DecisionBrain is actively investigating the topic of explaining Combinatorial Optimization results. Drawing from principles in both Social Sciences and Artificial Intelligence, they highlighted two key types of explanations, contrastive and counterfactual explanations, and discussed their relevance in decision-support systems. Link
This year DeclarativeAI (that includes DecisionCAMP as a co-event) will run the 19th International Rule Challenge, fostering friendly competition among innovative rule-oriented tools, prototypes, and applications tailored to research, industry, and government. Participants are invited to showcase their solutions to self-defined challenges and to propose open challenges for the community to address. Link
Ron Ross again brought up the question of “Exceptions to Business Rules” to light (https://lnkd.in/eyemPWF2). Ron defines an exception to the rules as a foreseen, explicit set of circumstances in which different-than-normal guidance is to be followed. He gave an example: seeing-eye dogs as an (explicit) exception to dogs not being allowed in a hospital. One comment says: “I heard that there are no exceptions to the BRs. I heard that there are only other BRs.”
This discussion is still as important as it was years ago when the vendors of Business Rules systems considered a more generic problem of Resolving Conflicts among Business Rules. In 2014, two BR vendors, Drools and OpenRules, independently addressed the problem of the diagnosis and resolution of business rule conflicts using Defeasible Logic. Link
Clive Spenser pointed to a very interesting Oct-2024 article by Martin Milani that makes the case for coupling deep learning with expert systems.
“In the next generation of AI, advanced expert systems will serve as the core “intelligence,” acting as the primary engine for decision-making. While deep learning systems excel at recognizing patterns, these perceptual learning methods represent a subservient and lower form of intelligence. Just as in humans, where perception informs but does not govern complex reasoning, deep learning systems will be subordinate to expert systems that provide structured, logical, and complex decision-making.” Link
Continue reading
Stéphane Dalbera posted today: “This quote elegantly describes the true art of software development. It’s not enough to dream up bold strategies or to architect utopias on whiteboards. The challenge lies in translating those strategic desires into tactical realities – code that runs, systems that scale, domains that make sense.
Continue readingCyc was a “child” of Doug Lenat, who devoted 40 years to trying to make AI more human. David Reed just blogged about it: “Is Cyc dead, as the obituary here claims? Did Deep Neural Nets kill it? I think they both have been part of a process that is killing AI by focusing only on ‘Representation of knowledge’. My own view is that the neural net training approach used in all LLMs today is not the way to get ‘intelligence’. Nor is the pure Cyc-style approach. Both focus only on manipulating Representation, and in particular representations amenable to simple computation machines (neural networks or symbolic logic). Intelligence, such as what we see in humans, which I call Cognition, isn’t primarily contained within the skull.” Link
P.S. Doug Lenat was our keynote speaker at DecisionCAMP in 2012 – see https://dmcommunity.org/2023/09/06/in-memory-of-doug-lenat/.
Stephane Dalbera: “The fantasy of programming in natural language didn’t emerge with Large Language Models; it’s been a topic of discussion for decades. It’s interesting to note that Edsger W. Dijkstra’s 1978 critique remains, in many aspects, highly relevant today.
In his paper titled “On the foolishness of ‘natural language programming’,” Dijkstra argued that the inherent ambiguities and imprecision of natural languages make them unsuitable for programming purposes. He emphasized the value of formal symbolisms in programming, stating that they are an effective tool for ruling out various forms of nonsense that are almost impossible to avoid when using natural language.
Dijkstra’s insights continue to resonate in contemporary discussions about the feasibility and desirability of natural language programming, especially in the context of advancements in AI and machine learning.
At a glance: The future will be unambiguous, or it will not be.” Link
An analytical firm assigns different cases to its analysts using the following rules:
Given a list of analysts with their current workload and a list of new cases, you need to help the firm to decide which analysts should be assigned to these cases. If there are multiple options, the firm prefers to minimize the overqualification when analysts are assigned to the cases below their levels. Can you create a decision service capable of addressing this and similar problems? Link
A team of researchers from UC Berkeley and UC San Francisco has unlocked a way to restore naturalistic speech for people with severe paralysis. This work solves the long-standing challenge of latency in speech neuroprostheses, the time lag between when a subject attempts to speak and when sound is produced. Using recent advances in artificial intelligence-based modeling, the researchers developed a streaming method that synthesizes brain signals into audible speech in near-real time. Link
SolverMax attempted to use an LLM tool to create an entire, non-trivial optimization model, with the AI doing all the programming. They asked Copilot to “Design a crop rotation optimization model“. Copilot responds with a comprehensive specification for this situation, but the entire process was not straightforward. It is described at https://www.solvermax.com/blog/can-ai-code-an-entire-optimization-model. Here is the conclusion:
Continue reading
Decision Management Community 
You must be logged in to post a comment.