Opinion: Artificial General Intelligence: A Distant Dream Anchored in Human Behavior

Do we have to understand ourselves first?

The aspiration for Artificial General Intelligence (AGI) has surged to the forefront of technological ambitions, illustrating a future where machines exhibit human-like adaptability and intelligence. This vision, visualized by characters like Iron Man’s Jarvis, is not just about engineering advanced systems but fundamentally transforming our interaction with technology. However, achieving AGI involves overcoming substantial challenges, notably in how we approach data diversity and integration.

The Current Landscape of AGI Development

Currently, AGI development is hamstrung by a reliance on narrowly defined data sets. Traditional models are fed structured data — text, numerical inputs, specific behavioral indicators — that fall drastically short of capturing the complex tapestry of human experience. For instance, in healthcare, while metrics like medication schedules and diagnostic results are easily quantifiable, subtler yet crucial data like emotional distress or socio-environmental impacts are often overlooked.

Consider a hypothetical scenario involving ‘Mary,’ a patient whose health deteriorates every Thursday, baffling the AGI system managing her care. It is only through human observation that a nurse discovers Mary’s worsening condition is linked to stressful visits from her in-laws. This scenario illustrates a critical gap in AGI’s capabilities: understanding the nuanced, often non-quantifiable human condition.

For AGI to truly mimic human cognitive functions, it must navigate an extensive array of data types, far beyond current capacities. Here are broad categories essential for AGI:

• Biological Data: Genetic codes, neurophysiological signals, hormonal fluctuations.
• Environmental Data: Living conditions, socioeconomic backgrounds, cultural contexts.
• Interactional Data: Emotional exchanges, social dynamics, conflict resolutions.
• Temporal Data: Behavioral changes over time, environmental shifts.
• Perceptual Data: Sensory inputs such as sights, sounds, and tactile sensations.

This diversity reflects just the surface of the necessary data spectrum, pointing to the immense scope of integration required.

Healthcare: A Model for AGI Data Integration

Healthcare is an exemplary domain where AGI could revolutionize data utilization. A sophisticated AGI system in healthcare could integrate from molecular to societal data points — patient genetics, real-time physiological data, and broader socio-economic conditions — to create nuanced care models. This level of integration would allow for predictions and interventions that are personalized and highly adaptive.

To illustrate, consider the following table which categorizes just a fraction of the potential data types and their applications in an AGI-driven healthcare system:

The integration of such diverse data types is not without challenges. It involves sophisticated data processing capabilities, advanced machine learning models, and rigorous ethical considerations, particularly regarding privacy and data security. Moreover, the need for an interdisciplinary approach — combining insights from computer science, psychology, sociology, and more — is crucial to develop AGI systems that are not only technically proficient but also socially and ethically aware.

I don’t think we will solve AGI until we all understand human behavior.

Teaser — for the past 25 years I have been studying human behavior and its relationship to engineering and computing. I may have a solution to AGI’s primary enemy. Email me for more information.