luka.

Photographic AI Designer

[luka. @LLLluka]

We don’t see colors.
We see the result of light interacting with the world.
AI doesn’t understand “red.”
It reproduces patterns of how red appears under certain conditions.
Color is not data.
It’s a phenomenon.

#AI
#ColorTheory
#ComputerVision
#GenerativeAI
#Design

[luka. @LLLluka]

興味深い論文。
単一モデル前提での観測としては自然だが、
マルチエージェント構成（役割分担＋検証ループ）では、
“collapse / explosive” の一部は
問題ではなく制御対象として扱える可能性がある。
観測レイヤーと運用設計レイヤーの差が
ここに出ている印象。

https://t.co/HKSxuuFs5r⁠�

[luka. @LLLluka]

補足

補足です。

このあたりは「可能」というより、 実際には制御対象として“設計される領域”でもあります。

運用設計まで含めると、見え方が少し変わる部分かなと。

[luka. @LLLluka]

@livgorton とても面白い視点だと思います。

単一モデル前提での挙動として見ると自然ですが、 実運用では役割分担や検証ループを前提にするため、

こうした挙動は“問題”というより “制御対象”として扱われるケースもあります。

前提モデルの違いで見え方が変わる部分かもしれません。

[luka. @LLLluka]

@livgorton とても面白い視点だと思います。

単一モデル前提での挙動として見ると自然ですが、 実運用では役割分担や検証ループを前提にするため、

こうした挙動は“問題”というより “制御対象”として扱われるケースもあります。

前提モデルの違いで見え方が変わる部分かもしれません。

[Liv @livgorton]

There are concerns that AI is very alien to us. I am not saying that is or will never be the case, or that if would mean AI is default safe (humans do harmful things!). But, the world where their decisions make sense to us is a better one and this is further evidence of that.

[luka. @LLLluka]

@AnthropicAI 興味深い観測だと思います。

単一モデル前提での挙動としては重要ですが、 マルチエージェント構成（役割分担＋検証ループ）では、 同様の挙動は設計で抑制可能な領域でもあると感じています。

観測と運用設計では、少しレイヤーが異なる印象です。

[luka. @LLLluka]

🚨AIは数学ができない？
→ ちょっと違う
GSM-Symbolicの結果：
・数値変える → 精度低下
・無関係な一文 → 崩壊
・複雑化 → 一気に劣化
これ全部同じ現象構造依存
AIは意味で理解してるんじゃなくて
構造とパターンで処理してる
だから問題は能力じゃない設計
論文＋検証
https://t.co/BBBq7LN2tL

[Nav Toor @heynavtoor]

🚨SHOCKING: Apple just proved that AI models cannot do math. Not advanced math. Grade school math. The kind a 10-year-old solves.

And the way they proved it is devastating.

Apple researchers took the most popular math benchmark in AI — GSM8K, a set of grade-school math problems — and made one change. They swapped the numbers. Same problem. Same logic. Same steps. Different numbers.

Every model's performance dropped. Every single one. 25 state-of-the-art models tested.

But that wasn't the real experiment.

The real experiment broke everything.

They added one sentence to a math problem. One sentence that is completely irrelevant to the answer. It has nothing to do with the math. A human would read it and ignore it instantly.

Here's the actual example from the paper:

"Oliver picks 44 kiwis on Friday. Then he picks 58 kiwis on Saturday. On Sunday, he picks double the number of kiwis he did on Friday, but five of them were a bit smaller than average. How many kiwis does Oliver have?"

The correct answer is 190. The size of the kiwis has nothing to do with the count.

A 10-year-old would ignore "five of them were a bit smaller" because it's obviously irrelevant. It doesn't change how many kiwis there are.

But o1-mini, OpenAI's reasoning model, subtracted 5. It got 185.

Llama did the same thing. Subtracted 5. Got 185.

They didn't reason through the problem. They saw the number 5, saw a sentence that sounded like it mattered, and blindly turned it into a subtraction.

The models do not understand what subtraction means. They see a pattern that looks like subtraction and apply it. That is all.

Apple tested this across all models. They call the dataset "GSM-NoOp" — as in, the added clause is a no-operation. It does nothing. It changes nothing.

The results are catastrophic.

Phi-3-mini dropped over 65%. More than half of its "math ability" vanished from one irrelevant sentence.

GPT-4o dropped from 94.9% to 63.1%.

o1-mini dropped from 94.5% to 66.0%.

o1-preview, OpenAI's most advanced reasoning model at the time, dropped from 92.7% to 77.4%.

Even giving the models 8 examples of the exact same question beforehand, with the correct solution shown each time, barely helped. The models still fell for the irrelevant clause.

This means it's not a prompting problem. It's not a context problem. It's structural.

The Apple researchers also found that models convert words into math operations without understanding what those words mean. They see the word "discount" and multiply. They see a number near the word "smaller" and subtract. Regardless of whether it makes any sense.

The paper's exact words: "current LLMs are not capable of genuine logical reasoning; instead, they attempt to replicate the reasoning steps observed in their training data."

And: "LLMs likely perform a form of probabilistic pattern-matching and searching to find closest seen data during training without proper understanding of concepts."

They also tested what happens when you increase the number of steps in a problem. Performance didn't just decrease. The rate of decrease accelerated. Adding two extra clauses to a problem dropped Gemma2-9b from 84.4% to 41.8%. Phi-3.5-mini from 87.6% to 44.8%. The more thinking required, the more the models collapse.

A real reasoner would slow down and work through it. These models don't slow down. They pattern-match. And when the pattern becomes complex enough, they crash.

This paper was published at ICLR 2025, one of the most prestigious AI conferences in the world.

You are using AI to help you make financial decisions. To check legal documents. To solve problems at work. To help your children with homework. And Apple just proved that the AI is not thinking about any of it. It is pattern matching. And the moment something unexpected shows up in your question, it breaks. It does not tell you it broke. It just quietly gives you the wrong answer with full confidence.

[luka. @LLLluka]

winter10 を投稿しました https://t.co/N5Os7M5FdB #エキサイトブログ #エキブロ

[luka. @LLLluka]

ご挨拶 を投稿しました https://t.co/3ue8WCihti #エキサイトブログ #エキブロ

[luka. @LLLluka]

ご挨拶 を投稿しました https://t.co/VJ2C3q7UEJ #エキサイトブログ #エキブロ

[luka. @LLLluka]

winter5 を投稿しました https://t.co/oHCNLP6W7j #エキサイトブログ #エキブロ

[luka. @LLLluka]

winter3 を投稿しました https://t.co/9VGEeBOesn #エキサイトブログ #エキブロ

[luka. @LLLluka]

winter２ を投稿しました https://t.co/StWvf0vUVW #エキサイトブログ #エキブロ

[luka. @LLLluka]

winter1 を投稿しました https://t.co/jNvpmvdDDX #エキサイトブログ #エキブロ

[luka. @LLLluka]

https://t.co/gVUgti68SL

[luka. @LLLluka]

Why does it only apply in some cases?
→ Because ethics is not a universal filter.
It is a structure we invoke when someone could be harmed.

[luka. @LLLluka]

I just published a new note:

🪞 “Ethical Breathing of AI Images:
The Three-Observer Model & the 6W3H-LITE protocol.”

As AI visuals become indistinguishable from reality,
we need ways to prevent misidentification and protect human dignity.

[luka. @LLLluka]

The framework proposes a shared decision process:
RUKU (structure/XYZ), BooBoo (ethics margin—soft safety),
and ACOS (final routing: Publish / Caution / Block).

note: https://t.co/F2EOJwyPKF

[luka. @LLLluka]

https://t.co/8Y6jxBopN4

[luka. @LLLluka]

AIで「検証」「構築」「骨格」という言葉が頻出していますが、
これらは雰囲気で行えるものではありません。

検証には

1) 事前定義
2) 条件統制
3) 再現性

が必須です。
外部の観測／安全レイヤーなしでは成立しません。

それらが揃わない“検証した”は、ただの感想です。