It's just counting the letter R's in a word strawberry.

就是數草莓這個單詞中的字母 R。

So let's start with the traditional, like existing model GPT-4.0.

是以，讓我們從傳統的、類似現有型號的 GPT-4.0 開始。

So as you can see the model fails on this.

是以，正如你所看到的，模型在這一點上是失敗的。

There are three R's, but the model says there are only two R's.

有三個 R，但模型說只有兩個 R。

So why does this advanced model like GPT-4.0 make such a simple mistake?

那麼，為什麼像 GPT-4.0 這樣先進的模型會犯如此簡單的錯誤呢？

That's because models like this are built to process the text, not with the characters or words.

這是因為這樣的模型是用來處理文本的，而不是處理字元或單詞。

It's somewhere between, sometimes called a sub-word.

它介於兩者之間，有時被稱為子詞。

So if we ask the question to a model, a question that involves understanding the notion of characters and words, the model can really just make mistakes because it's not really built for that.

是以，如果我們向一個模型提出一個問題，一個涉及理解字元和單詞概念的問題，這個模型可能真的會犯錯誤，因為它並不是為此而構建的。

So now let's go on to our new model and type in the same problem.

現在，讓我們進入新模型，輸入同樣的問題。

This is the O1 preview, which is a reasoning model.

這是 O1 預覽版，是一款推理機型。

So unlike the GPT-4.0, it starts thinking about this problem before outputting the answer.

是以，與 GPT-4.0 不同的是，它在輸出答案之前就已經開始思考這個問題了。

And now it outputs the answer.

現在，它給出了答案。

There are three R's in the word strawberry.

草莓一詞中有三個 R。

So that's the correct answer.

這就是正確答案。

And this example shows that even for seemingly unrelated counting problem, having a reasoning built in can help avoiding the mistakes because it can maybe look at its own output and review it and more just be more careful and so on.

這個例子說明，即使是看似無關的計數問題，內置的推理功能也能幫助避免錯誤，因為它可以查看自己的輸出結果並進行審查，從而更加小心謹慎，等等。