人工智能能幹什麼，蘋果也還沒想好｜金融時報

  • 2024-06-16 09:15
  • 外刊看世界
蘋果牌AI大模型的子彈，終於在2024年全球開發者大會（WWDC24）上射出。 
作爲本屆WWDC的壓軸內容，蘋果公司把自己多年對人工智能、機器學習等的開發經驗，與當前的生成式AI和大語言模型相結合的最新成果——Apple Intelligence（蘋果智能）推出，這是一種新型的個人智能系統。 
這個很蘋果的名字，引發了“蘋果再次用自己的方式詮釋了什麼是AI”的吐槽。至於Apple Intelligence，蘋果公司CEO蒂姆·庫克爲開發這套系統列出五大原則：強大的、符合直覺的、深度整合的、懂得個人的以及保護隱私的。 
蘋果公司在原來軟硬件生態的基礎上，賦予AI的最新能力，從而優化用戶的使用體驗，比如AI助手問答、文生圖、文本總結、跨應用聯動等。這些新功能似曾相識，並沒有超出預期，發佈會當天，蘋果股價收跌近2%。而在6月12日，蘋果公司股價開始大漲，市值首次突破3.3萬億美元，超過微軟，重奪全球市值第一的寶座。
請看《金融時報》的詳細報道：
Whenever a new technology comes along, the world has traditionally looked to Apple to know what to do with it. Under Steve Jobs, the company was renowned for converting groundbreaking technologies into the next must-have device or service.
每當一項新技術出現時，世界傳統上都指望蘋果(Apple)知道該如何利用它。在史蒂夫•喬布斯(Steve Jobs)的領導下，蘋果以將突破性技術轉化爲下一代必備設備或服務而聞名。
Not any more. This week’s hotly anticipated unveiling of Apple Intelligence — Apple’s audacious attempt to stamp its own brand on to generative artificial intelligence — was an uneven and unoriginal demonstration that seemed uncharacteristically derivative.
現在不是了。本週備受期待的蘋果智能(Apple Intelligence)發佈會，蘋果試圖將其品牌打造成生成式人工智能的代名詞，但展示效果不夠出色，缺乏創意，顯得不夠獨特。
There is little sense that Apple has edited down the possibilities of generative AI to prioritise the truly useful — the quality that most distinguished the company under Jobs. It may have eased worries that Apple has fallen badly behind in a key technology, but it leaves the question of who will shape the AI future wide open.
幾乎沒有跡象表明，蘋果已經縮減了生成式人工智能的可能性，以優先考慮真正有用的東西——這是喬布斯時代蘋果公司最具特色的品質。儘管這可能緩解了人們對蘋果在關鍵技術上嚴重落後的擔憂，但仍然讓誰將塑造人工智能未來的問題懸而未決。

第一，學習groundbreaking的用法。這個詞做形容詞，表示groundbreaking work involves making new discoveries, using new methods etc，創新的，開創性的，例：groundbreaking research 具開拓性的研究

第二，學習audacious的用法。這個詞做形容詞，表示showing great courage or confidence in a way that is impressive or slightly shocking，大膽的，例：the risks involved in such an audacious operation 如此大膽的行動中存在的風險

第三，學習derivative的用法。這個詞做形容詞，表示not new or invented, but copied or taken from something else – used to show disapproval，非獨創的，模仿他人的，抄襲的〔含貶義〕，例：a derivative text 依樣畫葫蘆的文章

The very versatility of generative AI feeds uncertainty about its likely impact. Will it usher in a more natural way of interacting with computers? Will it be felt mainly as a set of gee-whizz new features that add bells and whistles to the apps we use every day? Or, like OpenAI’s chatbot ChatGPT, will its most important effects be on the way we find information and navigate the broader digital world?
生成式人工智能的多功能性使人們對其可能產生的影響感到不確定。它會引領出一種更自然的與計算機交互方式嗎？它主要會被感知爲一組令人驚歎的新功能，爲我們每天使用的應用程序增添花哨的裝飾嗎？或者，就像OpenAI的聊天機器人ChatGPT一樣，它最重要的影響將體現在我們獲取信息和在更廣泛的數字世界中導航的方式上嗎？
The AI that Apple showed off this week tried to be all of these things. The risk is that, in rushing to make up for lost time, it will not do any of them particularly well.
蘋果本週展示的人工智能試圖成爲所有這些東西。風險在於，在急於彌補失去的時間時，它不會對其中任何一個做得特別好。
Start with the gee-whizz features. These were sprinkled liberally through Apple’s presentation and offer to give users ways to create personalised images and emojis and add automated writing prompts.
從那些令人驚歎的功能開始。這些功能在蘋果的演示中隨處可見，併爲用戶提供了創建個性化圖像和表情符號以及添加自動寫作提示的方式。
But features like these have become familiar in AI services from other companies. It seemed to be an exercise in box-checking. Google has a photo-editing tool that lets you erase things from pictures? We can do that. Microsoft’s new Recall feature keeps track of everything you’ve done on your device so you can find things again easily? Yes, we can do that, too.
但是這些功能在其他公司的AI服務中已經變得很常見。這似乎只是一個勾選方框的練習。谷歌(Google)有一個照片編輯工具，可以讓你從圖片中擦除東西？我們也可以做到。微軟(Microsoft)的新的Recall功能可以追蹤你在設備上所做的一切，以便你可以輕鬆找到東西？是的，我們也可以做到。

第一，學習versatile的用法。這個詞做形容詞，表示having many different uses，有多種用途的，多功能的，萬用的，例：The potato is an extremely versatile vegetable. 馬鈴薯是一種用途極廣的蔬菜。

第二，學習gee-whiz的用法。這個詞做形容詞，表示very good, in a way that is impressive and exciting，絕妙的，很棒的，例：gee-whiz graphics 超棒的圖像

第三，學習personalize的用法。這個詞做及物動詞，表示to design or change something so that it is suitable for a particular person，使〔某物〕符合某人的特定需求，例：All the products can be personalized to the client’s exact requirements. 所有產品都可以根據客戶的具體要求來設計。

Much will depend on how well Apple’s in-house AI models work — an important question, given the challenges such systems have in returning accurate results. But throwing everything at the wall to see what sticks has never been the Apple way.
很大程度上取決於蘋果公司自家的人工智能模型的表現——這是一個重要問題，考慮到這類系統在返回準確結果方面所面臨的挑戰。但是，將一切都試一試，看看哪個有效，從來不是蘋果的做事方式。
When it comes to an AI that can answer questions about the wider world, meanwhile, Apple’s in-house technology is far behind. So this week it announced a deal with OpenAI to bring ChatGPT to its devices, offering users the chance to get answers free of charge.
當涉及到能夠回答關於更廣闊世界問題的人工智能時，蘋果公司的內部技術遠遠落後。因此，本週它宣佈與OpenAI達成協議，將ChatGPT引入其設備，爲用戶提供免費獲取答案的機會。
Yet what might have sounded like a blockbuster alliance looked halfhearted. iPhone users will have to ask a question of Apple’s digital assistant, Siri, which will then ask for permission to put the same question to ChatGPT — hardly the “seamless integration” Apple likes to talk about.
然而，這個可能聽起來像一個轟動的聯盟卻顯得不夠用心。iPhone用戶將不得不向蘋果的數字助手Siri提問，然後Siri會請求許可將同樣的問題提給ChatGPT——這與蘋果喜歡談論的“無縫集成”相去甚遠。
Nor was it clear how this relationship will work in the longer term. Generating AI answers is expensive, and most tech observers expect Apple to eventually charge users for ChatGPT if the feature takes off.
同時，目前還不清楚這種關係在長期內將如何運作。生成AI答案是一項昂貴的工作，大多數技術觀察家預計，如果ChatGPT功能受歡迎，蘋果最終會向用戶收費。

第一，學習in-house的用法。這個詞做形容詞，表示working within a company or organization，公司[組織]內部的；在機構內部，例：We have an in-house training unit. 我們有一個內部培訓部。

第二，學習free of charge的用法。這個詞組表示with no cost，免費，例：Delivery is free of charge. 免費送貨。

第三，學習halfhearted的用法。這個詞做形容詞，表示done without much effort and without much interest in the result，半心半意的，敷衍了事的，不熱心的，例：Congress has made half-hearted attempts at finance reform. 國會嘗試了一下金融改革，卻沒有盡力而爲。

The arrangement also underlines the uncomfortable fact that Apple’s in-house AI is inferior. If the most advanced AI systems really do reach human-level intelligence then they could well become the core tech platforms on which future apps depend. For Apple, catching up will be key.
這一安排還凸顯了一個令人不悅的事實，即蘋果公司內部的人工智能技術相對較差。如果最先進的人工智能系統真的能達到人類水平的智能，那麼它們很可能會成爲未來應用程序所依賴的核心技術平臺。對於蘋果來說，迎頭趕上將是關鍵。
The possibility that AI will bring a more natural way to handle our everyday digital activities, is where Apple has the best chance to make an impact in the short term. It promises a souped-up Siri that will be able to pick through information in your email, messages and various apps to answer questions or take actions on your behalf.
AI有可能帶來一種更自然的方式來處理我們的日常數字活動，這是蘋果在短期內產生影響的最佳機會。它承諾提供一款升級版的Siri，能夠瀏覽您的電子郵件、信息和各種應用程序中的信息，以回答問題或代表您採取行動。
While this idea sounds deceptively simple, pulling it off will be hard. AI systems are probabilistic, meaning they make their best guess at returning the right answer. Apple has yet to show how well the new Siri works.
儘管這個想法聽起來簡單，但要實現它並不容易。人工智能系統是概率性的，意味着它們會盡力猜測出正確答案。蘋果尚未展示新版Siri的表現如何。
This raises broader questions about the future of technology. If Siri becomes the on-ramp to everything you do on an iPhone, reducing the need to open apps, what does that do to the many developers whose businesses depend on building direct relationships with Apple’s users?
這引發了關於技術未來的更廣泛的問題。如果Siri成爲iPhone上您所做的一切的入口，減少了打開應用程序的需求，那麼對於那些依賴於與蘋果用戶建立直接關係的許多開發者來說，這會產生什麼影響呢？
This is the same upheaval threatened by the latest AI feature in Google’s search engine, which aims to summarise information plucked from other websites. If AI systems run by big tech companies can filter the digital world in this way — and if they can use that information to start taking actions on our behalf — then what future is there for all the websites, apps and other software tools that we currently rely on?
谷歌搜索引擎的最新人工智能功能也面臨着同樣的動盪威脅，該功能旨在彙總從其他網站摘取的信息。如果大型科技公司運營的人工智能系統能以這種方式過濾數字世界，如果它們能利用這些信息開始代表我們採取行動，那麼我們目前依賴的所有網站、應用程序和其他軟件工具還有什麼前途可言？
Relief that Apple is finally bringing generative AI to its devices added more than $260bn to its stock market value this week. Over the next few years, Apple AI could fuel a massive hardware upgrade cycle. It will only work on a very small number of existing devices, so most customers will have to buy a new and more powerful iPhone, iPad or Mac to see Apple Intelligence.
蘋果終於將生成式人工智能引入其設備，這爲其股票市值增加了2600億美元。在接下來的幾年裏，蘋果的人工智能將推動大規模的硬件升級週期。它只能在極少數現有設備上運行，所以大多數顧客將不得不購買一部新的更強大的iPhone、iPad或Mac才能體驗蘋果智能。
But it doesn’t mean Apple has yet found the key that will unlock the full potential of generative AI.
但這並不意味着蘋果已經找到了能夠釋放生成式人工智能的全部潛力的關鍵。
外刊文章來源，2024年6月19日，金融時報，Even Apple cannot explain why we need AI in our lives
每天一測，模擬答題：
2025考研er，關注公衆號進入閱讀 ↓↓

掌握語言，是爲了換一個視角看世界
也許長，但必定值得耐心學習
願你看待這個星球的眼光能夠批判且不同
您怎麼看？