The birth of Prolog (1992) [pdf]

When I couldn't find the paper in arxiv, it didn't occur me at first to go back to the beginning and use a general search engine. When I did plug the title into DDG it took me to arxiv, which had the expected abstract and a working link to the pdf. But doing another search in arxiv returned "No Results", even though I had put in the exact title from the abstract I was just looking at.

A little more digging uncovered the fact that the paper in question has gone through 41 revisions, and somewhere along the line the title was changed from "Middle History of Logic Programming: Resolution, Planner, Edinburgh LCF, Prolog and the Japanese Fifth Generation Project" to "Inconsistency Robustness in Logic Programs", which is still rattling my brain a little bit.

edit: The working link I found is https://arxiv.org/abs/0904.3036v22

At MIT criticism of resolution which led to the development of the Planner language of Carl Hewitt and its use by Terry Winograd to implement a large portion of his PhD thesis which was extremely influential at the time... and this was promoted as an alternative to logic and as a procedural representation of knowledge as opposed to a declarative representation of knowledge. So this [muddied] the waters [so to speak] and it created enormous confusion. On the one hand you have resolution [...] logic and its applications to problem-solving, then you have the Planner language which is characterised as a procedural approach alternative to logic and yet it was clear to me that there where [similar to others] the Planner language had similarities to resolution which which were not understood and it was trying to understand that relationship and that similarity that I would say gave rise to logic programmign as I characterised it so far.

The podcast can be found here (I'm not afiliated with it in any way):

https://thesearch.space/

The passage above starts at around 29:53. My apologies for the poor quality of the transliteration.

>> Planner and Shrdlu (Winograd's Phd thesis based on Planner)

Terry Winograd's SHRDLU remains pretty much unsurpassed as a system even today. Wikipedia quotes an excerpt from Winograd's thesis that lists a dialogue between SHRDLU and its user. There's a particular exchange that blows my mind every single time I read it:

    Person: Does the shortest thing the tallest pyramid's support supports support anything green?
    Computer: YES, THE GREEN PYRAMID.

I mean, at that point he's just showing off! I can only imagine him snickering to himself imagining the faces of people reading that.

Still, for me personally resolution has proved to be the right way to go, historically. In particular, I'm prepared to bet that Terry Winograd would have saved himself a lot of pain had he been able to use Prolog's Definite Clause Grammars to write SHRDLU's natural English parser. In Prolog, a grammar, even a natural language grammar, is a theorem and parsing it is a proof, just like for every other program; in every other language a grammar is the input to a program that must be written specially for the task of parsing.