ChatGPT is really good

ChatGPT is really good at organizing my thoughts
==========================

🔹 Latest trends and direction of change in AI technology
The recent development of artificial intelligence (AI) technology is centered on three main trends: miniaturization, on-device utilization, and open-source technology spread. These changes are expected to have a major impact on the universalization of AI and the development of personalized services.

1️⃣ Direction of change in AI technology

📌 Limitations of existing AI services
✅ Server-Based AI from Large Enterprises

Most of the existing AI services are based on cloud servers (AWS, Google Cloud, Azure, etc.).
Users do not own the AI model directly, but must access AI services provided by enterprises through the Internet.

✅ Security concerns & data privacy concerns

There is a risk of personal information leakage because data must be transferred to a central server.
Sensitive data such as the user's voice, face, and document are stored on the cloud server.
The structure in which companies control the operational and pricing policies of AI services.

📌 New AI Trends (Small & On-Device AI)

🔹 AI의 소형화 (Model Compression & Quantization)
Research has been actively conducted to optimize large AI models such as GPT, LLaMA, and Stable Diffusion to be lightweight and viable on mobile phones and PCs.
Typical technologies:
Quantization: reduces model size and speeds up execution by converting the weights of AI models to FP32 → INT8/INT4.
Knowledge Distillation: Compressing key information of a large AI model into a small model for similar performance.

🔹 On-Device AI (On-Device AI)
Technology that runs AI directly on smartphones, PCs, and IoT devices without a network connection.
Key Advantages:
✅ Enhanced Security → Do not send data to an external server.
✅ Speed up → Network latency disappears, enabling quick response.
✅ Cost savings → No cloud server fees.
Case in point:
Apple Neural Engine (ANE) → Run AI features on the iPhone (e.g., face recognition, voice assistant)
Qualcomm Hexagon AI Engine → Optimizing AI Models on Android Devices
Meta's LLaMA 3 → Large language model that runs on a personal computer

2️⃣ Spread of open-source AI

✅ Democraticization of AI Technology
Previously, only large IT companies (Google, OpenAI, Microsoft, etc.) were able to develop high-performance AI models.
Recently, open-source AI models (e.g., LLaMA, Mistral, Whisper) have emerged, enabling individuals and small businesses to take advantage of high-performance AI.

✅ Representative open-source AI project
LLaMA 2 (Meta) → Powerful Natural Language Processing (NLP) AI, can run on personal computers
Mistral 7B → Light language model optimized for high-speed processing
Whisper (OpenAI) → Free Open Source Speech Recognition Model
Stable Diffusion → Image Generation AI, can run locally without the cloud

✅ Advantages of open-source AI
Affordable or free to use.
Can be customized to suit individuals and businesses' needs.
Creating an environment where AI technology is not monopolized and can be used by anyone.

3️⃣ Possibility of proliferation of personalized AI services
💡 miniaturization of AI + use of on-device + spread of open-source technology → Personalized AI services have emerged

✅ Meeting user needs and security needs
Users can directly manage their data and AI models.
Create personalized services using AI without cloud dependency.
For example:
AI assistant → Learn the schedules and habits of individual users to provide customized advice.
On-device translation AI → Live translation without the Internet.
AI chatbot running locally → enhanced privacy.

✅ Accelerate the pace of technological advancement
Due to the miniaturization of AI and the spread of open-source AI, personalized AI services are likely to spread faster than expected.
Not only companies but also general developers, researchers, and entrepreneurs can develop services using AI.

🔹 Conclusion: A New Paradigm of AI Technology

🔥 AI technology is fast moving from "large cloud server-centric" to "personalization & on-device."
✅ Research is actively conducted to optimize large AI models so that they can be executed on small devices.
✅ With the spread of open-source AI, individuals and small and medium-sized enterprises can also use AI.
✅ On-device AI is likely to become the new standard considering security and speed.

🚀 Forward-looking:

Large cloud-based AI services still exist, but the **"Local AI Era" **, where individuals run their own AI models, is likely to open in earnest.
The popularization of AI is expected to proceed faster than expected through cost reduction + security enhancement + user customization.