Poxy old computers (was RE: [extropy-chat] SPACE: Spirit Problems)

[Anders Sandberg]

Emlyn O'regan said:
> Give the kids Babbage engines and make 'em work it out.

My computer teacher in highschool almost did this. He was part of the team
who built the second Swedish computer (the BESK in 1953, see
http://www.treinno.se/pers/okq/besk.htm for pictures and
http://www.vethist.idehist.uu.se/english/newsletter/computers.html for a
description) and as an exam we got a copy of the instruction set and had
to write programs for it. The bootstrapping code was actually impressive.

> The bit you leave out when talking about coding on those old machines (ah,
> I
> remember my c64 fondly...) is that you were dealing almost exclusively
> with
> your own code in a simple, controlled, known environment. Pretty much the
> only alien code belonged to the OS, if you used such a bloated beast
> (takes
> up precious bytes).

That is a good point. I still tend to prefer to work with purely my own
code, which of course limits me to heavy numbercrunching neural network
programs. The shift to code re-use and more complex environments is very
important, and I notice that it was only reflected in the computer
education here a couple of years after I got sidetracked into
neuroscience.

So, what would the next step be? I think you are right about automatic
code generation, especially AI-generated code (we already have plenty of
code generators around, but they are fairly predictable and simple). The
step after that is IMHO likely robust adaptive code, where learning is an
integral part of the system and you program by setting the basic template
and learning parameters.

Each step seems to be a letting go. From handwritten machine code to
assembler that is compiled, to high-level languages, to software living in
an operating system, linking with other people's code, computers and data,
and then having other systems generate code and the code change itself.
Each time we give up control and certainty for flexibility and ability.

-- 
Anders Sandberg
http://www.nada.kth.se/~asa
http://www.aleph.se/andart/

The sum of human knowledge sounds nice. But I want more.