3.5 C
New York
Wednesday, February 21, 2024

We won’t achieve AGI without simulations

We won’t achieve AGI without simulations

The much needed AI paradigm shift

Photo by julien Tromeur on Unsplash

Not long ago, I found myself struggling with a thought-provoking dilemma: What would happen if a baby, devoid of interfaces for sensory inputs, were locked in a dark room for its entire life? The only interaction with the outside world would be through a small aperture for food, water, and waste disposal. This unsettling premise, while ethically sensitive, serves as a stark metaphor for understanding the current limitations of artificial intelligence. Simulating this rare case provides a perfect, albeit morbid, empirical method to examine why AI, in its current form, has reached an impasse and, unfortunately, will not achieve AGI status.

Let us explore this reasoning together.

At the heart of building AGI lies the question: what constitutes the essence of intelligence? For this sensory-deprived child, genetic predispositions act as a foundational ‘raw dataset’, akin to the initial programming of an AI. The initial synaptic connections formed in the womb can be compared to a neural network’s weights and biases. The innate tendencies formed by these predispositions dictate basic survival needs — hunger, thirst, elimination — but without sensory input, the scope of learning and evolution is severely restricted.

Apart from the connections developed in the womb, the child would develop only minimal synaptic connections, akin to those in a neural network. This child would be unable to comprehend language, process visual stimuli, recognize physical touch, or understand auditory signals. Now, you might argue that multimodal AIs can perform these functions without external stimuli to infer from. However, it’s crucial to understand the fundamental difference between merely processing data and possessing the intelligence to make autonomous decisions and form new conclusions.

While multimodal AIs can analyze and interpret various types of data, their ‘understanding’ is severely constrained by the parameters of their programming and the data they have been fed. They lack the intrinsic, experiential learning that comes from living in and interacting with a dynamic environment. This gap is the crux of the much-sought-after AGI. Yet, without a means to simulate such comprehensive, experiential learning, we remain frustratingly close…

Source link

Latest stories