the cost and time required to get a company off the ground and compete with well-funded competitors.
Similar to what we’ve seen with AWS transforming compute infrastructure into a utility-like shared resource, early pioneers in the AI space are now offering access to state-of-the-art systems and algorithms via the cloud. This enables companies to incorporate advanced AI capabilities into their products without incurring the time, risk, and capital investment required to build these systems entirely on their own.
How does a company integrate with cloud-based technologies? Early SaaS providers focused almost entirely on application delivery, but today’s businesses require more than applications, they need specific features and capabilities that can be directly incorporated into their systems and products. This is best accomplished through web-based APIs, which provide on-demand access to software algorithms in a programmatic fashion. For example, companies like AlchemyAPI provide APIs that enable deep understanding of text and images. AT&T provides APIs that can transcribe speech to text. IBM, through its Watson platform, allows computers to answer human-like questions.
What’s on the horizon? We can expect to see more of these advanced cognitive systems becoming available in the near future, enabling entirely new business models and product categories. However, there are a number of challenges. Software developers have decades of experience leveraging structured data for useful purposes, but are just getting started with unstructured data like text and images. AI infrastructure providers face the challenge of making these capabilities easily accessible, understandable, and reliable.
I don’t get up in the morning worrying if the electricity is going to be available, and a company executive should not be spending their days worrying if their cognitive computing infrastructure is operating effectively. These technologies should be as easy to use and reliable as flipping on a light switch, and that’s the challenge for today’s AI infrastructure providers.
Photo courtesy flickr / telstar
About the Author
Elliot Turner is a deep learning expert and the founder and CEO of AlchemyAPI, a cloud-based platform used by over 40,000 developers for creating big unstructured data applications that rely on the company’s innovations in natural language understanding, computer vision and question answering. His work has been cited by MIT Media Lab, IIT, Microsoft Research, Carnegie Mellon. You can find him at @eturner303