最近，國外幾家久負盛名的科技巨頭展現了他們的AI雄心。
例如蘋果舉辦WWDC 23，微軟召開Build 23，就連谷歌也在2月份舉辦了搜尋業務大會。
這些巨頭們的動作，無疑彰顯了生成式人工智慧（AIGC）的崛起，也帶動了一群先前對人工智慧不感興趣的團隊、機構。
現在這些大型科技公司全力押注人工智慧，值得注意的幾個標誌是：Google AI、Microsoft Copilot、Apple 機器學習以及OpenAI追求通用人工智慧。
蘋果的機器學習
#蘋果公司似乎對人工智慧這個詞「不感冒」。
在今年的WWDC 上，隻字未提“人工智慧”，以及“ChatGPT”等當前科技界內的一些更為流行的詞彙。蘋果所做的，只是簡單地提及了 7 次「機器學習（ML）」。
即使是在介紹他們準備了 7 年的 AR 眼鏡 Vision Pro 時，也只是表述為「使用了先進的編碼-解碼神經網路」。
實際上，蘋果對人工智慧的理解可能更加準確。他們在機器學習研究方面投入了大量資金，組建了一支由優秀的研究人員和工程師組成的團隊。他們將機器學習應用於各種項目，包括Siri、照片、健康和CarPlay，提升了使用者體驗。
今年，蘋果也更新了基於機器學習的新功能，如Live Text、Visual Look Up和Safety Check，這些功能都運行在iOS 16系統上。顯然，蘋果對人工智慧（機器學習）的重點：是改變使用者互動方式。
唯一有點的遺憾的事情，可能就是蘋果在數據方面不甚有競爭力，畢竟，庫克不只一次說過：我們是對用戶資料不感興趣的公司。的
換句話說，雖然蘋果的行動裝置令人著迷，但它們可能難以與Google、Meta這些佈局雲端運算的公司競爭。
GoogleAI
#最有名的當屬谷歌Brain技術，在人才培養上也有「人工智慧留才計劃（Google AI Residency Program）」。
Google在人工智慧演算法和系統方面有突破性的進展，推動了Google搜尋、Google翻譯和Google照片等基於人工智慧的產品和服務的創建。
GoogleI/O活動中重點強調了Bard，一個與OpenAI的ChatGPT競爭的聊天機器人。
面對Google的做法，一些業界人士認為：Google調整人工智慧運營，優先考慮快速推出產品的做法是「應對性「的，與其過去以創新為導向的做法有所偏離。
Google的母公司Alphabet多年來一直在投資人工智慧，並在2014年收購了DeepMind。最近，Alphabet將其Google研究團隊與DeepMind合併，整合人工智慧工作。然而，一些專家認為，這種整合應該早點進行，因為谷歌在其領先的人工智慧產品上沒有充分利用其優勢，導致在2022年落後於微軟。
Meta和自監督學習
#Meta從2017年的時候，就非常看好自監督學習。這幾年，他們也為業界提供了非常多的自監督學習演算法、框架。
例如，在影像分類和目標偵測等任務中取得了SOTA結果的SimCLR、SwAV和DINO。
Meta在2021年建立了Megatron，一個用於自監督學習演算法訓練的計算集群。在2022年，他們發表了Data2vec論文，介紹了一個跨語音、視覺和文字模態的SSL演算法。
正如Meta的Yann LeCunn多次強調的，他不相信RLHF。
"I think RLHF is hopeless because the space for wrong answers is very large, and wicked problems generally have long-tail distributions that RLHF cannot solve. Any system that does not experience the world and learn on its own will be subject to The data it learns from.”
Microsoft and Copilot
Microsoft has invested heavily in artificial intelligence in recent years, and its Copilot The project is one of the most ambitious examples of such investment. It is a powerful language model that can generate text, translate languages, and assist in a variety of creative tasks.
Copilot’s goal is to change the way people work and create by improving efficiency, inspiring creativity and enhancing inclusivity. Microsoft plans to provide Copilot service for free to Microsoft 365 users and as a standalone product. This tool has the potential to usher in the impact of artificial intelligence on the world, bringing benefits such as increased productivity, improved quality and expanded creativity.
Microsoft has also expanded the application of Copilot to the CRM and ERP fields, launching Dynamics 365 Copilot, and GitHub has also released Github Copilot for Business, which is an artificial intelligence programming for the public assistant.
OpenAGI
OpenAI CEO Sam Altman and other founders talk artificial intelligence on multiple platforms General Intelligence (AGI), with both interest and concern about the benefits and risks it may bring. Altman said in an interview with Lex Fridman that he believes AGI is "probably 10 to 20 years away" and is expected to have a "positive impact" on humanity, while emphasizing the importance of ensuring its responsible use.
Altman also admitted that AGI has some risks, such as being abused or causing massive unemployment. Altman emphasized the need to consider the pros and cons of AGI promptly. He works to develop safety guidelines and build a community of experts to promote responsible use of AGI. Altman's visit reflects growing global interest in AGI. As AGI becomes more of a reality, it’s important to think about its possible benefits and risks. Under Altman's leadership, OpenAI focuses on safe and ethical AGI development.
In a blog post, Altman and the other founders described their expectations for AGI, arguing that it could "solve some of the world's most pressing problems," such as climate change, Poverty and disease can also stimulate human creativity and wisdom.
However, they are also aware of the potential risks of AGI, including being used to create autonomous weapons or causing mass unemployment by replacing human jobs.
Amazon and cloud services
Amazon has invested heavily in artificial intelligence research, and its cloud services are artificial intelligence An important platform for intelligent development and deployment.
They are also committed to providing tools and resources to AI developers. Cloud-based platform SageMaker, for example, makes it easier to build, train and deploy ML models that can be used for a variety of applications such as fraud detection, churn prediction and product recommendations.
The recently released Falcon 40B is a large language model developed on Amazon Web Services (AWS). Falcon 40B is a versatile and robust tool for translation, Q&A, summarization, and image recognition, and is accessible through Amazon SageMaker JumpStart on AWS.
Reference source: //m.sbmmt.com/link/3e1804747c4cf0e9f098b445b1fff36c
以上是從谷歌到亞馬遜，科技巨頭們的AI痴迷的詳細內容。更多資訊請關注PHP中文網其他相關文章！
