Embedded Processing

There is a lot of energy being expended in the electronics industry to identify, create, and market the next killer application. The term killer application has been used to refer to a capability that draws in large numbers of new consumers for a particular system or platform. The iPhone from Apple and the Wii from Nintendo are two platforms that benefit from killer applications that rely on the system recognizing the user’s gestures instead of having to use conventional button-based and keyboard interfaces. My article about recognizing gesture interfaces gives an overview of human-machine interfaces.

I would like to share some thoughts about the future of killer applications and the technologies that will make them possible. It seems to me that when killer applications emerge, they transform some task that was too hard, too expensive, or impossible to do into a task that is feasible to perform.

The Visicalc and Lotus123 spreadsheet programs are an example of one of the first killer applications for computers. The spreadsheet program dramatically transformed how people performed analysis tasks. It made it much easier to setup an analysis. It automated the verification of the analysis results and made repeatability of the results feasible to a wider portion of the market. It vastly shortened the time needed to perform iterations of an analysis that enabled people to perform more exploration of options and what-if analysis than ever before.

The cell phone is its own killer application. It enables people to communicate with other people with less time latency and dependence on geographical location than using land lines. Many contemporary killer application candidates are piggy-backing on the cell phone platform. One of the key capabilities of the iPhone was its gesture interface that performed better than earlier attempts at touch and gesture interfaces in a cell phone platform, such as the IBM Simon that was launched 14 years prior to the iPhone.

The focus on gesture interfaces is logical because the easier it is for people to learn how to use a device, the lower the barrier to entry for people to take advantage of whatever capabilities the platform can deliver. However, gestures are not innate or universal – this means they must be taught and learned. As the market continues to evolve gesture interfaces, the room for improvement and killer applications based on solely interface improvements faces an eventual diminishing return on investment.

This leads me to where I think the next wave of killer application enabling technology will come from. The “Sixth Sense” and the “Siftables” projects are two exciting academic explorations that I think show great promise. Check out the videos for both the sixth sense and siftable projects at TED. However, both project presentations gloss over the underlying technologies that make these exciting ideas possible – that the machine must expand its awareness beyond the human-machine interface and interact with or sense the real world.

In the case of the sixth sense project, it relies on the system being able to recognize the user’s gestures to know what the user wants the system to do – such as take a photo. But the real magic occurs if the system can also recognize objects placed in its field of view and autonomously construct queries for online searches. I am very excited about these types of projects; however, this technology will require much more than current image recognition or marker technology to work in general purpose environments. I suspect that this technology will follow an analogous evolutionary path as the IBM Simon and the Apple iPhone. One of the challenges facing the Simon was that it required too much verification from the user. This frequent verification step can be intrusive to the user’s thought process. The iPhone was able to incorporate enough internal intelligence to remove much of the verification that made the Simon difficult to widely adopt.

The siftables project adds a tangible aspect to working with many small machines that make up a whole system. Like the sixth sense project, the magic is in how each of the small machines recognizes their relationship to the world and each other. This project excites me on another level though because it may also provide a platform to explore the future of multi-processing systems. I am currently finishing up a June article that tries to propose a taxonomy for multicore/processing configurations. I will complement the article with a post here about some of the things that I could not include in the final article.

I see the future of applications involving more machine recognition and intelligence. This will involve much more work in vision and audio recognition as well as context-based prediction, verification, and refinement. The scope of these technologies though will dwarf the consumer market-level killer applications of today because they will find their way into the myriad of embedded systems that consumers use all the time but are not even aware of. I believe increasingly autonomous systems are the direction of the next generation of killer applications, and for those to happen, we as a cross-disciplinary industry, will need to improve the machine’s ability to reliably predict, sense, verify, interact with, and recognize all of the things in its environment.