Research AI vs Applied AI vs Generative AI
- (Stanford University - Hank Ping Han Hsieh)
- Overview
Research AI focuses on developing new AI algorithms and techniques through experimentation, while Applied AI uses existing AI models to solve specific real-world problems in various industries, and Generative AI specifically creates new content like text, images, or sounds based on learned patterns, essentially generating original data rather than just analyzing existing data.
Essentially, research AI is about exploring the boundaries of AI, applied AI is about practical application, and generative AI is about creating new things based on existing patterns.
- Research AI
Research AI is rather academic in nature and requires a heavy dose of math across a variety of disciplines before you even get to those parts that are specific to AI.
- Goal: To push the boundaries of AI by exploring new algorithms, techniques, and theoretical concepts.
- Focus: Developing new AI models, investigating fundamental principles, and pushing the limits of what AI can do.
- Example: Developing a new deep learning architecture for generating highly realistic images.
This aspect of AI focuses on the algorithms and tools that drive the state of AI forward. For example, what neural network structures might improve vision recognition results? How might we make unsupervised learning a more generally useful approach? Can we find ways to understand better how deep learning pipelines come up with the answers they do?
- Applied AI
Applied AI refers to the use of AI technology and techniques to solve real-world problems or achieve specific goals. This may involve using AI to analyze data and make predictions or decisions, automate processes or tasks, improve the efficiency or effectiveness of the organization, or provide personalized experiences or recommendations to users.
- Goal: To leverage existing AI models to solve practical problems in specific domains like healthcare, finance, or manufacturing.
- Focus: Implementing AI solutions to improve existing processes, make predictions, and automate tasks.
- Example: Using machine learning to predict customer churn in a subscription service.
Applied AI systems use machine learning algorithms to analyze data and make predictions or decisions based on that data.
Applied AI can be used in a wide range of industries and applications, such as healthcare, finance, manufacturing, transportation, and customer service. The goal of applying AI is to use the capabilities of AI to create value and solve practical problems.
- Generative AI
Generative AI (GenAI) refers to systems that create new, original content (such as images, videos, and text) that did not exist before. These systems use algorithms to generate content based on patterns and data, and they can learn from their output and improve over time.
- Goal: To create new, original content like text, images, or music based on learned patterns.
- Focus: Generating novel data that doesn't exist yet, often used in creative fields like art, design, and writing.
- Example: Using a model to generate a realistic portrait of a fictional character.
Generative AI is commonly used in creative fields such as art, music and design, as well as in research and development.
With generative AI, the potential for research and testing is limitless, as it has the ability to generate raw material that has not yet been created.
Generative AI can enhance current algorithms by generating training data for new neural networks or developing advanced deep learning architectures. In essence, generative AI functions like a machine that creates improved machines.
- [The AI Universe]
- From AI Research To AI Production
From AI research to AI production refers to the process of taking an AI concept developed in a research setting and transforming it into a fully functional, practical application that can be used in a real-world production environment, involving steps like data preparation, model training, optimization, deployment, and ongoing monitoring to ensure its effectiveness in a live system.
Key phases about this transition:
- Research phase:This stage focuses on exploring new algorithms, techniques, and theoretical concepts in AI, often conducted in academic labs or research institutions, with the goal of pushing the boundaries of what's possible.
- Development phase: Once a promising AI concept emerges from research, it's further refined and developed into a working prototype, considering factors like data requirements, computational limitations, and potential real-world applications.
- Production phase: This involves scaling up the AI model, integrating it with existing systems, deploying it on appropriate hardware, and monitoring its performance in a live production environment to ensure it meets desired standards and can adapt to changing conditions.
- Challenges in Moving from Research to Production
To get from research into production takes far more than just a model. It takes a village of both research and engineering efforts in tandem to make these things work. It takes hardware, it takes scalable hosting, it takes DevOps, it takes great data science, and much more.
Thankfully, more and more startups are building solutions for each building block, and big players (Uber and Netflix come to mind) are joining in as they open source more and more of their tooling.
Key challenges about this transition:
- Data quality and quantity: Real-world data can be messy and inconsistent, requiring extensive data cleaning and preparation to train robust models.
- Scalability: Models developed in research labs might not be optimized for large-scale deployment, necessitating adjustments to handle high volumes of data efficiently.
- System integration: Integrating an AI model into existing systems can be complex, requiring careful consideration of compatibility and workflows.
- Monitoring and maintenance: Once deployed, AI models need continuous monitoring to detect performance issues and ensure they remain effective over time.
