Name: AI 是一項「產品」還是一個「功能」？ (AI the Product vs AI the Feature)
Uploaded: 2024-06-21T02:51:32.000Z
Duration: 8 min 16 s

動名詞

Okay, so I was listening to the Wired podcast this morning and they said something that really nailed the way I think about a lot of emerging technologies, especially AI.

好吧，今天早上我在聽《連線》播客的節目，他們說了一些話，真正觸動了我對很多新興技術的思考，尤其是人工智能。

過去簡單式

What they said was WWDC, which we just had from Apple, basically proved that AI is a feature, not a product.

他們所說的是 WWDC，我們剛剛從蘋果公司那裡瞭解到，WWDC 基本上證明了人工智能是一種功能，而不是產品。

進階程度副詞句型

And this hit so hard for me because that's the exact question that we've been asking ourselves for so long here about new stuff.

這對我的打擊很大，因為這正是我們長期以來一直在問自己的關於新東西的問題。

We've talked about this before, but Clubhouse back in the 2020s, like the pandemic era, it was this meteoric rise to success.

我們以前討論過這個問題，但早在 2020 年代，俱樂部會所就像大流行病時代一樣，如日中天。

There was this platform that was just live audio stage events that would disappear after they were over.

有這樣一個平臺，它只是實時音頻舞臺活動，活動結束後就會消失。

But with everybody sitting at home, it blew up in popularity.

但是，由於每個人都坐在家裡，它的受歡迎程度直線上升。

It skyrockets to the top of the app store.

Major new weekly shows were created on it.

新的大型周播節目也在此基礎上誕生。

There were huge, famous people participating in Clubhouses regularly.

有很多名人經常參加俱樂部會所活動。

But fast forward literally one year, maybe two, and suddenly Spotify had built this feature into their app.

但快進一年，也許是兩年後，Spotify 突然在其應用程序中內置了這一功能。

Discord had built the feature into their app.

Discord 在其應用程序中內置了該功能。

Slack 已將其內置到自己的應用程序中。

Even Twitter had built it into their app.

就連 Twitter 也在其應用程序中內置了這一功能。

And so suddenly it was just a feature inside of these larger apps.

於是，它突然就成了這些大型應用程序的一項功能。

So the question became, is Clubhouse, this huge thing that came out of nowhere, is this a product or is it just a feature?

於是問題來了，俱樂部會所這個突然冒出來的龐然大物，到底是產品還是功能？

And it turned out the success of all of the features and all the other apps that built it in meant that Clubhouse would just die.

結果發現，所有功能的成功以及所有其他應用程序的內置，都意味著 Clubhouse 將就此消亡。

And a lot of the other ones eventually ended up dying.

I think we still have Twitter spaces, but generally it just became a feature.

我想我們仍然有 Twitter 空間，但一般來說，它只是成為了一種功能。

And there are actually many examples of these, which is why the product versus feature question still keeps coming up over and over again.

實際上，這樣的例子有很多，這也是為什麼產品與功能的問題仍會反覆出現的原因。

So now the newest question, is AI a product or a feature?

那麼，現在最新的問題是，人工智能是產品還是功能？

Because we got to see what it looked like as a standalone product.

因為我們看到了它作為獨立產品的樣子。

We just had in 2024, we had the Humane AI pin, which was a dedicated piece of hardware specifically for engaging with an AI.

我們剛剛在 2024 年推出了 "人性化人工智能針"，這是一款專門用於與人工智能互動的專用硬件。

We got the Rabbit R1, another device that promised to be like this physical embodiment of an assistant you have everywhere.

我們得到了 Rabbit R1，它是另一款設備，承諾將成為你隨身助理的實體化身。

Now both these devices were bad and they didn't work very well, but let's say they did.

現在，這兩個裝置都很糟糕，它們都沒有很好地工作，但讓我們假設它們確實工作了。

Let's say they actually worked well and were fine.

假設它們真的運行良好，沒有問題。

And you can even consider like going to the ChatGPT website and using it there as AI as a product.

您甚至可以考慮訪問 ChatGPT 網站，將其作為 AI 產品使用。

But then fast forward now to just a few months later, and we just had Apple's WWDC, and we also just had Google IO and very different approach.

但幾個月後的今天，我們剛剛舉辦了蘋果的 WWDC，也剛剛舉辦了谷歌的 IO，兩者的做法截然不同。

Like look what happened with Apple, just like Clockwork, they went through and systematically added all these AI features sprinkled throughout a bunch of its operating systems.

看看蘋果公司的做法，他們就像發條一樣，有條不紊地將所有這些人工智能功能添加到其眾多作業系統中。

So for example, inside of any app with the ability to write text, you now have these writing tools that pop up that are powered by these new language models that can help you summarize or proofread or change the tone or style of your writing.

是以，舉例來說，在任何可以撰寫文字的應用程序中，現在都會彈出這些由新語言模型驅動的寫作工具，它們可以幫助你總結或校對，或改變你寫作的語氣或風格。

And then there's also a new Siri, you know, powered by these language models again.

然後還有一個新的 Siri，你知道，它又是由這些語言模型驅動的。

So it can hold conversations better and understand context better and use a semantic index to parse info about various files and things on your device and bring them into Siri's understanding.

是以，它能更好地進行對話，更好地理解上下文，並使用語義索引來解析設備上的各種文件和事物的資訊，並將它們納入 Siri 的理解範圍。

You can literally generate images as a feature on your device.

But the point is, it's clearly a very different way of thinking about AI for the consumer where it's just one of the features built into the thing that you use.

但問題是，對於消費者來說，這顯然是一種非常不同的人工智能思維方式，人工智能只是你使用的產品內置的功能之一。

Now I realize this isn't a perfect analogy.

現在我意識到這不是一個完美的比喻。

I think probably the biggest flaw being that, you know, when they integrated these features, like when Slack, when Twitter built Spaces, when they built these features, they didn't integrate Clubhouse into those bigger websites.

我認為最大的缺陷可能是，當他們整合這些功能時，比如 Slack、Twitter 建立 Spaces 時，當他們建立這些功能時，他們沒有將 Clubhouse 整合到這些更大的網站中。

They actually just took the idea of what Clubhouse is, which is just a live audio event on stage, and they built it themselves into their own apps.

實際上，他們只是將俱樂部會所的理念，也就是舞臺上的現場音頻活動，融入到自己的應用程序中。

於是，俱樂部會所就這樣壽終正寢了。

But in this specific case with Apple, it's actually a combination of two things.

但在蘋果公司的這個具體案例中，實際上是兩方面因素的結合。

It's them building a bunch of their own models to do a lot of these things on device, but then also them literally building a ChatGPT wrapper into a lot of their OS.

他們建立了大量自己的模型，可以在設備上做很多這樣的事情，但同時也在他們的作業系統中建立了一個 ChatGPT 封裝。

So ChatGPT actually gets more users this way.

是以，ChatGPT 實際上通過這種方式獲得了更多用戶。

So I thought this was fun to think about.

所以，我覺得這個想法很有趣。

Now there is no answer yet as far as is AI actually a feature or a product?

至於人工智能究竟是一種功能還是一種產品，現在還沒有答案。

But I think if history is any indication, I do think that more people in the long run are going to end up using this AI stuff as a feature more than going to like a standalone thing.

但我認為，如果歷史能說明什麼，我確實認為，從長遠來看，更多的人最終會把人工智能作為一種功能來使用，而不是獨立使用。

Like I was looking back for other examples of this, and I honestly found it really hard to find any examples of the other way around happening, where the individual product becomes far more successful than the same idea being baked as a feature into something larger.

就像我回過頭來尋找其他例子一樣，老實說，我發現真的很難找到任何反其道而行之的例子，在這種情況下，單個產品的成功率要遠遠高於將同樣的想法作為一個功能植入到更大的產品中。

Like I think TikTok is maybe the best example of this opposite version, where TikTok, it's a vertical video carousel with an algorithm that learns really quickly what video you want to see next, right?

我認為 TikTok 也許是這種相反版本的最佳範例，TikTok 是一個垂直視頻傳送帶，它的算法能快速瞭解你下一步想看什麼視頻，對嗎？

Like we've seen Instagram Reels develop that exact same thing.

就像我們看到的 Instagram Reels 一樣。

We've got all of YouTube, but inside of YouTube is this vertically scrolling carousel with an algorithm.

我們已經擁有了整個 YouTube，但 YouTube 內部是一個垂直滾動的旋轉木馬，其中包含一種算法。

But I still think today we would say TikTok is the most popular version of that idea.

但我仍然認為，今天我們可以說 TikTok 是這一想法最流行的版本。

So the standalone version is winning currently over it being built in as a feature.

是以，目前獨立版本比內置功能更勝一籌。

And maybe Snapchat is another one, kind of.

Like Snapchat Stories was one of the biggest features in Snapchat, and then that kind of got ripped into being a feature everywhere else.

就像 Snapchat Stories 曾是 Snapchat 最大的功能之一，但後來卻被其他地方的功能所取代。

Like everyone has Stories, but Snapchat by itself is doing well.

就像每個人都有 "故事 "一樣，但 Snapchat 本身也做得很好。

But anyway, the question is, do you have to get all the way to the level of Snapchat or TikTok in order to be successful as a standalone product to defeat the fact that your thing could just become a feature somewhere?

但不管怎麼說，問題是，你是否必須達到 Snapchat 或 TikTok 的水準，才能作為獨立產品獲得成功，從而戰勝你的東西可能只是某個地方的一個功能的事實？

But yeah, that is now the question with AI.

沒錯，這就是人工智能現在面臨的問題。

It just struck me as such a difference in approach between the product and the feature version of it.

這讓我覺得，產品和功能版本在方法上有如此大的差異。

And it also struck me that, yeah, it feels like Rabbit and Humane were kind of doomed from the start because there's no way they would also develop all the other benefits of the big things, those being smartphones.

我還突然意識到，是的，感覺 Rabbit 和 Humane 從一開始就註定要失敗，因為它們不可能同時開發智能手機等大型設備的所有其他優勢。

But also one more quick thing I wanted to highlight that I don't think got as much attention as it maybe deserved, which is all of these models that we've been talking about with WWDC, all the models under the umbrella of Apple Intelligence, the diffusion model, the image generation model, the language models are all built by Apple.

此外，我還想強調一件事，我認為它沒有得到應有的關注，那就是我們在 WWDC 上談論的所有這些模型，蘋果智能公司旗下的所有模型、擴散模型、影像生成模型、語言模型都是由蘋果公司構建的。

Like there was a whole moment on Twitter with a lot of confusion over just how integrated chat GPT is into iOS.

就像 Twitter 上有那麼一瞬間，人們對 GPT 哈拉功能與 iOS 的整合程度感到非常困惑。

I think it's actually not really that integrated at all.

我覺得其實根本沒有那麼融合。

So Apple had to go through the work of obviously making all these models, but also training them all.

是以，蘋果公司不僅要製作所有這些模型，還要對它們進行培訓。

是以，我們就此詢問了蘋果公司。

They've had to go through finding publicly available data and licensing and doing that whole dance and spending the millions and millions of dollars required to make these models work.

他們必須找到公開的數據，獲得許可，完成整個過程，並花費數百萬美元使這些模型發揮作用。

So that is a really higher barrier to entry, but it's only the once in a while that the request is complex enough or deals with enough real world data, which isn't in Apple's training, that it actually asks, okay, can I go out to chat GPT?

是以，這確實是一個較高的准入門檻，但只有偶爾請求足夠複雜或涉及足夠多的真實世界數據（這不在蘋果的培訓範圍內）時，蘋果才會實際詢問："好吧，我能出去聊聊 GPT 嗎？

And it asks every single time, which feels about as unintegrated as it gets.

而且每次都要問，這讓人感覺很不協調。

But generally, I think this is going to be something that's really hard for other companies, for any new companies to pull off.

但總的來說，我認為這對其他公司和任何新公司來說都是很難做到的。

This might be the last big set of models we get to see.

這可能是我們能看到的最後一組大型模型了。

So unless you're Apple or Google or Microsoft or open AI, or any of the other massive ones that are safe because they'll get integrated now, probably don't have much of a shot.

是以，除非你是蘋果、谷歌、微軟或開放式人工智能公司，或者是其他那些因為現在會被整合而很安全的大型公司，否則可能沒有什麼機會。

But either way, this has been really interesting to think about AI.

但無論如何，對人工智能的思考都非常有趣。

Or does it have to be one or the other and one wins and the other loses?

Let me know what you think in the comment section below.

請在下面的評論區發表您的看法。