AI Ecosystem
According to Dr. Andrew Ng, the AI ecosystem can be divided into four parts: [Link: https://scet.berkeley.edu/ai-is-the-new-electricity-insights-from-dr-andrew-ng/]
- Hardware Layer: The Hardware Layer consists of the hardware or devices used in AI processing. Chip manufacturers such as Nvidia, Intel, and AMD play a crucial role in this segment. Entering this market has high barriers due to the rapid development of chip technology and the substantial funding required for research and development.
- Cloud Infrastructure: Cloud Infrastructure provides the backbone for services like data storage and processing. Leading cloud service providers such as Microsoft Azure, Google Cloud, and Amazon Web Services (AWS) are key players in this segment. These infrastructures enable the processing of large datasets and efficient operations.
- Developer Tooling: Developer Tooling refers to the tools that developers use to build and improve AI. Examples include OpenAI and ChatGPT, which make it easier for developers to create AI. These tools help developers work more conveniently and quickly, enabling them to build more efficient AI.
- Application Layer: The Application Layer involves applying AI to create various applications. Examples include Bing Search using AI to summarize information, GitHub CoPilot to help with code writing, and Canva using AI for image and graphic design. Applying AI in real-world applications allows users to benefit from AI more effectively and conveniently.
Although software developers currently focus on Developer Tooling, to fully capture the benefits of AI, it should be applied at the Application Layer, where users can directly experience it. Using AI in real applications can enhance user experience (UX) and increase work efficiency. Detailed cost/benefit analysis helps ensure that investments are worthwhile and effective.
AI Application
Nowadays, everyone talks about the need to apply AI, but many companies fail to use it appropriately, leading to numerous problems. For instance, an AI chatbot that provides nonsensical answers can tarnish the company’s service reputation, or using AI to write news without fact-checking can cause misinformation.
(Image from MidJourney)
If we understand how AI works, we will realize that AI is like a fortune teller that provides predictions. Choosing to use AI should follow these principles:
- UX (User Experience): Applying AI in products or services should improve user experience, making usage more convenient and faster. If AI confuses or inconveniences users, it may deter them from using the service or product. For example, using AI in a product recommendation application should simplify the product search, not complicate it.
- Impact: Consider the potential impacts if AI makes errors, especially in cases with serious consequences. For instance, using AI for disease diagnosis without a physician’s review can pose high risks and lack reliability. Therefore, each case of AI usage should be carefully checked and risk-assessed.
- Accuracy: The accuracy of the AI model is crucial. If the model is not good enough, the results can be incorrect. For example, using AI to generate images might produce errors, such as extra fingers or people with two heads. Applying AI in situations that require high accuracy might not be suitable.
- Profitability: Investments in AI should consider cost-effectiveness and appropriateness. For example, purchasing AI tools like Co-Pilot for employees can be costly, and the benefits may not justify the expense, such as merely assisting with email writing or being unnecessary for some employees. It should be evaluated whether the AI can enhance work efficiency and provide a worthwhile return on investment.
If you are interested in applying AI or need training related to AI, feel free to consult with us at @davoy.