One critical area where IBM's Power Personal Systems Division is providing key technologies to operating-system vendors and ISVs (independent software vendors) is in value-added software extensions that take advantage of the capabilities of the PowerPC processor. The most important of these extensions enable what IBM calls human-centered computing.
In one sense, Power Personal systems are a triumph of technology transfer from IBM and university research labs to IBM products. For example, RISC processors like the PowerPC trace their heritage back to John Cocke's work on the Model 801 at IBM's Thomas J. Watson Research Center (Yorktown Heights, NY) in the mid-1970s. Research for IBM's instruction-set-translator technology comes from its Haifa, Israel, labs, and the notebook design is being done at the company's Yamato facility in Japan. IB
M is also relying heavily on its own research into the areas of speech recognition and AI.
The idea behind human-centered computing is simple: Make computers easier to use by having them conform to how you communicate, and not the other way around. The goal is to have users concentrate on the task at hand as opposed to operating the computer. IBM has married two trends to achieve this goal: Enabling the computer to understand a wider range of natural inputs, and enabling it to use richer data types such as high-quality audio and full-motion video.
Key to human-centered computing is enabling the computer to understand a range of inputs--what Karl Karlson, manager of the Human Centered Conversational Group (Boca Raton, FL), calls multimodal input. For example, when using a spreadsheet, it may be more appropriate to use touch to resize a window or speech to select a range of cells. Unlike today's keyboard-and-mouse interfaces, the human-centered paradigm doesn't constrain you to one method of input
.
Not only does human-centered computing enable a range of inputs--voice recognition, handwriting recognition, pointing, pen input, and touch--but it can make intelligent inferences about the inputs. This type of intelligence is supplied by agents, software modules that use technologies to understand your spoken or written requests and comply with them.
At first, agents will be constrained in their understanding to perhaps command and navigational tasks. As the technology progresses, however, they will develop the ability to perform complex tasks based on your input. One example Karlson uses is telling the computer to "call my wife." The agent would have to infer that call refers to making a telephone connection and equate wife with a specific entry in your address-book database. It may also need to decide, depending on the time of day, whether to place the call to a business or home phone.
Eventually, agents may need to communicate with other agents. Telling your computer to "set up a ma
nagers' meeting for early next week" may result in your agent contacting agents on other systems and negotiating a time and date that can fit into the schedules of the other managers. Such capabilities would require that agents be able to learn about your preferences, and this is a key point to the human-centered paradigm--the computer learns about you, as opposed to you learning about the computer.
The visible manifestation of agents is actors--graphical representations that provide a point of contact between you and the human-centered technologies. Actors are output devices that use speech as well as visible clues to indicate the progress of agents in carrying out your requests.
At Comdex in Las Vegas last November, IBM demonstrated Charlie, a 3-D actor that provided feedback from the system. The actors shipped initially with Power Personal systems will probably be visually simpler than Charlie, both to save processing cycles and to present a less intimidating image to end users. Thus, the fir
st actors may resemble Fred Flintstone more than they do Max Headroom. Some people may find actors a distraction, so IBM will provide an easy way to turn them off.
The Power Personal Systems Division is making the various human-centered technologies available to all operating-system vendors and will create an architecture for ISVs that will let their applications interact with the human-centered actors and agents available on a system. ISVs will also be able to create their own services. At this time, it is too early to tell the level at which different operating-system vendors will support the human-centered technologies, or how they will integrate them into their GUIs. IBM reports, however, that operating-system vendors are generally supporting the human-centered technologies.
Karlson expects that the integration of these technologies into operating systems and applications will be an evolutionary process. The technologies may first appear as a distinct operating-system layer and only graduall
y become more seamlessly integrated with the operating system.
The Power Personal Systems Division has said that the human-centered technologies are a defining element of their Power Personal systems. However, they have not said when specific technologies will appear, except that speech recognition will be included in the first systems.
Illustration: Called variously an actor or a conversational surrogate, Charlie is an output device that provides a focal point for interacting with your system. Shipping versions of Charlie are expected to be less realistic to be less computationally intensive and less intimidating.
Human-Centered Computing
