Symbolic AI, on the other hand, is based on human reasoning.

Symbolic AI, on the other hand, is based on human reasoning.

[LUM Magazine, Podcast] - Published on October 29, 2024in Science-Society

Logical and frugal, symbolic AI has not said its last word in the face of machine learning, which is now making its mark on the computing world. Researchers at Lirmm are betting on the hybridization of these two AIs.

In the 20th century, AI didn't need to call itself symbolic. Born in the 1950s, methods based on high-level abstract representations, known as "logic", dominated AI research. But over the past decade or so, with the development of computer processing power, masses of available data and new algorithms, neural AI - in reference to neural networks - has taken the lead. From 2012 onwards, historical AI had to differentiate itself from this new form of AI: it would become symbolic AI, because it relies on reasoning that mobilizes symbols. Marie-Laure Mugnier, computer scientist at Lirmm¹explains the dominant paradigm of this type of AI: "For a long time, work on AI was based on the postulate that, to be intelligent, you have to be able to reason. Researchers now rely on the representation of human knowledge in mathematical languages that enable the automation of reasoning. The researcher takes the example of simple deductive reasoning: Socrates is a man, men are mortal, therefore Socrates is mortal. These three assertions are linked together by a logical chain that the machine can reproduce.

The first major successes of this AI were in the medical field, to exploit medical knowledge bases and rapidly derive diagnostic information. "The MYCIN expert system was a precursor; exploiting a knowledge base of around 600 rules modeling a doctor's expertise, this program was able to identify the bacteria responsible for bloodstream infections, such as meningitis, and recommend antibiotic treatments. However, early expert systems were quite empirical, whereas today's knowledge-based systems are rigorously based on mathematical theories such as logic and probability." The logical approach is very different from that of deep learning, which relies on complex numerical calculations based on huge amounts of data. This is a major difference, since symbolic AI is by nature explicable, and therefore likely to be understood by users.

Analogy with the human brain

"Agronomists fromInrae came to see us at Lirmm precisely to put an end to "black boxes". So we worked on a project with the ABSys laboratory to develop an AI system capable of helping them design new agroecological systems", says Marie-Laure Mugnier. By exploiting plant databases built by ecologists and formalizing scientific knowledge on the relationships between plant functional traits and ecosystem services in the form of logical rules, the tool can identify species capable of providing certain ecosystem services. A crucial feature of this tool is that it can justify its results. "In viticulture, in particular, we have tested the identification of plants capable of fixing nitrogen, improving soil structure or storing water, which would be interesting for grassing vines (read: Integrates data and knowledge to support the
selection of service plant species in agroecology, in Computers and electronics in agriculture, 2024)."

Today, symbolic AI is still very effective in many fields, for example in solving problems modeled in terms of constraint systems. " It could be solving a Sudoku, but it could also be optimizing a car assembly line," she points out. To explain the difference between the two AIs, she proposes an analogy comparing them to the two systems that make up the human brain, according to Daniel Kahneman, psychologist and winner of the Nobel Prize in Economics in 2002: system 1, which is fast, unconscious and intuitive, used for pattern recognition, is neural AI, while system 2, which is slower, conscious, explicit and used for deduction, is symbolic AI.

Complex, high-level tasks

Neural AI has revolutionized the approach to computing in fields such as image and speech recognition, language translation and text generation. " But symbolic AI enables us to carry out complex, high-level tasks, which are still necessary for decision support, planning or collective deliberation", points out Marie-Laure Mugnier, who notes in passing that the craze for machine learning applied to everything shows its limits if the mass of data on which the algorithms run is insufficient: "I see a lot of students who, on company placements, develop neural AI tools but on data sets that are too small. And it doesn't work". This logical approach, based on a small amount of data, brings another advantage to symbolic AI: its frugality. For it is the exponential processing of ever-growing masses of data that is responsible for AI's ecological footprint.

At Lirmm, several research projects are focusing on hybrid AI aimed at combining the two approaches. " To take the example of agroecological systems, we could identify the plants growing naturally on a plot using image recognition from the Pl@ntNet application, then use our symbolic AI tool to determine their potential in terms of ecosystem services," explains Marie-Laure Mugnier.

1. Lirmm (UM, CNRS, Inria, UPVD, UPVM)
   ︎

Key words : IA, Artificial intelligence, LumLu, LUM Magazine, Research, Science, Company