First from left is Marvin Minsky, second from left is Alan Kay, then John Perry Barlow and Gloria Minsky.
Question: How would you interpret Marvin Minsky's idea that βComputer Science already has a grammar. What she needs is literature.
Alan Kay: The most interesting aspect of the record (including comments) is that there is no historical mention of this idea anywhere. In fact, more than 50 years ago in the 60s, there was a lot of talk about this and, as far as I remember, several articles.
I first heard about this idea from Bob Burton, in 1967 in graduate school, when he told me that this idea was part of Donald Knuth's motivation when he wrote The Art of Programming, the chapters of which were already circulating. One of Bob's main questions at the time was about "programming languages ββdesigned to be read by humans as well as machines." And that was the main motivation for COBOL design parts in the early 60s. And, perhaps more importantly in the context of our topic, this idea is seen in the very early and rather beautifully designed interactive language JOSS (mostly by Cliff Shaw).
As Frank Smith observed, literature begins with ideas worth discussing and writing down; it often partially generates representations and extends existing languages ββand forms; it leads to new ideas about reading and writing; and finally to new ideas that were not part of the original motive.
Part of the idea of ββ"literaturizing" is reading, writing, and referencing other articles that may be of interest. For example, Marvin Minsky's Turing Award lecture begins by saying that .
He meant that the most important thing in computing is the meaning, and how it can be considered and represented, in contrast to one of the big topics of the 60s about how to analyze programming and natural languages. For him, the most interesting thing about the thesis of graduate student Terry Winograd might be that at the same time it was not very correct in terms of English grammar (it was very good), but that it could give meaning to what was said and could justify what was said using this value. (This is a throwback to what Ken reports on Marvin's blog).
A parallel way of looking at "ubiquitously learned language". A lot can be done without changing the language or even adding a dictionary. It's just like with mathematical symbols and syntax it's very easy to write a formula. That's part of what Marvin is leading up to. It's funny that the Turing machine in Marvin's Computing: Finite and Infinite Machines (one of my favorite books) is a pretty typical 1-instruction computer (add 1 to a register and subtract 0 from the register and branch to the new instruction if the register is less than XNUMX - there are many options.)
This is a common programming language, but pay attention to the pitfalls. A reasonable solution to "ubiquitous learning" also needs to have certain kinds of expressive power that are likely to take more time to learn.
Don's interest in so-called "literate programming" led to the creation of an authoring system (historically called the WEB) that would allow Don to explain the actual program that was being written, and which included many features to extract parts of the program for human study. The idea was that a WEB document was a program, and the compiler could extract the compilable and executable parts from it.
Another early innovation was the idea of ββdynamic media, which was a popular idea in the late 60s and for many of us was an important part of interactive computing on a PC. One of the several motives for this idea was to have something like "Newton's Principia" where the "mathematics" was dynamic and could be run and tied to graphics etc. This was part of the motive to promote the idea of ββthe Dynabook in 1968 year. One of the terms that began to be used then was "active essay", where the kinds of presentation and argument that one would expect in an essay are enhanced by the fact that the interactive program is one of the many media types for a new type of document.
Some very good examples were made in Hypercard by Ted Kyler himself in the late 80s and early 90s. Hypercard wasn't directly configured for this - scripts weren't media objects for cards, but you could do some work and get scripts to show on cards and also make them interactive. A particularly provocative example was "Weasel", which was an active essay explaining part of Richard Dawkins' book "Blind Watchmaker", allowing the reader to experiment with a framework that used a kind of breeding process to find targeted sentences.
It's worth considering that while Hypercard was a near-perfect fit for the emerging Internetβand its widespread adoption in the early 90sβthe people who built the Internet chose not to embrace it, nor Engelbart's larger early ideas. And Apple, which had a lot of ARPA/Parc people in its research wing, refused to listen to them about the importance of the internet and how Hypercard would be great to start using a symmetrical read/write system. Apple refused to make a browser at a time when a really good browser would have been a big deal, and probably played a huge role in what the "public face" of the Internet turned out to be.
If we advance a few years, we find the sheer absurdity - almost obscenity even - of a web browser without a real development system (think how stupid wiki development should have even worked), and as one of many simple examples, a Wikipedia article like LOGO , which works on a computer, but does not allow the reader of the article to try programming LOGO from the article. This meant that what was important to computers was blocked from users in defense of different implementations of old media.
It's worth considering that Wikipedia was and is the main genre for thinking, inventing, implementing and writing the "computing literature" that is needed (and this certainly relates to both reading and writing many kinds of multimedia, including programming).
It's even more worth considering that I can't write a program here in this Quora answer - in 2017! β this will help to show what exactly I am trying to explain, despite the huge computer power that underlies this weak idea of ββinteractive media. The important question is βwhat happened?β is completely overlooked here.
To get an idea of ββthe problem, here's a 1978 system that we partly resurrected a few years ago to pay tribute to Ted Nelson and partly for fun.
(Please look here at 2:15)
The whole system is an early attempt at what I am now talking about over 40 years ago.
A striking example can be seen at 9:06.
Apart from "dynamic objects", one of the key considerations here is that "views" - the media that is visible on the page - can be rendered uniformly and independently of their content (we call them "models"). Everything is a "window" (some of them have explicit borders, and some don't display their borders). All of them are compiled on the project page. Another understanding was that since you have to compose and combine some things, make sure that everything is composable and compositional.
I think inexperienced users can be forgiven for not being able to critique bad designs. But programmers who make interactive media for users, and who don't care about learning about media and design, especially from the history of their own field, shouldn't get away with it so easily and shouldn't be rewarded for being they are "weaker".
Finally, a field without real literature is almost equivalent to the fact that this field is not a field. Literature is a way to keep great ideas alive in a new genre, and in the present and future thinking in the field. This is of course not present in the calculations to any useful extent. Like pop culture, computing is still most interested in what can be done without a lot of learning, and where the execution is more important than the consequences of the results. Literature is one of the mediums where one can move from the simple and immediate to the larger and more important.
We need it!
About GoTo School
- June 17-30, July 15-28, August 12-25
Source: habr.com