the IRQ conflict resolution use case from 1994 maps directly to modern infrastructure problems. kubernetes scheduling, network policy evaluation, IAM permission resolution - these are all constraint satisfaction problems where declarative logic would be more maintainable than the imperative spaghetti we currently write.
Open Policy Agent (OPA) essentially reinvented a subset of prolog with Rego for exactly this reason. the difference is Rego is purpose-built and limited, while embedding actual prolog gives you the full power of unification and backtracking. the tradeoff is debuggability - watching prolog backtrack through a complex rule set is painful without good tooling.
I started using Prolog in my self written home automation system over 20 years ago. At first I was using CORBA and I linked ACE/Tao into SWI-Prolog so that Prolog could catch and send CORBA messages. That worked for years but was too annoying to add new message types since a wrapper had to be written for each, plus threading had to be coordinated between C++ and Prolog. Eventually I ditched the CORBA stuff and switched to MQTT, but instead of binding the C++ and Prolog together I found and extended MQTT support for Prolog directly, actually I've mostly replaced the C++ parts of my HA system with Java. The Prolog is pretty nice the way I can now specify predicates for MQTT topic paths, and I use shared topics for scalability. Now all of this is running deployed in k3s.
> We recently installed Gateway multi-media kits on our PCs, but found the installation less than trivial because of conflicts in our interrupt (IRQ) channels. A simple expert system could have helped to resolve those IRQ conflicts. ... The sample program is set up to allow installation of two different devices, a 'Sound Blaster' and a 'Mitsumi CD- ROM'.
This was a real blast from the past. I wonder why more systems today don't have this kind of logic solving built in. Possibly, too many complex behaviours that are not cleanly quantified.
Ironically, once upon a time Prolog and logic programming in general were part of the cutting-edge of AI. There's quite a fascinating history of Japan's fifth-generation computing efforts in the 1980s when Japan focused on logic programming and massively parallel computing. My former manager, who is from Japan, earned his PhD in the 1990s in a topic related to constraint logic programming.
I remember when so-called "expert systems" written in Prolog or LISP were supposed to replace doctors. Then came the (first) AI winter after people realized how unrealistic that was.
Nowadays LLMs are supposed to replace doctors.. and that makes even less sense given that LLMs are error-prone by design. They will hallucinate, you cannot fix that because of their probabilistic nature, yet all the money in the world is thrown at people who preach LLMs will eventually be able to do every human job.
Even now NEC makes some cool massively parallel chips and accelerators that I wish were more mainstream because they look like they'd be fun to play with.
Louise (Patsantzis & Muggleton 2021) is a machine learning system that learns Prolog programs.
Louise is a Meta-Interpretive Learning (MIL) system. MIL (Muggleton et al. 2014), (Muggleton et al. 2015), is a new setting for Inductive Logic Programming (ILP) (Muggleton, 1991). ILP is a form of weakly-supervised machine learning of logic programs from examples of program behaviour (meaning examples of the inputs and outputs of the programs to be learned). Unlike conventional, statistical machine learning algorithms, ILP approaches do not need to see examples of programs to learn new programs and instead rely on background knowledge, a library of pre-existing logic programs that they reuse to compose new programs.
This is what was done by Douglas Lenat from late 1970-s on [1]. He did his work using Lisp, this thing does something close using Prolog.
There's another updated version of that prolog here along with some links, including an archived article from Microsoft Research on how it was (once upon a time) used in Windows NT network configuration:
https://github.com/opless/small-prolog
I often wonder what a Prolog implemented as an Objective-C like extension to C would look like. Since WAM has proper stack and heap IIRC, it might be possible to plug that in through some region-based memory management on C side. Is there some prior art like this?
FWIW (not much), around the time of that article, I reversed this: I used Arity Prolog for a morphological parsing program, with C calls for the bit fiddling (because I needed 64 bits, and the Prolog I was using only handled 16 bit strings).
prolog was doing chain-of-thought reasoning 30 years before LLMs made it cool. the idea of declaring facts and letting the system infer answers is basically what transformer attention does just at massive scale. still think logic programming is underrated for certain domains
Open Policy Agent (OPA) essentially reinvented a subset of prolog with Rego for exactly this reason. the difference is Rego is purpose-built and limited, while embedding actual prolog gives you the full power of unification and backtracking. the tradeoff is debuggability - watching prolog backtrack through a complex rule set is painful without good tooling.
This was a real blast from the past. I wonder why more systems today don't have this kind of logic solving built in. Possibly, too many complex behaviours that are not cleanly quantified.
I see three stories already.
Nowadays LLMs are supposed to replace doctors.. and that makes even less sense given that LLMs are error-prone by design. They will hallucinate, you cannot fix that because of their probabilistic nature, yet all the money in the world is thrown at people who preach LLMs will eventually be able to do every human job.
The second AI winter cannot come soon enough.
[1] https://en.wikipedia.org/wiki/Eurisko
In case you weren't aware, people are using Prolog with LLMs;
https://news.ycombinator.com/item?id=42039527
https://news.ycombinator.com/item?id=45712934
Disclaimer: I'm no expert on LLMs.