LPaaS as Micro-Intelligence: Enhancing IoT with Symbolic Reasoning

Abstract

In the era of Big Data and IoT, successful systems have to be designed to discover, store, process, learn, analyse, and predict from a massive amount of data-in short, they have to behave intelligently. Despite the success of non-symbolic techniques such as deep learning, symbolic approaches to machine intelligence still have a role to play in order to achieve key properties such as observability, explainability, and accountability. In this paper we focus on logic programming (LP), and advocate its role as a provider of symbolic reasoning capabilities in IoT scenarios, suitably complementing non-symbolic ones. In particular, we show how its re-interpretation in terms of LPaaS (Logic Programming as a Service) can work as an enabling technology for distributed situated intelligence. A possible example of hybrid reasoning-where symbolic and non-symbolic techniques fruitfully combine to produce intelligent behaviour-is presented, demonstrating how LPaaS could work in a smart energy grid scenario.

Fifty Years of Prolog and Beyond

“…Novel opportunities may then arise by bridging Prolog with other well-established research areas. This is, for instance, what happened with the Multi-Agent System community, where Prolog and LP have been extensively exploited in the last decades as the technological or conceptual basis for tens of agent-oriented technologies (Calegari et al, 2021). Similarly, the Prolog language has been proposed within the scope of Distributed Ledger Technologies (a.k.a.…”

“…tuProlog (Denti et al, 2001;tuProlog Home, 2021) is a relatively recent, research-oriented system which is the technological basis of several impactful works at the edge of the multi-agent systems and logic programming areas, such as TuCSoN (Omicini and Zambonelli, 1998), ReSpecT (Omicini and Denti, 2001), and LPaaS (Calegari et al, 2018). The main purpose of tuProlog is to make Prolog and LP ubiquitous (Denti et al, 2013).…”

“…Explainable AI: While all such issues are being tackled by the eXplainable AI community (XAI) (Gunning, 2016) by leveraging on a plethora of sub-symbolic tricks (Guidotti et al, 2019), an increasing number of works recognize the potential impact of symbolic AI models and technologies in facing these issues, such as the works by Calegari et al (2018 or Cyras et al (2021). It seems clear that symbolic inferential capabilities will be crucial for transitioning into the sustainable and humanity-serving AI that is urgently needed .…”

Logic-Based Technologies for Intelligent Systems: State of the Art and Perspectives

“…In the recent years, the historical dichotomy between the "two souls" of AI has been reconciled, in favor of a comprehensive vision where symbolic and sub-symbolic approaches are seen as complementary-rather than in a competition-so that they mutually soften their corners [155][156][157] (see Figure 1). While symbolic approaches are well suited for relatively small-sized problems implying complex but exact tasks-possibly relying on structured data-sub-symbolic approaches are best suited to use cases processing big (possibly huge) amounts of possibly unstructured data-where errors, or lack of precision, are tolerated to some extent, if unavoidable.…”

Human-Error Prevention for Autonomous Edge Software Using Minimalistic Modern C++

“…Formal methods [13][14][15] and logic programming [16] extract consistency specification from embedded software. However, after extraction, executable software is generally implemented by programming language.…”