# Why isn't ChatGPT AGI? (Or is it?)

## Метаданные

- **Канал:** Conner Ardman
- **YouTube:** https://www.youtube.com/watch?v=7eexMbak0mc
- **Дата:** 29.04.2026
- **Длительность:** 1:22
- **Просмотры:** 1,783

## Описание

What makes AGI? Is ChatGPT already there?


Prepping for your frontend interviews? Use code "conner" for a discount on my product FrontendExpert:
https://www.frontendexpert.io/conner


🎬 TikTok: https://tiktok.com/@connerardman
💼 LinkedIn: https://linkedin.com/in/connerardman
📸 Instagram: https://instagram.com/connerardman
📢 Twitter/X: https://x.com/connerardman
💻 Video/Coding Gear (affiliate): https://www.amazon.com/shop/connerardman


Business/brands 👉 youtube@connerardman.com

## Содержание

### [0:00](https://www.youtube.com/watch?v=7eexMbak0mc) Segment 1 (00:00 - 01:00)

What exactly can Chat GPT not do that an AGI would be able to do? Because I think this is an important distinction, and it's one that a lot of people actually have trouble coming to terms with. So, great question. The difference becomes much clearer when you frame it as capability gaps. Persistent self-directed goals. Chat GPT only responds to prompts, no long-term goals or initiative, and doesn't decide to do things. Point two here, in true understanding versus pattern matching. So, Chat GPT predicts likely text based on training, can simulate understanding extremely well, but it doesn't actually understand in a grounded way. An AGI would actually have some internal model of reality. It would understand cause and effect, not just correlations, and have real common sense. So, for example, Chat GPT can explain physics, AGI could discover new physics. Next, we also have learning continuously from real-world experience. So, Chat GPT doesn't learn after deployment, no real-time learning, no memory unless explicitly implemented. It can't improve itself from experience autonomously versus AGI would learn like a human. It would try something, fail, and update its mental model. Long-term planning and execution. So, Chat GPT can help plan, but it doesn't execute, whereas AGI presumably would, I guess, need like an infinite context window or at least a better way of actually handling that.

---
*Источник: https://ekstraktznaniy.ru/video/50153*