Taku's Blog++: OpenAI's New Model Has Surpassed Humans!? This Is the Birth of Artificial General Intelligence (AGI)...!

Check out Takumi’s NEW English youtube channel 🎵
↓↓↓
https://www.youtube.com/@takuway

Yokohama Arena was a time when everyone connected all at once, like Wi-Fi, and downloaded something extraordinary.

Ten thousand people in a frenzy.

So many friends gathered from all over the country!

With no reasons, no maneuvering, no reservations, everyone became one, hands wide open!

Congratulations on the award~~

Fion, what a guy!

It was a time when the mechanism for becoming free landed not as words, but deep in the gut.

Afterward, at Kaoru Nakajima's place.

One delicious dish after another...

The conversation, the drinks, the energy — all of it was incredible!

Thank you so much!

Tomorrow is the costume fitting and the cast meeting!

The movie is moving forward, closer and closer to the start of shooting~~~

↓↓↓ Please register here ↓↓↓

Open Chat “The Society Supporting Director Takumi Yamazaki”

GPT‑6 Astra was unbelievable! Overwhelming at PC operation and programming tasks

↓↓↓

First, watch this!

https://www.youtube.com/shorts/IVSpKxbPPYQ

And then...

1. Introduction: Where AI Stands “Now,” and Its Use in Politics

Background: AI holds power great enough to change not only education and the economy, but a nation's security and diplomacy as well [00:46]。
Practical Use (AI as a Private Tutor): What matters is not simply having it read textbooks, but having AI and simulators actually used and experienced in the political arena — in briefings for the Prime Minister, for example [01:10]。

2. Differences in AI Development Between the U.S. and China

Here is what Anno-san saw for himself on the ground in each country, the U.S. and China.

The U.S. (Software — the Brain Side):
- They have entered the final sprint toward realizing “AGI (Artificial General Intelligence),” an intelligence that surpasses human beings [02:45]。
- The pace of development is extremely fast, and there's an end-game atmosphere in the air [02:55]。
China (Hardware — the Body Side):
- In AI's brain itself (LLMs), they are chasing the U.S. at a furious pace [03:06]。
- On the other hand, in the technology of actually putting “physical AI” (humanoid robots) — AI with bodies — to work in society, China leads the world [03:21]。

3. The Future Economy and Social Structure AI Will Bring

As AI evolves further, how will our lives and our economy change?

A New Economic Sphere Called “AI to AI”:
- Until now, the basis of the economy has been “humans buying and consuming things” [04:45]。
- Going forward, however, a new kind of economic exchange that never passes through humans — “AI selling services to, and buying them from, other AI (AI to AI)” — may grow very large [02:00]。
Growth and Abundance for the Whole World:
- A world may be coming in which AI's power produces goods and services in vast quantities and prices fall [05:37]。
- Even if human purchasing power declines, the economy between AIs will grow so much that society as a whole is expected to keep developing [52:17]。

Video → [http://www.youtube.com/watch?v=sJH1-7ymNC4]。

Programming and 3D design, which used to demand “difficult study only specialists could do,”have entered an era where anyone can do them easily, just by telling the AI in words what they want to make

① What Is “GPT-6 Astra”?

How It Differs From Previous AI: The AI's performance has been powered up further, and it is now newly available in ChatGPT and compatible apps (such as Cursor) [01:16]。
What's Amazing About It: Even without difficult specialized knowledge, it will carry out complex computer work for you from instructions given in plain words [00:35]。

② Integration With the Specialized Software “Blender”

Background: “Blender” is specialized software for creating 3D (three-dimensional) models [00:22]. Normally, mastering it takes years of study and a high level of skill.
What Changed: But by combining it with this new capability (GPT-6 Astra),simply asking, in natural language, “Make me a near-future motorcycle”is now enough for the AI to automatically build a realistic 3D model [00:30]。

③ Taking It Further: Turning What You Made Into a Game

What Else It Can Do: Not just building a 3D model — you can then tell the AI, “Make a game where I can drive the motorcycle you just made” [03:36]。
The Result: The AI writes the program automatically, and a game where you can actually drive the motorcycle you made is finished [03:48]. It even reproduces the physics of what happens when the motorcycles collide [04:11]。

④ Other Examples of Use

Beyond 3D modeling and games, you can easily create things like a “manga-making app and viewer” just by giving instructions [33:22]。

In Summary (What Makes This So Amazing?)

Programming and 3D design, which until now demanded “difficult study only specialists could do,”have entered an era where anyone can do them easily, just by telling the AI in words what they want to make — that is what this video explains, demonstrations included [00:54]。

http://www.youtube.com/watch?v=BLyMK4MwmWA

_______________________________

AI has come to handle external tools and OS operations as deftly as a human, and cleared unknown, rule-free problems with high accuracy — which means thata de facto AGI (Artificial General Intelligence) has finally become reality. That's what this is.

Sam Altman and the others at OpenAI claim that with this achievement, humanity has attained AGI

The video “OpenAI's New Model Has Surpassed Humans!? This Is the Birth of Artificial General Intelligence (AGI)...”

（https://www.youtube.com/watch?v=jC1jN2gWFyU) — here is a theoretical, step-by-step summary of its contents.

OpenAI's New Model Release and Its High Scores
- OpenAI's new model “GPT-6 ASTRA” has been released, and it recorded a very high score on a test that has it solve games it has never seen (the ARC-AGI 3 test) [00:02]。
- The results differ by mode: 62.7% in standard mode, and an accuracy of 99.9% in provider-adapter mode [00:26]。
How the Test Modes Work, and How They Differ
- Standard Mode: A mode in which the problems are solved by the AI model alone, using no external tools or notes whatsoever [00:43]. It's equivalent to a flesh-and-blood human solving hard problems purely from memory, inside their own head [01:18], and the score was 62.7% [00:38]。
- Provider-Adapter Mode: A mode in which results are reached using external tools (calculation via Python programs and the like) and notes alongside — just as a human would when solving a difficult math problem [00:54]. In this mode it achieved a score of 99.9% [00:32]。
Views on Whether AGI (Artificial General Intelligence) Has Been Achieved
- Sam Altman and the others at OpenAI claim that with this achievement, humanity has attained AGI [01:38]。
- Since humans themselves don't memorize every step of a calculation in their heads either, the analysis goes, a model that can produce results by making appropriate use of external tools can be said to have effectively achieved AGI [01:54]。
The Leap Beyond Conventional Generative AI, and Why It Matters
- This test is extraordinarily difficult: with no rulebook and no advance explanation, the model has to work out the definition of the rules and the solution through trial and error [02:17]。
- Generative AI has been criticized as nothing more than “a fill-in-the-blank drill that statistically predicts the next word,” but this model can now understand and judge for itself problems it knows nothing about, a computer's UI/UX, and even operations at the OS level [02:41]。
- Security limits aside, you could say a powerful agent has been born — one that can handle anything autonomously on an OS, just as a human would [03:25]。

_______________________________

Doubts About AI's Evolution, and the Conventional View [00:00]
- Among general users, AI up to now has only been put to work on things like drafting emails or looking up particular information, and a doubt lingers: “even if a new model comes out, won't it be much the same?”
How Fundamentally Different the New Model Is [00:05]
- In response: the new model is “something else entirely” from what came before, and has reached a level where its value is dramatically different even for ordinary people with no specialized knowledge.
The Biggest Change for Ordinary People, and Its Use Cases [00:19]
- Even an ordinary, non-engineer, humanities-background user can, if they have an idea for something they want to make, use this model to achievethe creation of complex, highly polished software.
A Transformation of the Old Process [00:45]
- Until now, when making a game, doing DIY electronics work, or running a scientific research simulation, users had to sit glued to a desk building prototypes, compiling documentation, writing code, and checking that it ran — work demanding an enormous amount of effort [00:49]. With the arrival of the new model, there is hope of breaking free from that inefficient old workflow [00:56]。