+ All Categories
Home > Documents > Artificial intelligence in the man/machine interface: Useful tools are emerging from AI research

Artificial intelligence in the man/machine interface: Useful tools are emerging from AI research

Date post: 25-Aug-2016
Category:
Upload: kathy-lawrence
View: 216 times
Download: 0 times
Share this document with a friend
3

Click here to load reader

Transcript
Page 1: Artificial intelligence in the man/machine interface: Useful tools are emerging from AI research

Artificial intelligence in the man/machine

interface Useful tools are emerging from AI research

by KATHY LAWRENCE

Abstract: Artificial intelligence programs may be too expensive for many organizations, but the tools coming out of AI development are also being marketed, such as languages and environments. A version of LISP has been used to produce a manlmachine interface development environment which can be used for design and training.

Keywords: data processing, programming languages, software techniques, artificial intelligence.

Kathy Lawrence is editor of Data Processing.

Artificial Intelligence Ltd, Intelligence House, Merton Rd, Watford, Herts WDl 7BY, UK.

T here are few organizations will- ing to pay vast amounts for the benefit of artificial intelligence

(AI) programs, according to a special- ist UK AI company. However, many

techniques are emerging from AI research carried out over the last 20 years, and a market is developing for these tools, says David Catton, a director of Artificial Intelligence Ltd. Current users of AI tools are those involved in large research projects for the Alvey and Esprit programmes, academics, hi-tech industries and major software houses. Research laboratories of major companies are also buying tools, in some cases in

244 0011-684x186/050244-03$03.00 0 1986 Butterworth &Co (Publishers) Ltd. data processing

Page 2: Artificial intelligence in the man/machine interface: Useful tools are emerging from AI research

systems

case they might be missing out on something competitors are doing. So far though, says Catton, there are few people who know how to exploit AI

techniques.

Getting started in AI

Catton offers advice to those who are thinking of starting in AI. This in- cludes a careful look at the costs involved, which will be much in excess of the hardware and software costs. He is much in favour of starting

small, with a demonstrator project. A feasibility study is important, as is access to a knowledge engineer. One option is to subcontract the work to a software house with AI experience and let inhouse staff learn from the

project. Catton is not keen on the current

range of shell systems available, which he argues are built around one particular method of problem resolu- tion, which may well not fit the

problem in question.

Artificial Intelligence Ltd therefore concentrates on supplying general en- vironments, such as the LISP language and the workstations to run the soft- ware. These are still expensive

machines, but Catton believes that with the arrival of the IBM PC RT the personal computer will begin to be useful in the AI environment.

Benefits of AI techniques

So why use artificial intelligence tech- niques? Many of the systems in ques- tion can be produced using conven- tional programming techniques, but they may be inelegant, expensive and time-consuming. An important differ- ence between programming the con- ventional way and programming the AI way is flexibility. In standard applications the requirement is known, and the program can be developed in a structured manner. In AI the exact end-requirement is not always known, and the programmer must have the freedom to experiment and change direction during develop- ment. This is programming under uncertainty. An example is the layout of a newspaper page, where the layout artist obeys a few guidelines - such as advertisements being placed around the edge - to build the page, and then adjusts it until it is accept- able.

The AI languages and tools allow

Example of lhe Trillium screen with windowing and a simulated control panel

~0128 no 5 june 1986

this sort of flexibility in design to take place in programming. In LISP, for example, decisions are deferred until run time, whereas in conventional programming a commitment is made on specification, on compilation, on

linking and at run time.

Man/machine interface development environment

These artificial intelligence tools and techniques can be used to develop

other tools. An example of this is a man/machine interface development tool called Trillium. Trillium was

originally developed at Xerox Palo Alto Research Centre (PARC) to speed up the design cycle for the

man/machine interface of new ranges of photocopiers. Artificial Intelligence

Ltd has developed it into a general tool which is now being sold and supported by the company. Trillium is written in INTERLISP-D and there-

fore uses the language’s features to make the tool more flexible.

An example of the use of such an environment might be the design of a control panel for guiding unmanned vehicles. The user interface is likely to consist of a screen, plus controls such as push-buttons, knobs, dials and so on. Trillium can be used to design and simulate this control panel.

The features of the Xerox work- station, on which Trillium runs, allow windowing. So in this example there may be one window showing a design of the control panel and another

showing the movement of the vehicle. If the graphic representation of a dial is turned to the right (using a mouse to move the item on the screen in this case) the effect of this on the move- ment of the vehicle can be viewed in the second window.

This can look like a very expensive version of an arcade game, but is a great deal more sophisticated. For example, the control panel can be edited in mid-simulation, and a joy- stick might be put in the place of the dial. The joystick is chosen from a

245

Page 3: Artificial intelligence in the man/machine interface: Useful tools are emerging from AI research

vast range of pop-up menus which appear in windows on the screen as requested. The joystick may be re- quired to do the same job as the previous dial, but provides a different

interface for the user.

Developing the interface with Trillium

The company recommends a number of stages in the development of a man/machine interface. The two major phases are to analyse the inter- action between user and system and then to implement it, using the modelling environment of Trillium.

In the analysis phase, the tasks of the user and the relationship between those tasks and the functions of the system must be defined. Each task must be defined in terms of require-

ments for the interface. Once the definition has taken place,

the interface can then be modelled using the environment.

The first action is to break every- thing down into items. A graphic symbol on the screen such as a circle, is an item. A circle by itself is a ‘primitive’ item. A ‘composite’ item

might be a rotary knob for display and for control. It is a combination of primitives.

The primitives either appear on the screen and are known as ‘explicit’ or exist for functional purposes and are not seen by the user, in which case they are ‘implicit’ primitives. Explicit

primitives are:

artwork, where a graphic or piece of text appears purely as inform- ation on the screen, displayers, where the system is providing changing information to the user e.g. height, speed, fuel available in an aircraft, controllers, which the user manip- ulates to pass information to the system e.g. dials, joysticks, push- buttons.

The implicit primitives, which do not

246

r _-_--- I I I ; irillbmengine , ]

----------

Figure 1. Items defined within the manlmachine interface and their function within TrSum

appear to the user but carry out

important functions are:

initialisers, which are used to set up the system before it can be used inhibitors, which stop certain actions taking place if others have

occurred sensors, which can detect the loca- tion of a pointing device such as a

mouse implicators, which allow data from one activity to be passed onto another groups of those above.

All of these item types are used within the system of Trillium (see Figure 1). The user reads displayers on the screen and uses controllers to pass information back to the system. With- in the system representation are the initialisers, inhibitors, implicators and sensors. These are all driven by the Trillium engine.

Having given a type to each item, the next step is to make a choice from

the repertoire for each type. For

example, a displayer could be text or artwork chosen from a wide range of fonts, backgrounds, textures and line

features. As there in no need to give a type to variable in LISP the system itself will sort out what is text and what has numerical values. All the display values are derived from the cell values in the engine which drives

the system. Selections are made via a number of menus.

Many items will have action forms to fill in. In the previous example, a number of actions must occur if a dial is turned, including messages to be sent to the user.

Once items have been defined and set up the interface can then be implemented.

The Trillium environment can be used to design interfaces. Alternative- ly it can be used as a simulator for training purposes.

Trillium will cost about 53000 for the full environment, plus the cost of the workstation. I7

data processing


Recommended