A sidebar to the article “A Guided Tour of Visi On,” published in Byte, issue 6/1983, pp. 276.

Looking at the next generation of software beyond Visi On, and I know that’s very far to look ahead, when do you think voice will become a factor in the user interface? When do you expect to have to deal with that?

Voice input... in some ways I guess you can consider it a factor today.

Because of TI’s announcement of the Professional Computer?

Well, because there are several things that can take very limited amounts of voice but not a general function. I think we’re going to see much further integration over the next 10 years.

Where will the integration increase?

In the underlying data structures, so that you really do have common data. Each program understands all the context at some level or another of every other program’s data.

What stands in the way of that now?

Basically a definition of the underlying data base. It’s a broad subject, not just to understand that something is of a certain type, numeric and alpha and so on, but also to understand all the semantics and all the metadata that go with that type, that say, for example, that this program is a spreadsheet and therefore it has the following kinds of information as far as formulas and recalculation orders; or the program is a word processor and therefore it understands all the semantics about justification and margining and the rest.

If you look ahead somehow, if you could develop an underlying context that understood all of that, there’s no reason why the tools that go on top of it can’t be – boy, am I talking ahead – much smarter than such tools are today. Therefore you really can do what the 1-2-3s and Context MBAs are trying to do: make one product that does everything. What MBA and 1-2-3 do is integrate some functions very problematically and nonextensibly. But if you had something that could understand all of this kind of structure then, you really would approach Smalltalk’s objective. You would have lots of classes and subclasses of data that understand all the manipulations that are done on them, and all you would have to do would be little processes to change that. Therefore you really could be just typing along and in the middle of your text have a spreadsheet, which will recalculate any time and will understand any time you make a change in this model over here that this document must be updated the next time you look at it or print it out. Those things are obviously an extension of where we are trying to go with Visi On, but they aren’t here now.

Voice input and the next levels above that could be much more highly interactive. The problem with voice input is that you’ve got to deal with the office, and in many offices you can’t do a lot of talking. The real technical problem, of course, is just getting the computer to understand voice.

That’s sort of coupled with knowledge-based systems as well. If you start looking at the concepts of artificial intelligence, somewhere along the way we’re going to start seeing voice input in inference processing, and use of artificial intelligence will work its way into these systems. But I see a highly usable system like that as about 10 years away. You are going to see special cases of that all the way along. After all, we’ve seen special cases of windowing systems for 15 years.

When do you think the features of expert systems will begin to appear?

You’re just asking my personal opinion?

Yes.

In the next couple of years, you’re going to see people coming out with expert systems that have some utility. I don’t think you’re going to see the Visicalc of expert systems for five to seven years – that is, an expert system that is so conceptually easy to use that it will be generally adapted by a large group of people. The expert systems that come out sooner will be special-purpose, will involve special training, and will do a very small segment of a problem. But we aren’t close enough there in the technology. That’s just my opinion.

What about progress in languages? Do you think that we’re going to see languages that incorporate features like the new applications? Is Smalltalk going to become popular? Do you think people will ever program with languages? Will languages reach the necessary stage of ease of use, or do you think that’s just fantasy?

I don’t think that people will generally write programs in the sense we’re talking about writing programs today because I don’t think many people easily think at the level of logic it fakes to do a program. Even concepts of iteration are hard for people to understand. I do think that people will be able to develop programs by interacting with systems that are being developed now, maybe inference processing systems.

But I don’t think people will sit down and learn languages even as simple as BASIC or Smalltalk to do much programming on a general basis. The potential world market for the computer/personal computer/desktop workstation is an inverted pyramid – this is an Apple model, by the way – and we’ve seen only the first couple of percent of the tip of that pyramid. To get farther into the pyramid, you have to get farther and farther from the actual need for people to understand computers and programming.