Artificial intelligence (AI) has gained immense popularity with the emergence of generative AI models such as Dall.E and ChatGPT-4. While these AI models have shown remarkable successes and occasional failures, they have sparked debates about the scope and dangers of advanced AI. However, what can we truly learn about natural intelligence, such as our own, from these generative AI models? As a philosopher and cognitive scientist, I have dedicated my career to understanding the human mind, and I believe that examining the contrast between AI and natural intelligence can provide valuable insights.
Generative AI models learn by creating predictive patterns in various types of data, allowing them to generate new versions of that data. ChatGPT, for example, specializes in generating text based on user prompts. However, natural intelligence encompasses much more than just text generation. Our brains are constantly learning to predict sensory information received through our senses, with the ultimate goal of selecting actions that help us survive and thrive in our environments.
This is where the key difference lies. Natural intelligence incorporates predictions of how our actions will impact our subsequent sensory experiences. For instance, if I accidentally step on my cat’s tail, my brain has learned that the resulting sensory input will include wailing, squirming, and possibly pain from a retaliatory scratch. This type of learning allows us to distinguish between cause and simple correlation, which is crucial for effective action in the world.
In contrast, current AI models, like ChatGPT, primarily predict specific types of data, such as sequences of words. While this provides a window into our world, it lacks the crucial ingredient of action. AI models only have access to verbal descriptions of actions and their typical effects, but they cannot practically intervene in the world to test and improve their predictions. This limitation is not only practical but also has deeper significance. Biological minds anchor their knowledge to the world by interacting with it and observing the cause-and-effect relationships. This experiential grounding enables us to truly understand sentences like “The cat scratched the person who trod on its tail.” Our generative models are shaped by our experiences through actions.
Looking ahead, it is possible that future AI models could develop anchored models by actively experimenting in the real world. In some domains like online advertising and social media, algorithms already adjust their behavior based on specific effects on users. If more powerful AI systems utilize this kind of action loop, they might start to resemble the grounding in action and experience that characterizes natural intelligence.
In conclusion, while generative AI models have provided valuable insights and advancements, there are significant divergences between artificial and natural intelligence. Natural intelligence not only predicts sensory information but also integrates actions and their consequences, contributing to a deeper understanding of the world. As AI continues to evolve, exploring these distinctions will help us appreciate the unique qualities of both artificial and natural intelligence.
The source of the article is from the blog regiozottegem.be