After discovering Don Crabb’s thoughts on AI, which I wrote about yesterday here, I did a little more digging in the Internet Archive. This turned up an incredible treasure trove of files collected by David Craig called Apple Lisa Document and Media Collection, which included a photocopy of Jef Raskin’s interview in the amazing book by Susan Lammers titled Programmers at Work, which can be checked out for reading on archive.org here or online at this website created by Lammers).
Jef Raskin, who wrote the user manual for the Apple II and founded the team that would go on to launch the Macintosh computer among other accomplishments, was an important figure in the first phase of the personal computer industry. Toward the end of his interview in Lammers’ book, she asks him about AI:
INTERVIEWER: What do you feel artificial-intelligence programs can contribute to society?
RASKIN: Artificial intelligence teaches us a lot about ourselves and about knowledge. Any reasonable artificial-intelligence program will not fit on a very inexpensive machine, at least not these days.
Real artificial intelligence is something like religion. People used to say that just above the sky were heaven and angels. Then you get a rocket ship out there, and now you know that’s not true. So they change their tune. As soon as you accomplish something, it is no longer artificial intelligence.
At one point, it was thought that chess-playing programs encompassed artificial intelligence. When I was a graduate student, you could get a Ph.D. in artificial intelligence by learning to program chess. Now you can buy a chess player for $29.95 and nobody calls it artificial intelligence. It’s just a little algorithm that plays chess.
First, there’s a problem of definition. Then it gets more complicated. People say that programs should understand natural language, but our utterances are too inexact for a computer, or anybody, to figure out what is meant to be done; that’s why we have programming languages. If anyone’s ever worked from a spec prepared in English, they know that you can’t write a program from it because it’s not exact. So if human beings can’t do it, there is almost no way we can expect to make a machine do that kind of thing. When you’re dealing with so-called artificial-intelligence programs, the computers have got to learn a vocabulary. Let’s say you have five commands and you want the machine to understand any possible English equivalent to them. But it won’t understand any English equivalent: One person might say, “Get employee number,” while an Englishman might say, “Would you be so kind as to locate the numerical designation for our employee.’ That’s exactly the complaint AI people are trying to solve.
A lot of the promise of artificial intelligence is misunderstood. What artificial intelligence has already taught us about the nature of languages is wonderful. So, do I think artificial intelligence is worthwhile? Absolutely. Do I think it’s going to turn out great products? A few. Do I think it’s going to fulfill the promise that you read about in the popular press? Not at all. Will I be putting a lot of money into artificial intelligence? Nope (qtd. in Lammers 243-244).
Lammers, Susan. “Jef Raskin.” Programmers at Work: Interviews with 19 Programmers Who Shaped the Computer Industry. Microsoft Press, 1989, pp. 227-245.
What he said has some resonance today. There seems to be the same kind of effect in computers that we see in other fields. For lack of a better phrase, it’s the “so, what have you done lately?” question. Once one hurdle is accomplished, its importance or significance gets erased by the passage of time and people’s attention. Deep Blue beat Kasparov at chess? Great, what’s next? AlphaGo beat Lee Sedol at Go? Okay, what’s next? ChatGPT can do your homework? Super, what’s next? With each milestone, the preceding success seems diminished and becomes the $29.95 chess player that Raskin refers to above.
However, as AI’s capabilities increase, it seems to be edging further toward ubiquity. It’s already ever present in many aspects of our lives, such as business, finance, advertising, and photography, that we are not necessarily cognizant of or paying attention to. Now, it’s creeping into computer and smartphone operating systems (similar to Don Crabb’s observations that I wrote about yesterday) and some of the software that we use for daily productivity (email, word processing, and integrated development environments for programming). Perhaps its the eventual ubiquity of AI that will make it feel mundane instead of a radical technological development as imagined in the heady cyberpunk era represented most clearly by William Gibson’s Neuromancer (1984).
But, there’s something else that Raskin talks about in his interview that has some relevance to AI. After he left Apple when Steve Jobs took over the Macintosh project, he founded Information Appliance, Inc. to build and market an add-on card for the Apple IIe called the SwyftCard. This card contained a ROM for an all-in-one piece of software that contained word processing, communication, calculation, printing, and programming capabilities. He explains:
Watch this. There is no disk in the drive, and I want to type a message, “Remember to bring home some milk.” How do you like that? I turn it on and start typing. No need for commands, no insert, no getting to the editor, I can just start typing.
Now I want to print the message and put it in my pocket, so I can use it later. I press a single key, and it prints. Isn’t that convenient . . . .
We can do calculations easily. Before, whenever I was using the word processor and wanted to do a calculation, I’d get out my pocket calculator and have to use a separate calculating program, or get up SideKick; on the Mac, you call up the calculator and paste it into your document. We also have telecommunications capability.
INTERVIEWER: All in the same program?
RASKIN: Sure. There is no difference between all the applications. What’s a word processor? You use it to generate text, move it around, change it if you make a mistake, and find things. What’s a telecommunications package? You use it to generate text, or receive text generated by someone else. Instead of it coming in from a keyboard or out from a printer, it comes in or out over a telephone line. And what’s a calculator? You use it to generate numbers, which are just text, and the answer should come back into your text. So, one day it dawned on me, if these applications do the same thing, why not have one little program that does them all?
INTERVIEWER: Well, what is this product you’ve developed to cover all of these features?
{Raskin holds up a simple card.]
RASKIN: It’s called a SwyftCard (qtd. in Lammers 233).
Lammers, Susan. “Jef Raskin.” Programmers at Work: Interviews with 19 Programmers Who Shaped the Computer Industry. Microsoft Press, 1989, pp. 227-245.
It seems to me that we’re heading toward a great collapsing of software into generative AI. As large language models learn more with increasing amounts of training data, they reveal new capabilities that emerge from the resulting trained models. Will we type and eventually talk to our computers to tell it what we want to accomplish without having to worry about having x, y, or z programs installed because the AI can do those things in an all-in-one fashion as the Nth degree of Raskin’s SwyftCard? Time, of course, will tell.