Generative Ai vs. AGI: Strengths and Weaknesses of Large Language Models


A moderately detailed consideration of interactive LLMs as cognitive systems is given, focusing on LLMs circa mid-2023 such as ChatGPT, GPT-4, Bard, Llama, etc.. Cognitive strengths of these systems are reviewed, and then careful attention is paid to the substantial differences between the sort of cognitive system these LLMs are, and the sort of cognitive systems human beings are.

It is found that many of the practical weaknesses of these AI systems can be tied specifically to lacks in the basic cognitive architectures according to which these systems are built. It is argued that incremental improvement of such LLMs is not a viable approach to working toward human-level AGI, in practical terms given realizable amounts of compute resources. This does not imply there is nothing to learn about human-level AGI from studying and experimenting with LLMs, nor that LLMs cannot form significant parts of human-level AGI architectures that also incorporate other ideas.

Practical applications of LLMs to data analytics and reasoning are then discussed, with a focus on what they can do on their own, and how their functionality can be expanded by bringing them further in an AGI direction via hybridizing them with other tools such as knowledge graphs, reasoning engines and evolutionary learning systems. One approach described involves using LLMs to guide semantic parsing that maps natural language expressions into formal logic expressions, which can then be input into mathematical and commonsense theorem-provers. Another approach involves fine-tuning LLMs on quantitative data contained in CSV data files via using LLMs to interpret the row and column headers in the data files. Another involves use of evolutionary learning to create hypotheses to be evaluated by an LLM. The use of these three techniques together enables biomedical analytics based on a combination of natural language, genomic and clinical data plus background knowledge comprising structured bio-ontologies and large-scale data repositories.

Learning objective is to provide the attendee with a broader understanding of how the capability of LLMs for learning and reasoning can be enhanced by integrating them with AI tools drawn from other paradigms, such as reasoning engines and evolutionary learning processes. Open source tools involved are: the Llama2 LLM , and the OpenCog Hyperon AGI toolkit


Dr. Ben Goertzel is a cross-disciplinary scientist, entrepreneur and author. He leads the SingularityNET Foundation, the OpenCog Foundation, and the AGI Society which runs the annual Artificial General Intelligence conference.

Dr. Goertzel also chairs the futurist nonprofit Humanity+, serves as Chief Scientist of AI firms Rejuve, Mindplex, Cogito and Jam Galaxy, all parts of the SingularityNET ecosystem, and serves as keyboardist and vocalist in the Jam Galaxy Band, the first-ever band led by a humanoid robot.

As Chief Scientist of robotics firm Hanson Robotics, he led the software team behind the Sophia robot; as Chief AI Scientist of Awakening Health he leads the team crafting the mind behind Sophia’s little sister Grace.

Before entering the software industry Dr. Goertzel obtained his PhD in mathematics from Temple University in 1989, and served as a university faculty in several departments of mathematics, computer science and cognitive science, in the US, Australia and New Zealand.

Open Data Science




Open Data Science
One Broadway
Cambridge, MA 02142

Privacy Settings
We use cookies to enhance your experience while using our website. If you are using our Services via a browser you can restrict, block or remove cookies through your web browser settings. We also use content and scripts from third parties that may use tracking technologies. You can selectively provide your consent below to allow such third party embeds. For complete information about the cookies we use, data we collect and how we process them, please check our Privacy Policy
Consent to display content from - Youtube
Consent to display content from - Vimeo
Google Maps
Consent to display content from - Google