What is Strong AI?
Strong artificial intelligence (Strong AI), also known as artificial general intelligence (AGI), is a theoretical AI system that can reason, self-teach, and perform tasks across various problem domains, similar to a human being. It can generalize knowledge and apply it across functions, plan ahead using current knowledge, and modify behavior to adapt to changing environments. Current AI technologies all function within a set of pre-determined parameters. Strong AI remains a theoretical goal due to various processing and evaluation limitations.
What is the difference between strong and weak AI?
Weak AI, or narrow AI, solves specific problems after training with data from that domain. For example, you must fine-tune a large language model on terminology from a particular field before it can act as a chatbot for customers in that industry. On the other hand, strong AI performs any task without having directly related training. Using human-like cognitive understanding, strong AI can adapt to new skills and learn as it does so.
Another vital distinction between strong and weak AI is that weak AI needs human interference to function. Researchers must define parameters and offer training data to improve the accuracy of weak AI in its function. Strong AI does not require manual input as it develops intelligence over time like human beings.
While weak AI specializes in a small range of specific functions, strong AI rapidly solves new problems, completes novel tasks, and applies previous knowledge to distinct situations, just like a human would.
What are the approaches in strong AI research?
Strong artificial intelligence uses a broader range of approaches than weak AI, allowing it to complete more complex tasks that reflect human intelligence. Strong AI systems achieve a high level of creativity, memory, perception, and deep learning ability to achieve human capabilities.
Here are some of the approaches researchers can use in strong AI research.
Symbolic AI
The symbolic approach to AI research aims to represent human thoughts as a complex logic network. AI uses an expanding network of ‘If’ statements to move through potential answers in a way that attempts to reflect human thought patterns. This approach to AI development allows it to interpret and understand ideas at a high level.
However, the downside to the symbolic approach is that more subtle tasks, like perception and common sense, are complex to recreate through the logic network.
Connectionist AI
The connectionist approach to AI research uses neural network architecture to recreate the structure of the human brain. As AI systems interact with stimuli, they can change the transmission pathway of information to reflect the changing pathways of brain neurons in neural networks.
Large language models specializing in natural language processing (NLP) use a connectionist approach. This strategy combines the logical processing of symbolic AI structures with the ability to exhibit human-like low-level cognitive skills.
Universal AI
The universal approach to AI aims to create a universal formula that strong AI can apply across different situations that may arise. This approach focuses on the calculation level, attempting to develop a theoretical solution to human intelligence that researchers can then repurpose to advance how strong AI functions.
Whole organism architecture
The whole organism architecture approach to AI development and research believes that strong AI needs to learn by interacting with the physical world. By giving strong AI a physical interface, it can explore and interact with its environment to develop new skills and enhance its intelligence.
Hybrid approach
Finally, the hybrid approach to strong AI research aims to blend different approaches to achieve more effective results. Instead of attempting to get results from one school of thought, a hybrid approach can combine different principles and research methods to enhance artificial general intelligence.
What are the tests to evaluate strong artificial intelligence research?
Researchers use a range of tests to evaluate the efficacy of strong AI. Here are some of the most prominent tests.
Turing test
Developed by Alan Turing, the Turing test aims to determine whether an evaluator can identify differences between human-generated and computer-generated output. If the evaluator can spot differences, the machine fails the test. If a strong AI can pass the test, it will have a high level of intelligence.
Researchers take this test one step further when testing strong artificial intelligence. They use the Extended Turing test, which evaluates various tasks in different domains and compares them to human-generated output in each area. Strong AI needs to be able to generate output that matches human output.
Since weak artificial intelligence only focuses on one task, it is likelier to fail the Extended Turing Test. Strong AI replicates a higher level of human behavior and aims to pass this extended test to prove its intelligence.
Chinese Room Argument
The Chinese Room Argument (CRA) is a philosophical thought experiment John Searle proposed in 1980. In it, he imagines a scenario in which a person is locked inside a room with a rulebook for understanding a foreign language.
Searle supposes that if the man received questions in that same foreign language under the door, he could create accurate responses using the rulebook. While the people outside the door would then think he understood the language, in reality, he wouldn’t understand the context of the output he creates.
This scenario undermines the Turing test, suggesting that a machine can perform a task without actually understanding the task it completes. The human mind has a semantic and mental context that a machine may not be able to produce just by having access to the same information.
What are the prominent technologies in strong artificial intelligence research?
Strong AI uses a range of emerging technologies to further the development of artificial intelligence.
Robotics
Robotics is an engineering practice that involves creating mechanical systems that can automatically perform physical tasks. In strong artificial intelligence, robotics is vital in the whole organism architecture approach. This discipline can produce physical robotic machines with sensors that enable strong AI to perform physical tasks.
Deep learning
Deep learning is the training of artificial intelligence systems to process data in a way that resembles thought processes in the human brain. Training these neural networks with deep learning helps strong AI to recognize patterns, process text and audio, and understand various types of information.
Natural language processing
Natural language processing allows strong AI systems to interpret and comprehend human language. NLP combines several technologies, like machine learning and advanced computational linguistics. It decomposes language into phonetic representations, transforming them into mathematical tokens that a computer program can then make predictions about.
Generative artificial intelligence
Generative AI extends deep learning, allowing strong AI systems to generate content from training data. By training a model with a dataset, generative AI can respond to queries and produce accurate responses based on its initial data.
Computer vision
Computer vision technology allows strong AI to understand, analyze, and process spatial and visual data. By drawing information from cameras, computer vision can help AI to navigate the real world, classify objects, and process images.
How can AWS support your artificial intelligence efforts?
AWS provides managed artificial intelligence services that help you train, deploy, and scale generative AI applications. Organizations use our AI tools and foundational models to innovate AI systems with their own data for personalized use cases.
Amazon Bedrock is a fully managed service wherein developers can use API calls to access generative AI models they deploy. You can select, customize, train, and deploy industry-leading foundational models on Bedrock to work with proprietary data.
Amazon SageMaker Jumpstart helps software teams accelerate AI development by building, training, and deploying foundational models in a machine-learning hub.
Use Amazon Elastic Compute Cloud UltraClusters to power your generative AI workloads with supercomputing GPUs to process massive datasets with low latency.
You can also quickly and easily add intelligence to your applications with dozens of pre-trained, fully managed AI services without needing ML training or expertise.
Get started with AI on AWS by signing up for an AWS account today.
Browse all cloud computing concepts
Browse all cloud computing concepts content here:
Did you find what you were looking for today?
Let us know so we can improve the quality of the content on our pages