全球人工智能教育及研究領(lǐng)導(dǎo)者、DeepLearning.AI創(chuàng)始人吳恩達，4月21日于知乎發(fā)布本文。

Dear friends,

The competitive landscape of large language models (LLMs) is evolving quickly. The ultimate winners are yet to be determined, and already the current dynamics are exciting. Let me share a few observations, focusing on direct-to-consumer chat interfaces and the LLM infrastructure and application layers.

First, ChatGPT is a new category of product. It’s not just a better search engine, auto-complete, or something else we already knew. It overlaps with other categories, but people also use it for entirely different purposes such as writing and brainstorming. Companies like Google and Microsoft that are integrating LLMs into existing products may find that the complexity of switching not only technologies but also product categories raises unique challenges.

OpenAI is clearly in the lead in offering this new product category, and ChatGPT is a compelling direct-to-consumer product. While competitors are emerging, OpenAI’s recent move to have ChatGPT support third-party plugins, if widely adopted, could make its business much more defensible, much like the app stores for iOS and Android helped make those platforms very defensible businesses.

Second, the LLM infrastructure layer, which enables developers to interact with LLMs via an API, looks extremely competitive. OpenAI/Microsoft leads in this area as well, but Google and Amazon have announced their own offerings, and players such as Hugging Face, Meta, Stability AI, and many academic institutions are busy training and releasing open source models. It remains to be seen how many applications will need the power of the largest models, such as GPT-4, versus smaller (and cheaper) models offered by cloud providers or even hosted locally, like gpt4all, which runs on a desktop.

Finally, the application layer, in which teams build on top of LLMs, looks less competitive and full of creativity. While many teams are piling onto “obvious” ideas — say, building question-answering bots or summarizers on top of online content — the sheer diversity of potential LLM-powered applications leaves many ideas relatively unexplored in verticals including specialized coaching and robotic process automation. AI Fund, the venture studio I lead, is working with entrepreneurs to build applications like this. Competition feels less intense when you can identify a meaningful use case and go deep to solve it.

LLMs are a general-purpose technology that’s making many new applications possible. Taking a lesson from an earlier era of tech, after the iPhone came out, I paid $1.99 for an app that turned my phone into a flashlight. It was a good idea, but that business didn’t last: The app was easy for others to replicate and sell for less, and eventually Apple integrated a flashlight into iOS. In contrast, other entrepreneurs built highly valuable and hard-to-build businesses such as AirBnB, Snapchat, Tinder, and Uber, and those apps are still with us. We may already have seen this phenomenon in generative AI: Lensa grew rapidly through last December but its revenue run appears to have collapsed.

Today, in a weekend hackathon, you can build a shallow app that does amazing things by taking advantage of amazing APIs. But over the long term, what excites me are the valuable solutions to hard problems that LLMs make possible. Who will build generative AI’s lasting successes? Maybe you!

One challenge is that the know-how for building LLM products is still evolving. While academic studies are important, current research offers a limited view of how to use LLMs. As the InstructGPT paper says, “Public NLP datasets are not reflective of how our language models are used. . . . [They] are designed to capture tasks that are easy to evaluate with automatic metrics.”

In light of this, community is more important than ever. Talking to friends who are working on LLM products often teaches me non-intuitive tricks for improving how I use them. I will continue trying to help others wherever I can.

Keep learning!

Andrew

親愛的朋友們，

大型語言模型 (LLMs) 的競爭格局正在迅速打開。最終贏家尚未出爐，但目前的形勢已經(jīng)令人興奮。我想分享一些觀察結(jié)果，重點關(guān)注直接面向消費者的聊天接口以及LLMs基礎(chǔ)設(shè)施和應(yīng)用程序?qū)印?/p>

首先，ChatGPT是一個新的產(chǎn)品類別。它不僅僅是一個更好的搜索引擎——能自動完成檢索，及其他我們已經(jīng)知道的功能。ChatGPT與其他類別有一些重疊，但人們也將其用于了完全不同的目的，如寫作和頭腦風(fēng)暴。谷歌和微軟等公司正在將LLMs集成到現(xiàn)有產(chǎn)品中，這樣做可能不僅需要轉(zhuǎn)換技術(shù)，還要轉(zhuǎn)換產(chǎn)品類別，這就帶來了獨特的挑戰(zhàn)。

OpenAI在提供這種新的產(chǎn)品類別方面顯然處于領(lǐng)先地位，ChatGPT就是一種引人注目的直接面向消費者的產(chǎn)品。雖然競爭對手不斷涌現(xiàn)，但OpenAI最近讓ChatGPT支持第三方插件的舉措——一旦被廣泛采用，可能會使其業(yè)務(wù)更具防御性——會像iOS和Android的應(yīng)用商店使這些平臺的業(yè)務(wù)更具防御性一樣。

其次，LLMs的基礎(chǔ)設(shè)施層使開發(fā)人員能夠通過API與LLMs進行交互，這看起來極具競爭力。OpenAI和微軟在這一領(lǐng)域也處于領(lǐng)先地位，谷歌和亞馬遜也爭相發(fā)布了自己的產(chǎn)品，而Hugging Face, Meta, Stability AI等公司和許多學(xué)術(shù)機構(gòu)都在忙著訓(xùn)練和發(fā)布開源模型。有多少應(yīng)用程序需要用到像GPT-4這樣的最大型模型，而不是云提供商提供的更?。ǜ阋耍┑哪Ｐ?，甚至是本地托管的模型（比如運行在桌面上的gpt4all）還有待觀察。

最后是應(yīng)用程序?qū)印ｉ_發(fā)團隊建立在LLMs的基礎(chǔ)上，看起來競爭不那么激烈，且充滿創(chuàng)造力。雖然許多團隊都在嘗試“顯而易見”的想法——比如在在線內(nèi)容的基礎(chǔ)上構(gòu)建問答機器人或摘要器。但LLMs支持的潛在應(yīng)用程序的多樣性，使得許多想法在專業(yè)指導(dǎo)和機器人過程自動化等垂直領(lǐng)域還未被充分探索。我領(lǐng)導(dǎo)的風(fēng)投公司AI Fund正在與企業(yè)家合作開發(fā)這樣的應(yīng)用程序。當你能夠確定一個有意義的用例并深入解決它時，競爭的感覺就不那么激烈了。

LLMs是一種通用技術(shù)，它使許多新的應(yīng)用成為可能。在iPhone問世后，我從早期科技時代吸取了教訓(xùn)花費1.99美元購買了一個能把手機變成手電筒的應(yīng)用程序。這是個好主意，但這筆生意沒能持續(xù)多久：這款應(yīng)用很容易被其他人復(fù)制，售價也更低，最終蘋果將手電筒集成到了iOS系統(tǒng)中。相比之下，其他企業(yè)家建立了價值更高和開發(fā)難度更大的業(yè)務(wù)，如AirBnB、Snapchat、Tinder和Uber，這些應(yīng)用程序至今仍在被使用。我們可能已經(jīng)在生成式人工智能中看到了這種現(xiàn)象：Lensa（一款火爆的照片編輯器）在去年12月的使用量增長迅速，但收入?yún)s不盡如人意。

現(xiàn)在，你可以在一個周末進行的黑客馬拉松中構(gòu)建一個簡單的應(yīng)用程序，通過利用厲害的API來實現(xiàn)驚人的結(jié)果。但從長遠來看，令我興奮的是LLMs能為解決難題提供有價值的解決方案。誰將打造生成式人工智能的長期成功？也許就是你！

我們面臨的一個挑戰(zhàn)是，構(gòu)建LLMs產(chǎn)品的技術(shù)訣竅仍在不斷發(fā)展。雖然學(xué)術(shù)研究很重要，但目前的研究對如何使用LLMs只提供了有限的幫助。正如InstructGPT論文所說，“公共NLP數(shù)據(jù)集并不能反映我們的語言模型是如何被使用的……（它們）被設(shè)計用來捕捉那些容易用自動指標進行評估的任務(wù)?！?/p>

鑒于此，社群的作用比以往任何時候都更加重要。與從事LLMs產(chǎn)品開發(fā)工作的朋友交談能帶給我一些直覺以外的技巧來改進對這些產(chǎn)品的使用。我將繼續(xù)盡我所能去幫助別人。

請不斷學(xué)習(xí)！

吳恩達