Introduction to AI: Know This and You're Halfway There
<Table of Contents>
- Introduction: The Need for Proper Knowledge About AI
- Definition of AI: Structure, Learning, and Performance Capabilities
- Unit of AI: Models
- Deep Learning: Accelerating AI Development
- Important Moments in AI Development
- AI Terms Worth Knowing
- The Power and Influence of AI
- AI Industry and Future
- Conclusion: Getting Friendly with AI
Hello! In this article, I've summarized the content of "Introduction to AI: Know This and You're Halfway There." It covers basic AI concepts to current industry trends in a way that even non-specialists can understand. Let's take a look!
Introduction: The Need for Proper Knowledge About AI
The author mentions that they were inspired to write this article after a phone call with their mother. Their mother shared information about AI, but much of it was incorrect. This made them realize that while interest in AI is growing, there's a lack of sources providing accurate knowledge.
"Compared to the enormous interest in AI, there are too few places providing the quality knowledge about AI vision that we really need to know."
Definition of AI: Structure, Learning, and Performance Capabilities
AI (Artificial Intelligence) means artificially created intelligence, which must meet three conditions:
- Structure: Just as our intelligence comes from the brain, AI needs a structure (code).
- Learning Ability: Through this structure, it should be able to learn data and knowledge and improve its capabilities.
- Performance Ability: It should be able to perform specific tasks based on what it has learned.
For example, Chat GPT has a code structure, learns from our inputs, and has the ability to generate appropriate responses.
Unit of AI: Models
The unit for counting AI is 'models.' Just as we have Samsung's Galaxy and Apple's iPhone for smartphones, AI has models like Google's Gemini, OpenAI's GPT, and Meta's Llama. And just as the iPhone has versions from 1 to 16, AI models also have versions like GPT-1, 2, 3, 4.
Deep Learning: Accelerating AI Development
An essential concept in AI history is 'deep learning.' In the past, to distinguish between cats and dogs, developers had to code individual features like eye shape and ear shape, which had limitations.
Deep learning creates a brain-like structure (perceptrons interconnected like neurons), and when this 'brain' is exposed to lots of data (cat and dog photos), the AI learns by itself.
"Deep learning accelerated AI development. It showed that by feeding lots of data to this brain, it could perform any task very well, which showcased its potential."
Important Moments in AI Development
- March 2016, AlphaGo vs. Lee Sedol: A historic moment when a machine defeated a human, sparking widespread interest in AI
- November 2022, Chat GPT emergence: Interest in AI exploded as people could directly experience AI capabilities in everyday life
AI Terms Worth Knowing
1. LLM (Large Language Model)
A large language model specialized in natural language (language we use). It's called 'large' because it has many structures that act like neurons, and most language models created by major companies fall into this category.
2. Generative AI
AI that creates content (videos, audio, photos, etc.) that didn't exist before. For example, a feature in Photoshop that automatically generates a beautiful sky in the background of a photo.
3. Multimodal AI
AI that can process various types of information, not just language but also audio, video, images, etc. Most recent AI models (e.g., GPT-4) are multimodal AIs.
4. AGI (Artificial General Intelligence)
AI with thinking abilities that can judge and solve problems on its own, which is the goal of current AI developers. It's still a technology of the distant future, but we're moving toward it.
The Power and Influence of AI
AI is currently being used in various ways:
- Generative AI: Used for generating images, videos, and audio, which can be used in advertising or movie special effects, but also has the risk of being misused for crimes like deepfakes or voice phishing.
- Performance Assistant: Through AI like Chat GPT, you can get support for various tasks such as planning travel itineraries, writing invitation messages, or replying to emails.
AI Industry and Future
AI Value Chain
The AI industry should be understood not as a single independent technology but as a value chain where various technologies influence each other.
- AI models (brain): Developed by big tech companies like Google, Meta, Microsoft, Tesla
- Hardware (semiconductors, GPU, HBM): Led by companies like NVIDIA, SK Hynix
- Infrastructure (energy, data centers): Requires efficient energy technologies like SMR (Small Modular Reactor)
- Application areas: Expanding to robotics (humanoids, etc.), bio AI, etc.
Conclusion: Getting Friendly with AI
Let's actively use AI instead of fearing it.
"The most important thing is not to fear AI. With just the content I've explained, you now have much more knowledge about AI."
The author suggests that by experiencing various free AI tools and becoming familiar with them, you can develop a sense for AI and resolve concerns related to career paths or job searches.
And as an assignment for the above article, they asked readers to try asking Chat GPT or Gemini more than 5 curious questions and making it do more than 3 tedious tasks! Would you like to try it too?