Personal tools

Neuroscience and Human Level of AI

Stanford University_080921E
[Stanford University]


- Overview

Advances in artificial intelligence (AI) have enabled the creation of AIs that can perform tasks previously thought only humans could, such as translating languages, driving cars, playing world-championship-level board games, and extracting protein structures. However, each of these AIs is designed and exhaustively trained for a single task and is able to learn only what is required for that particular task.

Recent AI that produces fluent text (including conversations with humans) and produces impressive and unique art may give the false impression that the brain is at work. But even these are specialized systems that perform narrowly defined tasks and require extensive training.

Combining multiple AIs into a single AI that can learn and perform many different tasks remains a daunting challenge, let alone pursue the full range of tasks humans perform or take advantage of the range of experience available to humans to reduce the amount of data that would otherwise be required to understand how perform these tasks. The current best AIs in this regard, such as AlphaZero and Gato, can handle various tasks that fit a single model, such as playing games. Artificial general intelligence (AGI) capable of performing a wide range of tasks remains elusive.

Ultimately, AGI needs to be able to effectively interact with each other and people in a variety of physical and social contexts, integrate the various skills and knowledge required to do so, and learn flexibly and efficiently from these interactions.

Building an AGI boils down to building an artificial mind, albeit a vastly simplified one compared to a human mind. To build an artificial mind, you need to start with a cognitive model.


- Human-Level of AI

Even if we reach a state where AI can behave like a human, how can we be sure it can continue to behave in this way? We can base the humanization of AI entities on:

  • Turing Test
  • The Cognitive Modelling Approach
  • The Law of Thought Approach
  • The Rational Agent Approach


- Convergence of Neuroscience and AI

Artificial intelligence (AI) is a field of computer science that involves using machines to simulate human intelligence so that machines can acquire problem-solving and decision-making abilities similar to those of the human brain. Neuroscience is the scientific study of brain structure and cognitive function. Neuroscience and AI are interrelated. These two fields promote each other and make progress together. 

Neuroscience theories have brought many unique improvisations to the field of AI. Biological neural networks implement complex deep neural network architectures and are used to develop a variety of applications, such as word processing, speech recognition, target detection, etc. Additionally, neuroscience helps validate existing AI-based models. 

Reinforcement learning in humans and animals has inspired computer scientists to develop reinforcement learning algorithms in artificial systems, enabling these systems to learn complex strategies without explicit instructions. This learning helps build complex applications such as robot-based surgeries, self-driving vehicles, gaming apps, and more. 

In turn, AI, with its ability to intelligently analyze complex data and extract hidden patterns, is perfect for analyzing very complex neuroscience data. 

Large-scale simulations based on AI can help neuroscientists test their hypotheses. By interfacing with the brain, AI-based systems can extract brain signals and generate commands based on the signals. These commands are fed into devices such as robotic arms to help paralyzed muscles or other body parts move. 

AI has multiple use cases in analyzing neuroimaging data and reducing the workload of radiologists. Neuroscience research contributes to the early detection and diagnosis of neurological diseases. Likewise, AI can be effectively applied to the prediction and detection of neurological diseases.


[More to come ...]






Document Actions