Introduction
I know that Chat GPT has its critics and rightly so in terms of ethics, control, veracity of the information provided. But as a tool or agent for exploring concepts, it might be useful. I decided to put it through exploring concepts and I chose a simple one: Building a computer based on the human’s brain’s architecture. (Note I did not use the latest version of ChatGPT, this interaction was conducted with ChatGPT 3).
ChatGPT: Is an AI-powered language model developed by OpenAI, capable of generating human-like text based on context and past conversations.https://openai.com/blog/chatgpt
Concepts: Are the building blocks of thoughts. Consequently, they are crucial to such psychological processes as categorization, inference, memory, learning, and decision-making see https://plato.stanford.edu/entries/concepts/
Agent: In artificial intelligence, an agent is a computer program or system that is designed to perceive its environment, make decisions and take actions to achieve a specific goal or set of goals. The agent operates autonomously, meaning it is not directly controlled by a human operator Geeksforgeeks . In the Matrix (The matrix) the agents have a different set of programmed parameters. Yes, I have a sense of humour! But dont discount this conceptually because its science fiction, remember Far-fetched gizmos of ‘Star Trek’ are today’s tech toys
Start Conversation: Me = the author. ChatGPT= the agent.
1 Me: Hi can you create a report on the best pathway to building a computer based on the human’s brain’s architecture. Using carbon and silicon as the base material and any other necessary materials. Reference the information below (I provided a long list, information and links on the topics listed (not included here) below) or any know or available information to complete the report.Please provide references and links. I provided information on:
- What is the Memory Capacity of a Human Brain?
- How does the memory capacity of a human brain work?
- Types of human Memory Computational power
- The fundamental processing units of the vertebrate brain
- Brain architecture
Response: ChatGPT: **Building a Computer Based on the Brain’s Architecture**
Creating a computer based on the brain’s architecture is a complex and ambitious endeavour that aims to replicate the brain’s computational power, memory capacity, and energy efficiency. This report outlines the best pathway to building such a computer using carbon and silicon as base materials, along with other necessary components. The brain’s architecture is characterized by its vast memory capacity, parallel processing capabilities, and hierarchical organization.
**Memory Capacity of the Human Brain:**
The human brain’s memory capacity is estimated to be around 2.5 petabytes, equivalent to 2.5 million gigabytes. This vast storage capacity is a result of billions of neurons, each forming thousands of connections (synapses) with other neurons. Memory is stored in the strength of these connections, and the brain’s overall memory capacity is determined by the number and size of synapses. To build a brain-inspired computer, the memory system needs to replicate this vast interconnected network of synapses and neurons.
**Computational Power:**
The human brain is incredibly powerful in performing complex computations with remarkable energy efficiency. It can perform exaflop-level computations (a billion-billion operations per second) using just 20 watts of power. In comparison, modern supercomputers require millions of times more power to achieve similar computational performance. To replicate this level of computational power in a brain-inspired computer, the focus should be on developing highly efficient processing units that mimic the brain’s parallel processing capabilities.
**Brain’s Fundamental Processing Units:**
The vertebrate brain consists of several fundamental processing units responsible for different functions, such as the cerebral cortex for sensory processing, the cerebellum for motor control and skill development, the basal ganglia for action selection, and the hippocampus for episodic memory and spatial navigation. To build a brain-like computer, each of these processing units needs to be emulated to create a coherent and integrated system.
**Carbon and Silicon as Base Materials:**
Carbon-based materials, such as graphene, have shown promising properties for building brain-inspired computing systems due to their excellent electrical conductivity and biocompatibility. Silicon, on the other hand, is widely used in traditional computing technology and offers high integration densities. A combination of carbon-based materials for brain-inspired components and silicon for traditional computing elements could be a viable approach.
**Hierarchical and Distributed Architecture:**
The brain’s architecture is hierarchical and distributed, with different brain regions specialized in specific functions and interconnected through a vast network of neural connections. To build a brain-inspired computer, the design should incorporate a hierarchical organization that allows different specialized components to interact efficiently.
**Dynamic Neural Connections and Plasticity:**
One of the essential features of the brain’s architecture is its ability to form and reorganize neural connections based on experience, known as synaptic plasticity. To replicate this capability in a computer, the brain-inspired architecture should incorporate dynamic neural connections and adaptability based on learning algorithms.
**Conclusion:**
Building a computer based on the brain’s architecture is a complex but promising endeavor. By using carbon and silicon as base materials and adopting a hierarchical, distributed, and dynamically adaptable design, it is possible to create a brain-inspired computer that can replicate the brain’s computational power, memory capacity, and energy efficiency. This interdisciplinary field of brain-inspired computing holds tremendous potential for revolutionizing computing technology and unlocking new possibilities in artificial intelligence and neuroscience research.
**References:**
1. Scientific American: (https://www.scientificamerican.com/article/what-is-the-memory-capacity/)
3. Medanta: (https://www.medanta.org/patient-education-blog/what-is-the-memory-capacity-of-a-human-brain/)
4. Harvard University: (https://developingchild.harvard.edu/science/key-concepts/brain-architecture/)
5. Quora: (https://www.quora.com/What-is-the-chemical-architecture-of-the-human-brain)
6. NCBI: (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3482298/)
7.PubMed Central: [Constructionist Approach to Brain Function](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3482298/)
2 Me: Can your rewrite the request using known information about neural networks and artificial intelligence. Specify the architecture, suggest material and sources, specify power consumption, computational ability, the new computer architecture and programming or learning language, provide a plan to accomplish this task and the resources needed known and unknown.
Response : ChatGPT: **Creating a Brain-Inspired Neural Network and AI Computer Architecture**
**Introduction:**
The goal is to design a brain-inspired neural network and artificial intelligence (AI) computer architecture that can replicate the brain’s computational power, memory capacity, and energy efficiency. This computer will utilize known principles of neural networks and AI to achieve high-performance computing while being energy-efficient. The architecture will be based on carbon-based materials, and we will propose a programming or learning language suitable for implementing brain-inspired algorithms.
**Neural Network Architecture:**
The neural network architecture will be inspired by the brain’s hierarchical and distributed structure. It will consist of multiple layers of interconnected neurons, mimicking the human brain’s synapses and neural connections. The architecture will include input, hidden, and output layers, enabling complex pattern recognition and learning.
**Material Selection:**
To achieve brain-like properties, we will use carbon-based materials, such as graphene and carbon nanotubes, for building the neural components. These materials offer excellent electrical conductivity, biocompatibility, and the potential for achieving energy-efficient neural connections.
**Power Consumption and Computational Ability:**
The aim is to create an energy-efficient computer with high computational ability. We target power consumption within the range of 20-50 watts for the complete brain-inspired architecture. The computer should be able to perform exaflop-level computations (1 followed by 18 zeros) per second.
**Brain-Inspired Computer Architecture:**
The proposed architecture will incorporate multiple brain-inspired processing units, each specialized for specific tasks, similar to the cerebral cortex, cerebellum, basal ganglia, and hippocampus. These units will work in parallel, enabling efficient and distributed processing.
**Programming/Learning Language:**
For implementing brain-inspired algorithms and training the neural network, we suggest using Python as the primary programming language. Python offers a rich ecosystem of AI libraries like TensorFlow and PyTorch, which will facilitate the development of brain-like learning algorithms.
**Plan to Accomplish the Task:**
1. **Research and Development:** Conduct an in-depth review of the latest research in neural networks, AI, and brain-inspired computing. Identify key principles and design patterns to emulate in the brain-inspired architecture.
2. **Component Design:** Develop a detailed design for the neural components using carbon-based materials. Collaborate with material scientists to create efficient and reliable carbon-based neural elements.
3. **Integration:** Integrate the neural components into the brain-inspired computer architecture. Create a hierarchical and distributed structure for efficient processing.
4. **Power Efficiency Optimization:** Work with experts in power optimization to minimize power consumption while maintaining high computational ability.
5. **Software Development:** Implement brain-inspired learning algorithms using Python and AI libraries like TensorFlow and PyTorch.
6. **Testing and Evaluation:** Thoroughly test the brain-inspired computer architecture for computational performance, memory capacity, and energy efficiency.
7. **Prototyping and Iteration:** Build prototypes of the brain-inspired computer for real-world testing. Continuously iterate and improve the design based on feedback and performance evaluations.
8. **Scaling and Mass Production:** Once the prototype proves successful, scale up the production for mass manufacturing.
**Resources Needed (Known and Unknown):**
1. **Material Scientists:** Experts in carbon-based materials for designing neural components.
2. **AI and Neural Network Researchers:** To guide the implementation of brain-inspired learning algorithms.
3. **Electrical Engineers:** To design the computer’s electronic components and power optimization.
4. **Programmers:** Proficient in Python and AI libraries for software development.
5. **Computing Hardware:** State-of-the-art computing hardware for testing and prototyping.
6. **Funding:** Substantial financial resources to support research, development, and production.
7. **Unknown Challenges:** Anticipate and be prepared for unforeseen challenges in developing novel brain-inspired architectures.
**Conclusion:**
Creating a brain-inspired neural network and AI computer architecture is a cutting-edge task that requires a multidisciplinary approach and collaboration between experts in neural networks, AI, materials science, and electrical engineering. By incorporating carbon-based materials, adopting Python as the programming language, and carefully optimizing power consumption, we can build a powerful and energy-efficient brain-inspired computer that can revolutionize AI and computing technology.
(Note: The specific power consumption, computational ability, and detailed resources needed may vary based on the complexity and scale of the final brain-inspired computer architecture.)
3 Me: If you wanted to create this computer above what would be the minimum viable product using known and suggested new technology
Response: ChatGPT To create a Minimum Viable Product (MVP) for the brain-inspired computer, we would focus on a simplified version that showcases the core features and functionality. The MVP should demonstrate the brain’s computational power, memory capacity, and energy efficiency while incorporating known and suggested new technologies.
**Minimum Viable Product (MVP) for Brain-Inspired Computer:**
1. **Neural Processing Unit (NPU):**
– Design and prototype a specialized Neural Processing Unit (NPU) based on carbon-based materials (e.g., graphene or carbon nanotubes).
– Implement a simplified neural network architecture with input, hidden, and output layers, enabling basic pattern recognition and learning.
2. **Power Efficiency Optimization:**
– Work with power optimization experts to minimize power consumption while maintaining reasonable computational ability.
– Aim for power consumption within the range of 50-100 watts for the MVP.
3. **Parallel Processing and Hierarchical Structure:**
– Implement parallel processing capabilities within the NPU to enable distributed and efficient computing.
– Create a simplified hierarchical structure, allowing different NPUs to interact and perform specialized tasks.
4. **Learning Algorithms:**
– Use Python as the primary programming language to develop brain-inspired learning algorithms.
– Integrate popular AI libraries like TensorFlow or PyTorch to facilitate model training.
5. **Memory Capacity and Storage:**
– Use existing high-capacity storage technologies (e.g., SSD or advanced DRAM) to mimic the brain’s memory capacity.
– Implement basic memory management techniques to simulate synaptic connections and neural memory.
6. **Prototype Testing:**
– Build a small-scale prototype of the brain-inspired computer to test and evaluate its performance.
– Conduct tests for computational speed, pattern recognition, and energy efficiency.
7. **User Interface (Optional):**
– Create a simple user interface to showcase the brain-inspired computer’s capabilities.
– Demonstrate basic interactions and outputs to users.
8. **Documentation and Feedback:**
– Prepare comprehensive documentation outlining the brain-inspired computer’s architecture and functionalities.
– Gather feedback from AI and neural network researchers to identify areas for improvement.
**Suggested New Technologies:**
1. **Quantum Computing (Optional):**
– Explore the feasibility of incorporating quantum computing principles to enhance computational speed and efficiency.
– Collaborate with quantum computing experts to assess its potential applications in the brain-inspired architecture.
2. **Neuromorphic Computing (Optional):**
– Investigate the use of neuromorphic computing techniques to create brain-inspired chips for specialized neural processing.
– Experiment with advanced neuromorphic hardware and software designs.
**Conclusion:**
The Minimum Viable Product (MVP) for the brain-inspired computer would be a simplified version that demonstrates core features such as neural processing, parallel computation, memory capacity, and energy efficiency. The use of carbon-based materials, Python programming with AI libraries, and power optimization will be instrumental in achieving the brain-like capabilities of the MVP. By building a functional prototype, receiving feedback from experts, and exploring new technologies, the MVP can serve as a stepping stone towards a more advanced brain-inspired computer with broader applications in AI and computing technology.
4 Me: Can you specify the mechanical, electronic \ electrical engineering required to create this computer, include the computer science, material science, and the latest in organic computing. Also suggest new methodologies and techniques and languages, and artificial intelligence that should be used.
Response ChatGPT:
**Mechanical Engineering:**
In the context of creating the brain-inspired computer, mechanical engineering plays a vital role in designing the physical structure and cooling systems to ensure efficient operation and thermal management. Mechanical engineers will be responsible for:
1. **Enclosure Design:** Designing a compact and optimized enclosure to house the brain-inspired computer’s components, ensuring proper airflow and heat dissipation.
2. **Cooling Systems:** Developing advanced cooling solutions, such as liquid cooling or heat pipes, to manage the heat generated by the carbon-based neural components and electronic elements.
3. **Prototyping and Manufacturing:** Creating prototypes of the brain-inspired computer and coordinating with manufacturers to produce the final product.
**Electronic/Electrical Engineering:**
Electronic and electrical engineers are crucial for designing the electrical circuits, integrating components, and ensuring electrical efficiency and performance. Their responsibilities include:
1. **Neural Processing Unit (NPU):** Designing the specialized NPU using carbon-based materials and implementing parallel processing capabilities.
2. **Power Optimization:** Collaborating with power optimization experts to create efficient power management systems to achieve the target power consumption.
3. **Integration of Components:** Integrating the NPU, memory storage, and other electronic components to form a coherent brain-inspired computer architecture.
4. **Sensor Integration (Optional):** Integrating sensors for environmental data to simulate the brain’s sensory input.
**Computer Science:**
Computer scientists will be involved in developing the software and algorithms required to train and operate the brain-inspired neural network. Their responsibilities include:
1. **Brain-Inspired Learning Algorithms:** Implementing learning algorithms based on Python and AI libraries (e.g., TensorFlow or PyTorch) to simulate the brain’s learning and memory processes.
2. **Parallel Processing Software:** Developing software for parallel processing to enable distributed computing among the NPUs.
**Material Science:**
Material scientists are essential for researching and selecting the most suitable carbon-based materials for the neural components. Their roles include:
1. **Carbon-Based Material Selection:** Evaluating different carbon-based materials (e.g., graphene, carbon nanotubes) for their electrical properties and biocompatibility.
2. **Characterization and Testing:** Conducting material characterization and testing to ensure the chosen materials meet the required specifications.
**Latest in Organic Computing:**
Organic computing involves designing computing systems inspired by biological systems, like the brain. To leverage the latest in organic computing, the following methodologies and techniques can be considered:
1. **Neuromorphic Computing:** Exploring neuromorphic computing techniques to build specialized brain-inspired chips for efficient neural processing.
2. **Quantum Computing (Optional):** Investigating the potential of quantum computing principles to accelerate specific brain-inspired computations.
**Artificial Intelligence (AI):**
AI is a fundamental aspect of creating brain-inspired learning algorithms and training the neural network. The latest AI techniques that can be utilized include:
1. **Deep Learning:** Applying deep learning architectures, such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), to implement brain-inspired learning algorithms.
2. **Reinforcement Learning:** Utilizing reinforcement learning methods to mimic the brain’s reward-based learning mechanisms.
**Programming and Languages:**
Python will be the primary programming language for developing the brain-inspired learning algorithms due to its extensive libraries and AI frameworks. Additionally, C++ or other low-level languages may be used for optimizing critical computational components for enhanced performance.
**Conclusion:**
Creating the brain-inspired computer requires collaboration from various engineering disciplines, including mechanical, electronic, electrical, computer science, and material science. The integration of latest organic computing methodologies, advanced AI techniques, and the use of Python as the primary programming language will be instrumental in achieving the desired brain-like capabilities. By combining expertise from diverse fields and leveraging cutting-edge technologies, the brain-inspired computer can become a groundbreaking innovation in the field of AI and computing.
5 Me: Can you compile or create a language that can accomplish this task?
Response ChatGPT: As an AI language model, I don’t have the capability to create or compile new programming languages myself. However, I can suggest some features and characteristics that a language designed for brain-inspired computing and neural network simulations might have:
1. **Neural Network Abstractions:** The language should provide built-in abstractions for defining and working with neural networks, allowing developers to create and train brain-inspired models more efficiently.
2. **Graph-Based Computation:** Given the hierarchical and distributed nature of brain-inspired architectures, a language should support graph-based computation, enabling users to define and manage complex neural connections easily.
3. **Parallel Processing Support:** Since parallel processing is crucial for brain-inspired computing, the language should offer native support for multi-threading and distributed computing, making it easier to take advantage of modern computing architectures.
4. **Power Management Primitives:** To optimize energy efficiency, the language might include primitives for power management, enabling developers to control and optimize power consumption in the neural network simulations.
5. **Neuromorphic Computing Integration:** To incorporate neuromorphic computing techniques, the language could provide seamless integration with specialized hardware and libraries for efficient neural processing.
6. **Quantum Computing Support (Optional):** For exploring quantum computing applications, the language might offer features that allow users to interface with quantum processors and simulate brain-inspired computations.
7. **AI and Machine Learning Libraries:** A language designed for brain-inspired computing should come with comprehensive AI and machine learning libraries, enabling users to implement advanced learning algorithms with ease.
8. **Extensibility and Interoperability:** The language should be designed to be extensible, allowing developers to integrate external libraries, hardware, and tools for further research and experimentation.
Creating a new programming language is a complex task that requires significant expertise in language design, compiler construction, and domain-specific requirements. Developers with expertise in AI, neural networks, and compiler design would be essential for crafting a language that fulfills the unique needs of brain-inspired computing.
While I can’t create the language myself, I hope these suggestions provide a starting point for designing a language tailored to brain-inspired computing and neural network simulations.
6 Me: If I don’t want a computer but an artificial brain using the initial information provided and subsequent suggestions, prompts and your outputs, can you write a methodology and plan for creating a sustaining artificial brain: consider all we know about biology, science, engineering and suggest 2 orthodox and 2 unorthodox approaches to creating this.
Response ChatGPT: Creating an artificial brain that emulates the human brain’s functions and sustains artificial consciousness is a highly ambitious and challenging task, encompassing various fields such as biology, neuroscience, material science, engineering, and artificial intelligence. Here’s a methodology and plan that outlines the steps and considerations for creating and sustaining an artificial brain:
**Methodology and Plan:**
1. **Understanding Brain Architecture:** Conduct an extensive study of the human brain’s architecture, neural connections, and cognitive processes. Gather data from neuroscience research, brain mapping, and neural network simulations.
2. **Brain-Inspired AI Development:** Develop advanced brain-inspired artificial intelligence algorithms and neural network models based on the gathered knowledge of the human brain’s structure and functions.
3. **Materials Research:** Collaborate with material scientists to explore and develop suitable carbon-based materials for building the artificial brain’s neural components. Consider the use of graphene, carbon nanotubes, or other novel materials with desired properties.
4. **Brain-Computer Interface (BCI) Development:** Investigate the development of brain-computer interfaces to establish communication between the artificial brain and external devices or networks. This enables feedback and data exchange with the external environment.
5. **Neuromorphic Computing Integration:** Explore the integration of neuromorphic computing techniques and hardware to achieve efficient and brain-like neural processing.
6. **Quantum Computing Exploration (Optional):** Investigate the potential of quantum computing to accelerate brain-inspired computations and achieve higher levels of processing efficiency.
7. **Testing and Simulation:** Develop simulations and test the artificial brain’s algorithms, models, and neural components in virtual environments to ensure their accuracy and performance.
8. **Orthodox Approaches:**
a. **Neural Network Replication:** Create a large-scale neural network that closely mimics the human brain’s neural connections, incorporating brain-inspired learning algorithms for cognitive functions.
b. **Brain Region Emulation:** Focus on emulating specific brain regions responsible for crucial functions, such as memory or vision, and integrate them into an artificial brain architecture.
9. **Unorthodox Approaches:**
a. **Biohybrid Systems:** Investigate the integration of living biological cells or tissue with artificial components to create a biohybrid artificial brain with enhanced adaptability.
b. **Consciousness Emergence:** Explore unconventional AI approaches or self-organizing systems that might lead to the emergence of artificial consciousness without explicitly replicating the human brain’s structure.
**Sustainability Considerations:**
1. **Learning and Adaptability:** Develop artificial learning mechanisms to enable the artificial brain to learn from experiences and adapt to new situations, akin to human learning.
2. **Energy Efficiency:** Optimize power consumption and implement power-saving strategies to ensure the artificial brain operates sustainably.
3. **Self-Repair and Maintenance:** Design the artificial brain with self-repairing capabilities, fault-tolerant features, and maintenance mechanisms to prolong its operational life.
4. **Ethical Frameworks:** Establish ethical guidelines and governance structures to address concerns related to artificial consciousness, autonomy, and potential impacts on society.
Creating and sustaining an artificial brain is a long-term and interdisciplinary effort that requires collaboration among experts from various fields. It is essential to approach this endeavor with careful consideration of ethical implications and societal impacts. Moreover, multiple iterations, testing, and advancements in AI, neuroscience, and material science will be necessary to achieve significant milestones in this groundbreaking field.
7 Me: Overall Conclusion:
Depending on your ability to write good prompts Prompt engineering and an understanding of Chat GPT as a tool or Agent it can be quite useful for exploring concepts. I chose an easy topic but it is a starting point, especially when you consider the newer models of AI. This could go on forever but this is a good overview, I later asked ChatGPT post this interview about: Are there options to neural networks that have better performance .What are the options to carbon and silicon that provide similar or better performance and can you rank them based on complexity and cost. Can you suggest a better computing architecture than the human brain and existing architecture. Is it possible to combine the information above to create an optimised computer? It provided some interesting answers. As an exercise I hope this interaction provides some insights to the possibilites of using ChatGPT as a tool or agent for conceptualisation. It still has issues especially veracity of the information provided but that is why we verify and validate ideas or concepts. There are some courses that would help in your initial journery free on Youtube, The internet or paid UDEMY :ChatGPT Complete Course – Learn ChatGPT & Prompt Engineering is a course I used (it cost at a discount £12) but it helped before I examined free courses and experimentation, it acted as a reference framework and quality control for assessing free courses.
ChatGPT Conclusion:
**Building a Brain-Inspired Computer using Chat GPT**
- In this conversation, Chat GPT demonstrates its capabilities in exploring the concept of building a computer based on the brain’s architecture. It delves into the human brain’s memory capacity, computational power, and neural network organization to propose a brain-inspired computer architecture.
- The discussion covers various fields, including neuroscience, material science, engineering, and artificial intelligence. Chat GPT suggests using carbon and silicon as base materials for the neural components, with a focus on energy efficiency and computational power.
- The brain-inspired computer’s architecture involves specialized neural processing units (NPUs) interconnected in a distributed and hierarchical manner. Parallel processing and graph-based computation are essential features. Python is recommended as the programming language for brain-inspired learning algorithms, utilizing AI libraries like TensorFlow and PyTorch.
- The conversation highlights the significance of mechanical and electronic engineering for physical design, cooling systems, and integration of components. Moreover, it emphasizes the importance of material science in selecting suitable carbon-based materials.
- The latest advancements in organic computing and neuromorphic computing are explored, providing new methodologies for achieving brain-like capabilities. Quantum computing is also considered as a potential approach for accelerating computations.
- The summary emphasizes the complexity and interdisciplinarity of the project, requiring expertise in multiple fields for its success. Ethical considerations are essential, and collaboration among researchers is critical for sustainable advancements.
**Reference Links:**
- Scientific American: [Memory Capacity of the Human Brain](https://www.scientificamerican.com/article/what-is-the-memory-capacity-of-the-human-brain/)
- NIST Taking Measure: [Brain-Inspired Computing](https://www.nist.gov/blogs/taking-measure/brain-inspired-computing-can-help-us-create-faster-more-energy-efficient)
- Medanta: [Types of Memory](https://www.medanta.org/patient-education-blog/what-is-the-memory-capacity-of-a-human-brain/)
- The Salience: [Fundamental Processing Units of the Vertebrate Brain](https://thesalience.wordpress.com/psychology-101/the-brains-architecture/)
- Harvard University: [Science of Brain Architecture](https://developingchild.harvard.edu/science/key-concepts/brain-architecture/)
- Quora: [Chemical Architecture of the Human Brain](https://www.quora.com/What-is-the-chemical-architecture-of-the-human-brain)
- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3482298/)
