OpenAI 宣布推出迄今為止功能最強大、最一致的 AI 模型 GPT-4o1

該模型目前可供 ChatGPT 訂閱者和 API 用戶使用。這個新模型帶來了許多新的可能性，我們很可能會在使用者的創作中看到這些可能性的實現。

OpenAI has recently announced the launch of its latest AI model, GPT-4o1. This new model boasts improved reasoning capabilities, which is why it takes a noticeable amount of time to respond—roughly equivalent to how long it would take a human to think through a problem before answering. The model is designed for applications in physics, chemistry, biology, and coding and has outperformed its predecessors in several benchmark tests in these domains.

OpenAI 最近宣布推出其最新的人工智慧模型 GPT-4o1。這種新模型擁有改進的推理能力，這就是為什麼它需要相當長的時間才能做出回應——大致相當於人類在回答之前思考問題所需的時間。該模型專為物理、化學、生物學和編碼領域的應用而設計，在這些領域的多項基準測試中表現優於其前身。

According to OpenAI, GPT-4o1 is also safer than previous models. When subjected to a user safety test, GPT-4o1 scored 89 out of 100, while its predecessor, GPT-4o, scored only 13 out of 100.

據 OpenAI 稱，GPT-4o1 也比以前的模型更安全。在進行使用者安全測試時，GPT-4o1 的得分為 89 分（滿分 100 分），而其前身 GPT-4o 的得分僅為 13 分（滿分 100 分）。

However, this new model is not without its limitations. Despite being more advanced and secure, GPT-4o1 cannot browse the web or work off, from, or with user-uploaded images.

然而，這種新模式並非沒有限制。儘管更加先進和安全，GPT-4o1 無法瀏覽網頁或使用用戶上傳的圖像進行工作。

Here is o1, a series of our most capable and aligned models yet:https://t.co/yzZGNN8HvDo1 is still flawed, still limited, and it still seems more impressive on first use than it does after you spend more time with it. pic.twitter.com/Qs1HoSDOz1

這是o1，我們迄今為止最強大、最一致的一系列模型：https://t.co/yzZGNN8HvDo1 仍然有缺陷，仍然有限，而且它在第一次使用時看起來仍然比在你花了更多時間使用它之後更令人印象深刻。 pic.twitter.com/Qs1HoSDOz1

The model is currently available to ChatGPT subscribers and API users. With this new model comes many new possibilities that we will likely see come to fruition in user creations. But one thing I wonder is, why does the AI model replicate human thinking in terms of the time it takes to respond? I am sure it comes with benefits and also introduces an element of realism to artificial intelligence, but most people are not using AI chatbots to access a counterpart that resembles a human; most are using it to access accurate information quickly.

該模型目前可供 ChatGPT 訂閱者和 API 用戶使用。這個新模型帶來了許多新的可能性，我們很可能會在使用者的創作中看到這些可能性的實現。但我想知道的一件事是，為什麼人工智慧模型在回應時間方面會複製人類的思維？我確信它帶來了好處，也為人工智慧引入了現實主義元素，但大多數人並沒有使用人工智慧聊天機器人來存取類似於人類的對應對象；大多數人都使用它來快速獲得準確的資訊。

When I gave GPT-4o1 a try, it took 39 seconds to think through the prompt I gave it, but it did not even give me an output! I had to prompt the model a second time to get the output I desired, and then it took 36 more seconds to think—1 minute and 5 seconds total—to give me a response for a set of fairly simple instructions.

當我嘗試 GPT-4o1 時，我花了 39 秒思考我給的提示，但它甚至沒有給我輸出！我必須再次提示模型以獲得我想要的輸出，然後又花了 36 秒的時間思考（總共 1 分 5 秒）才給我一組相當簡單的指令的回應。

OpenAI is not losing a customer in me for that wait time, and I understand the trade-off of having better information even though it delivered at a slower rate. Still, the task I gave the chatbot was fairly simple, and the one-minute five-second wait time felt long. That being said, I can’t imagine how long the chatbot takes to think and respond when given a task that is actually complex.

OpenAI 並沒有因為等待時間而失去我的客戶，而且我理解獲得更好的資訊的權衡，即使它以較慢的速度傳遞。不過，我給聊天機器人的任務相當簡單，一分五秒的等待時間感覺很長。話雖這麼說，我無法想像當給出一個實際上很複雜的任務時，聊天機器人需要多長時間來思考和回應。

Another tech giant, Google (NASDAQ: GOOGL), is under scrutiny in the EU for its AI model. Ireland’s Data Protection Commission has opened an inquiry into Google’s Pathway Language Model 2 (PaLM 2) regarding how the model handles personal data. Since Google’s European headquarters are located in Dublin, the Irish watchdog acts as the company’s lead regulator for the bloc’s privacy rulebook, which is why they were the entity to raise the issue. The inquiry will inspect whether Google breached its obligations under GDPR regarding how it processed citizens’ personal data in the EU.

另一家科技巨頭Google（納斯達克股票代碼：GOOGL）的人工智慧模型正在歐盟接受審查。愛爾蘭資料保護委員會已對 Google 的 Pathway Language Model 2 (PaLM 2) 展開調查，以了解該模型如何處理個人資料。由於Google的歐洲總部位於都柏林，愛爾蘭監管機構充當該公司歐盟隱私規則手冊的主要監管者，這就是他們成為提出該問題的實體的原因。此次調查將檢查谷歌在處理歐盟公民個人資料的方式方面是否違反了 GDPR 規定的義務。

Tech giants facing obstacles when operating in other countries is not a new issue; previously, Meta (NASDAQ: META) had to suspend the use of its AI model in Brazil for similar privacy concerns that the Brazilian regulator had about its AI models. In many cases, it’s an uphill battle for these AI companies to operate in many places worldwide. Data, including personal data, is crucial when it comes to training and optimizing AI models. At the same time, many countries have strict rules regarding how a company can use consumer data, which tend to be in stark contrast to how the tech giants would like to use the data in relation to their models.

科技巨頭在其他國家運作時面臨障礙並不是一個新問題。先前，由於巴西監管機構對其人工智慧模型的隱私擔憂，Meta（納斯達克股票代碼：META）不得不暫停在巴西使用其人工智慧模型。在許多情況下，這些人工智慧公司要在全球許多地方開展業務是一場艱苦的戰鬥。在訓練和優化人工智慧模型時，包括個人資料在內的資料至關重要。同時，許多國家對公司如何使用消費者資料有嚴格的規定，這往往與科技巨頭如何使用與其模型相關的數據形成鮮明對比。

At Apple’s GlowTime event, it showcased the latest version of the iPhone—the iPhone 16—and the AI features that will be coming to the phone. The company also emphasized that these new phones have a new chip at their core that specializes in handling the AI operations that the company has deemed “Apple Intelligence.”

在 Apple 的 GlowTime 活動中，它展示了最新版本的 iPhone（iPhone 16）以及即將在手機上推出的 AI 功能。該公司還強調，這些新手機的核心有一個新晶片，專門處理該公司所謂的「蘋果智慧」的人工智慧作業。

Apple (NASDAQ: AAPL) describes Apple Intelligence as “the personal intelligence system that helps you communicate, work, and express yourself. It combines generative models with your personal context to deliver intelligence that is most useful and relevant to you.”

蘋果（NASDAQ：AAPL）將 Apple Intelligence 描述為「幫助您溝通、工作和表達自己的個人智慧系統。它將生成模型與您的個人背景相結合，以提供對您最有用和最相關的情報。

These Apple Intelligence features will be available on all iPhone 16 models, the iPhone 15 Pro and 15 Pro Max, as well as iPads and Macs. But more importantly, Apple Intelligence is software-dependent and will only be available on iOS 18.1, which is expected to launch in October.

這些 Apple Intelligence 功能將在所有 iPhone 16 機型、iPhone 15 Pro 和 15 Pro Max 以及 iPad 和 Mac 上提供。但更重要的是，Apple Intelligence 依賴軟體，並且只能在預計 10 月推出的 iOS 18.1 上使用。

The upcoming AI features include improvements to Siri, custom writing tools, photo and video enhancement tools, natural language processing features, and more—but is AI branding really moving the needle for consumers?

即將推出的 AI 功能包括 Siri 的改進、自訂書寫工具、照片和影片增強工具、自然語言處理功能等等，但 AI 品牌真的能為消費者帶來好處嗎？

Recent research has shown that branding a product as an

最近的研究表明，將產品品牌化為