3 ms·

There's something quite illuminating with this first "horror", where they basically say "it's OK to report wrong answers, because you can check the answers". I

by rtpg 5mo ago

There's something quite illuminating with this first "horror", where they basically say "it's OK to report wrong answers, because you can check the answers". I don't think I've ever felt like it's OK for my program to provide a list of answers where some are right and some are wrong, but reading this... and generally believing in P != NP.... maybe that's a decent way of looking at some stuff!

DonHopkins 5mo ago
Sometimes the Biorhythm program on my Apple ][ failed to produce correct answers. But it sure was great for impressing cool hippie chicks. https://www.youtube.com/watch?v=jYoY1cwAd90 https://www.youtube.com/watch?v=jYoY1cwAd90
hedora 5mo ago
The article server is offline, but I assume they found out that prolog rule evaluation depends on the order the rules are presented in the program. If so, the language they thought they were using (and that they should actually use) is datalog, not prolog. Datalog has declarative semantics: All facts that are derivable from the base database and the rules will be derived by the interpreter, and it will not add extra hallucinated facts. If that's not true, it's a bug in the runtime, not in the language.
- cbarrick 5mo ago
  www.metalevel.at is run by Prolog legend Markus Triska, author of CLP(FD)/CLP(Z). So it's not that they "discovered" anything about Prolog; they already knew the language inside out. This article explains how to appropriately use Prolog declaratively and with full generality.
Zarathustra30 5mo ago
I've actually run into this in the wild, with regards to sales forecasting. A program we were using returned zero if the error bars on a forecast were over 100%. For example, selling somewhere between 1 and 7 units, but averaging 3. Returning 3 was "wrong", but infinitely more correct than retuning 0.
- taneq 5mo ago
  That sounds like zero was meant to be treated as a special error value? If so then “bad forecast” is more correct than “everything’s fine here, situation normal”, no? Of course this is far from the best way to do this.
cwillu 5mo ago
iirc, shor's algorithm for factoring relies on this.

by rtpg 5mo ago

DonHopkins 5mo ago

Sometimes the Biorhythm program on my Apple ][ failed to produce correct answers. But it sure was great for impressing cool hippie chicks. https://www.youtube.com/watch?v=jYoY1cwAd90 https://www.youtube.com/watch?v=jYoY1cwAd90

hedora 5mo ago

The article server is offline, but I assume they found out that prolog rule evaluation depends on the order the rules are presented in the program. If so, the language they thought they were using (and that they should actually use) is datalog, not prolog. Datalog has declarative semantics: All facts that are derivable from the base database and the rules will be derived by the interpreter, and it will not add extra hallucinated facts. If that's not true, it's a bug in the runtime, not in the language.

cbarrick 5mo ago
www.metalevel.at is run by Prolog legend Markus Triska, author of CLP(FD)/CLP(Z). So it's not that they "discovered" anything about Prolog; they already knew the language inside out. This article explains how to appropriately use Prolog declaratively and with full generality.

Zarathustra30 5mo ago

I've actually run into this in the wild, with regards to sales forecasting. A program we were using returned zero if the error bars on a forecast were over 100%. For example, selling somewhere between 1 and 7 units, but averaging 3. Returning 3 was "wrong", but infinitely more correct than retuning 0.

taneq 5mo ago
That sounds like zero was meant to be treated as a special error value? If so then “bad forecast” is more correct than “everything’s fine here, situation normal”, no? Of course this is far from the best way to do this.

cwillu 5mo ago

iirc, shor's algorithm for factoring relies on this.