“The reality is your machine is brilliant. But it needs our language. And without it, it’s just a tin box that lights up,” Bill Gates reportedly said in 1976 as he cut a deal to sell software to a pioneering PC maker. That deal set the stage for the creation of Microsoft and the multibillion dollar PC business that dominated the tech market for a generation. The story illustrates how during a time when everyone was focused on the business of computing hardware, Gates astutely grasped the importance of software, particularly the operating system.
Fast forward to the post-PC era of 2017, and there’s a new technology on the rise: artificial intelligence. However, the AI market today is a ball of confusion, populated by a multiplying horde of algorithms, each with a different purpose, capability and specialty. So, what’s needed to bring order out of this chaos?
The answer is a new kind of operating system, one that corrals the stampede of algorithms, places them into an organized group and orchestrates them to deliver real value that meets customers’ specific AI needs. Without such an operating system, the vast array of brilliant AI technologies could become like Gates’ lighted tin boxes: items of curiosity whose value remains forever untapped.
AI comes of age
Just five years ago, publicly available machine-learning algorithms literally didn’t exist. Today there are more than 5,000 such algorithms commercially available, Veritone estimates.
Technology giants like Microsoft, Google and Amazon are offering these engines as a service with variable pricing based on usage, payable by credit card. Thousands of software engineers and data scientists now see the path to widespread adoption and explosive growth, and more are joining their ranks every day. Because of this, we expect the number of new, specialized machine learning algorithms to increase dramatically over the next five years.
The gold rush is on.
Anarchy in the AI
But like any gold rush, this rapid growth is leading to a boomtown mentality, with inhabitants honoring few rules or formal structures as they pursue their fortunes. Technologies of all shapes and sizes are being lumped together into various AI categories, ranging from highly specialized solutions to the broadest possible approaches.
For example, the transcription segment includes general-purpose solutions for converting speech-to-text, alongside algorithms that are designed for much more narrow uses, such as taking dictation of Spanish phrases or medical terms. All these engines get stamped with the transcription moniker, despite their radical variances in capabilities.
The very notion of AI also is open to widely different interpretations. These days it seems like everything is being called AI, from specialized applications like computer vision systems, to mass-market digital assistants like Apple’s Siri, to robots and drones, to ambitious efforts to apply human-level artificial intelligence to solve major problems, like IBM Watson.
Making sense of cognitive engine technology
For companies hoping to harness the capabilities of cognitive engines, this situation presents major challenges. Companies have specific needs that may change over time. These needs could be best served by one of the multitude of algorithms on the market—or a combination of different algorithms.
For example, imagine a company that’s building a robot. This effort would require one algorithm to make the robot talk, another to allow it to walk, a third to enable it to perceive and understand the world around it—and so on.
The difficulty involved with identifying the right engines to perform each task would be daunting, with a high likelihood that the company wouldn’t pick the best engines for each task.
Get with the system
The best solution to this problem is the aggregation of many specialized cognitive engines into a single ecosystem. Veritone has developed a platform designed to satisfy a wide variety of use cases across any industry.
The Veritone technical stack is analogous to the architecture of a personal computer:
- Cognitive engines are like peripheral devices, each unique but able to plug into the architecture with relative ease via standard APIs.
- Cognitive classes such as transcription, object recognition, and sentiment analysis are the equivalent of device drivers or the software layer that enables specific types of devices, such as keyboards or mice, to work with the system.
- The next level is the most critical: Veritone Conductor, which corresponds to the operating system. Veritone Conductor intelligently orchestrates the entire cognitive ecosystem, matching specialized algorithms to find the right solution for each user’s needs. This is machine learning applied to machine learning.
- Every PC requires storage and memory, and Veritone also has both. Media is stored via cloud providers, and Veritone’s proprietary temporal elastic database (TED) stores the resulting metadata from cognitive processing in a cohesive, time-coded format.
- No PC would be complete without native and third-party apps to make it useful to each individual user. The Veritone platform includes native apps such as the Veritone CMS for ingesting and processing, Veritone Discovery for searching and monitoring, Veritone Collections for sharing and extending, and Veritone Analytics. And like any modern platform, it is extensible to third-party app developers via a robust set of APIs.
Beyond the tin box
Veritone enables organizations big and small to take advantage of the collective power of specially-tuned engines available for a wide range of uses, used individually as point solutions or as customer defined groups. This ecosystem can simplify the process of finding the best engine for whatever task is required, making artificial intelligence accessible, understandable, and useful for a vast variety of tasks.
Furthermore, by combining the capabilities of multiple engines each designed to perform a specific task, the Veritone Platform makes significant strides toward the capabilities of so-called artificial general intelligence (AGI), the still unrealized goal of 60 years of AI research to create a single machine that can perform all cognitive tasks as well as humans can.
Just as Bill Gates helped enable the PC revolution by wielding the power of the operating system, Veritone’s mission is to allow AI to realize its potential as the most important technology of the 21st century.
Tyler Schulze is vice president, strategy & development at Veritone. He serves as general manager for developer partnerships, cognitive engine ecosystem, and media ingestion for the Veritone Platform. Learn more about our platform and join the Veritone developer ecosystem today.