Houbing Herbert Song, Ph.D., an IEEE Fellow, explains what neuro-symbolic AI is and why it might be a pathway toward Artificial General Intelligence (AGI). This article originally appeared in Insight Jam, an enterprise IT community that enables human conversation on AI.
AI is advancing rapidly. According to The Impact of Technology in 2024 and Beyond: an IEEE Global Study, AI helps detect and predict events quickly, such as outbreaks, unauthorized or unsafe drone operations, bias, cybersecurity threats, and malicious activities, driving innovation and competition in a range of application domains including environmental sustainability, space tech and exploration, smart cities, manufacturing, agriculture, energy, healthcare and medicine, and transportation.
On October 30, 2023, President Biden signed an Executive Order on the Safe, Secure, and Trustworthy Development and Use of AI. Building Artificial General Intelligence (AGI), a powerful form of AI that could theoretically rival humans, has been a distant goal. However, there are three major challenges associated with state-of-the-art (SOTA) AI algorithms: they lack generalizability (i.e., AI models are only as good as the data they are trained on), transparency and interpretability (i.e., AI models are “black box” models: opaque, non-intuitive, and difficult for people to understand), and robustness (i.e., imperceptible perturbations to AI inputs could altering its output).
AGI of the future will be characterized by three capacities:
- Grounding: AGI systems must understand the concepts they reason over and operate with.
- Instructiblity: AGI systems can be proven experimentally to change their behavior appropriately in response to explicit feedback provided by even non-expert users.
- Alignment: AGI systems must be judged by how well their operations align with expectations of objective truths in a domain and correspond to societal expectations and human intentions in their operations.
Neuro-symbolic AI, which integrates neural networks with symbolic representations, has emerged as a pathway towards AGI because of its potential to enable humans to understand and trust their behaviors, generalize to new situations, and deliver robust inferences, and strengthen AI in terms of grounding, instructiblity, and alignment.
“Neuro-symbolic” bridges the gap between two distinct AI approaches: “neuro” and “symbolic.” On the one hand, the word “neuro” implies the use of neural networks, especially deep learning, which is sometimes referred to as sub-symbolic AI or connectionism. This technique is known for its powerful learning and abstraction ability, allowing models to find underlying patterns in large datasets or learn complex behaviors. On the other hand, “symbolic” refers to symbolic AI or symbolism. It is based on the idea that intelligence can be represented using symbols like rules based on logic or other representations of knowledge, such as logical constraints, equations, finite state machines, relational graphs, and visual concepts.
In the history of AI, the first wave of AI emphasized handcrafted knowledge. In that era, computer scientists focused on constructing expert systems to capture the specialized knowledge of experts in rules that the system could then apply to situations of interest. The second wave of AI emphasized statistical learning, with computer scientists focused on developing deep learning algorithms based on neural networks to perform various classification and prediction tasks. The third wave of AI emphasizes integrating symbolic reasoning with deep learning, i.e., neuro-symbolic AI, and computer scientists focus on designing, building, and verifying safe, secure, and trustworthy AI systems.
There have been many advances in the emerging area of neuro-symbolic AI, such as logic neural networks, logic tensor networks, physics-informed neural networks and scientific machine learning, graph neural networks, neuro-symbolic programming, neuro-symbolic visual question answering, verification and validation, testing and evaluations of neuro-symbolic AI, neuro-symbolic transfer learning, and neuro-symbolic reinforcement learning, among others.
Neuro-symbolic AI, the integration of connectionism with symbolism, can create safe, secure, and trustworthy AGI systems, including healthcare and medicine, finance, criminal justice, autonomous and cyber-physical systems, and high-performance computing applications. However, transformative advances are needed to enable the safe, secure, and trustworthy development and use of neuro-symbolic AI towards AGI.
