The future of AI lies in developing hybrid models that combine data-driven machine learning techniques with knowledge-based approaches.
To achieve true cognition and understanding, AI systems need to integrate external knowledge structures and symbolic reasoning, going beyond purely data-driven methods.
In order to achieve human-like understanding, AI models need to bridge the gap between computation and mind by combining symbolic reasoning, knowledge representation, and cognitive abilities.
Deep dives
Hybrid models for a new wave of AI
The podcast episode discusses the need for a new wave of artificial intelligence and the potential of hybrid models. The panel highlights the limitations of current empirical and data-driven approaches in achieving strong AI, particularly in conversational agents, natural language understanding, and self-driving cars. They explore the idea of combining knowledge-based approaches with data-driven machine learning techniques to overcome these limitations and drive progress in AI. The panelists emphasize the importance of integrating modules with specialized capabilities and creating systems that can reason, plan, and act using a rich knowledge structure. They emphasize that the future of AI lies in developing hybrid models that combine the strengths of different approaches.
The challenge of knowledge acquisition in AI
The panelists discuss the challenge of knowledge acquisition in AI and the limitations of purely data-driven methods. They argue that learning from data alone is insufficient for achieving true cognition and understanding. While data-driven techniques are valuable for certain tasks and pattern recognition, they are limited in their ability to acquire new knowledge through abstract reasoning and deduction. The panelists highlight the importance of integrating external knowledge structures and explicit symbolic reasoning into AI systems to enable conceptualization, interpret complex situations, and perform analytical reasoning. They propose that future AI systems need to go beyond learning from data and incorporate a hybrid approach that combines neural networks with knowledge extraction and symbolic reasoning.
The significance of symbolic reasoning and the role of human-like understanding
The panelists delve into the significance of symbolic reasoning in AI and its role in achieving human-like understanding. They emphasize the importance of capturing deeper meanings and abstract categories beyond simple pattern recognition. They discuss how humans have the ability to use abstract reasoning, deduction, and conceptual hierarchies to understand and interpret information. They argue that computational processes alone, such as neural networks, cannot fully explain human cognition and consciousness. The panelists highlight the limitations in current AI systems that mainly rely on data and emphasize the need to bridge the gap between computation and mind by integrating symbolic reasoning, knowledge representation, and cognitive abilities into AI models.
The Importance of Context in Understanding Images
Context plays a crucial role in understanding images. There are two levels of context to consider: domain context and instant context. Domain context refers to how objects are perceived within the context of a specific task or domain. For example, an object that appears brown and long in an office setting may be more likely to be interpreted as a cigar. Instant context refers to how specific aspects of an image are interpreted based on the question or task at hand. Context helps in refining interpretations and providing a feedback loop, enhancing understanding. It is argued that meaning cannot be derived from an image without considering context, as the possibilities of interpretation are infinite.
The Limitations of Deep Learning and the Importance of Reasoning
While deep learning has made significant progress in image and pattern recognition, it falls short in areas that require reasoning and deduction. Deep learning can approximate solutions based on the data it is trained on, but it cannot acquire factual knowledge or engage in analytic thinking. Reasoning, deduction, and understanding complex concepts are uniquely human capabilities that cannot be learned solely from observations or experiences. Deep learning can achieve animal-like intelligence, but it cannot replicate the cognitive abilities that involve reasoning, planning, and understanding language. The limitations of deep learning highlight the need for additional cognitive processes beyond surface-level data processing.
It has been over three decades since the statistical revolution overtook AI by a storm and over two decades since deep learning (DL) helped usher the latest resurgence of artificial intelligence (AI). However, the disappointing progress in conversational agents, NLU, and self-driving cars, has made it clear that progress has not lived up to the promise of these empirical and data-driven methods. DARPA has suggested that it is time for a third wave in AI, one that would be characterized by hybrid models – models that combine knowledge-based approaches with data-driven machine learning techniques.
Joining us on this panel discussion is polymath and linguist Walid Saba - Co-founder ONTOLOGIK.AI, Gadi Singer - VP & Director, Cognitive Computing Research, Intel Labs and J. Mark Bishop - Professor of Cognitive Computing (Emeritus), Goldsmiths, University of London and Scientific Adviser to FACT360.
Moderated by Dr. Keith Duggar and Dr. Tim Scarfe
https://www.linkedin.com/in/gadi-singer/
https://www.linkedin.com/in/walidsaba/
https://www.linkedin.com/in/profjmarkbishop/
#machinelearning #artificialintelligence
Get the Snipd podcast app
Unlock the knowledge in podcasts with the podcast player of the future.
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode
Save any moment
Hear something you like? Tap your headphones to save it with AI-generated key takeaways
Share & Export
Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode
