A conversation with Kevin Scott: What’s next in AI
A conversation with Kevin Scott: What’s next in AI
与 Kevin Scott 对话：人工智能的未来展望
Artificial intelligence systems powered by large language models today are transforming how people work and create, from generating lines of code for software developers to sketches for graphic designers. Kevin Scott, Microsoft’s chief technology officer, expects these AI systems to continue to grow in sophistication and scale—from helping address global challenges such as climate change and childhood education to revolutionizing fields from healthcare and law to materials science and science fiction. 如今，由大型语言模型驱动的人工智能系统正在改变人们的工作和创作方式，从为软件开发人员生成代码行，到为图形设计师绘制草图，无所不包。微软首席技术官 Kevin Scott 预计，这些人工智能系统将在复杂性和规模上持续增长——从帮助应对气候变化和儿童教育等全球性挑战，到彻底改变医疗保健、法律、材料科学和科幻小说等领域。
Scott recently shared his thoughts with us on the impact of AI for knowledge workers and what’s next in AI. The biggest takeaways: Advances in large AI models and generative AI will continue to boost productivity, creativity and satisfaction. AI will enable scientific breakthroughs and help the world solve some of its biggest challenges. As these models become platforms and Microsoft continues to responsibly scale AI advancements for customers, the cloud, infrastructure investments and a strong responsible AI approach are critical. Scott 最近与我们分享了他对人工智能如何影响知识工作者以及人工智能未来发展的看法。核心要点包括：大型人工智能模型和生成式人工智能的进步将持续提升生产力、创造力和工作满意度。人工智能将推动科学突破，并帮助世界解决一些最严峻的挑战。随着这些模型逐渐成为平台，且微软持续为客户负责任地扩展人工智能技术，云服务、基础设施投资以及稳健的“负责任人工智能”方针至关重要。
In your mind, what were some of the most important advancements in AI this year? When we were heading into 2022, I think just about everybody in AI was anticipating really impressive things to take place over the next twelve or so months. But now that we’re pretty much through the year, and even with those lofty expectations, it’s kind of genuinely mind-blowing to look back at the magnitude of innovation that we saw left-to-right in AI. The things that researchers and other folks have done to advance the state-of-the-art are just light years beyond what we thought possible even a few years ago. And almost all of this is a result of the incredibly rapid advancement that has happened with large AI models. 在你看来，今年人工智能领域最重要的进展有哪些？在进入 2022 年时，我想人工智能领域的每个人都预料到未来十二个月左右会发生非常令人印象深刻的事情。但现在我们几乎走完了这一年，即便有着如此高的期望，回首这一年人工智能领域全方位的创新规模，依然令人感到震撼。研究人员和其他从业者在推动技术前沿方面所做的工作，比我们几年前认为可能的程度领先了数光年。而这一切几乎都是大型人工智能模型极其快速进步的结果。
The three things I’ve been most impressed by this year were the launch of GitHub Copilot, which is a large language model-based system that turns natural language prompts into code and has this dramatic positive impact on developer productivity. It opens up coding to a much broader group of people than we’ve ever had before, which is awesome because so much of the future is dependent on our ability to write software. 今年最让我印象深刻的三件事：首先是 GitHub Copilot 的发布。这是一个基于大型语言模型的系统，能将自然语言提示转化为代码，并对开发人员的生产力产生了巨大的积极影响。它让编程向比以往任何时候都更广泛的人群开放，这非常棒，因为未来的很大一部分都取决于我们编写软件的能力。
The second thing is these generative image models such as DALL∙E 2 that have become very popular and more accessible. A fairly high degree of skill is required to sketch and draw and master all of the tools of graphic design, illustration and art. An AI system such as DALL∙E 2 doesn’t turn ordinary people into professional artists, but it gives a ton of people a visual vocabulary that they didn’t have before—a new superpower they didn’t think they would ever have. (Editor’s note: All images in this post except for Kevin Scott’s photograph were generated by a producer using DALL∙E 2.) 第二件事是像 DALL∙E 2 这样变得非常流行且更易于使用的生成式图像模型。素描、绘画以及掌握图形设计、插画和艺术的所有工具需要相当高的技能。像 DALL∙E 2 这样的人工智能系统并不能让普通人瞬间成为专业艺术家，但它赋予了大量人群一种他们以前不具备的视觉词汇——一种他们从未想过自己会拥有的新“超能力”。（编者注：本文中除 Kevin Scott 的照片外，所有图片均由制作人员使用 DALL∙E 2 生成。）
We’ve also seen that AI models are becoming more powerful and delivering even more substantial gains for the problems that they’re being used to solve. I think the work on protein folding this year has been really good throughout the technology industry, including the work that we’ve done with David Baker’s laboratory at the University of Washington, the Institute for Protein Design with RoseTTAFold, and helping that with a bunch of advanced AI to do transformational things. And so that’s just tremendously exciting. Anything that’s a force multiplier on science and medicine is just net beneficial to the world because those are where some of our biggest, nastiest problems live. That’s a big, impressive year. And I think next year will be better. 我们还看到，人工智能模型正变得越来越强大，在解决各种问题时带来了更显著的收益。我认为今年整个科技行业在蛋白质折叠方面的工作非常出色，包括我们与华盛顿大学 David Baker 实验室、蛋白质设计研究所合作的 RoseTTAFold 项目，并利用大量先进的人工智能技术来完成变革性的工作。这令人无比兴奋。任何能成为科学和医学领域“力量倍增器”的事物，对世界都是净收益，因为我们最棘手、最严峻的问题往往就存在于这些领域。这是伟大且令人印象深刻的一年，我相信明年会更好。
Where do you see AI technology having the greatest impact next year and beyond? I think with some confidence I can say that 2023 is going to be the most exciting year that the AI community has ever had. And I say that after really, genuinely believing that 2022 was the most exciting year that we’d ever had. The pace of innovation just keeps rolling in at a fast clip. I talked about GitHub Copilot already, and it’s amazing. But it’s the tip of the iceberg for what large AI models are going to be able to do going forward—extrapolating the same idea to all kinds of different scenarios for how they can assist in other kinds of intellectual labor beyond coding. 你认为人工智能技术在明年及未来几年将在哪里产生最大的影响？我可以自信地说，2023 年将是人工智能社区有史以来最令人兴奋的一年。我之所以这么说，是因为我真心认为 2022 年已经是有史以来最令人兴奋的一年了。创新的步伐正在快速推进。我已经谈到了 GitHub Copilot，它非常了不起。但这只是大型人工智能模型未来能力的一角——将同样的理念推演到各种不同的场景中，看看它们如何协助编程以外的其他智力劳动。
The entire knowledge economy is going to see a transformation in how AI helps out with repetitive aspects of your work and makes it generally more pleasant and fulfilling. This is going to apply to almost anything—designing new molecules to create medicine, making manufacturing “recipes” from 3D models, or simply writing and editing. For example, I’ve been playing around with an experimental system I built for myself using GPT-3 designed to help me write a science fiction book, which is something that I’ve wanted to do since I was a teenager. I have notebooks full of synopses I’ve created for theoretical books, describing what the books are about and the universes where they take place. With this experimental tool, I have been able to get the logjam broken. 整个知识经济将经历一场变革，人工智能将帮助处理工作中重复性的部分，使工作变得更加轻松和充实。这几乎适用于任何领域——设计用于制造药物的新分子、从 3D 模型制作制造“配方”，或者仅仅是写作和编辑。例如，我一直在尝试使用 GPT-3 为自己构建一个实验系统，旨在帮助我写一本科学幻想小说，这是我从青少年时期就想做的事情。我的笔记本里写满了为理论书籍创作的梗概，描述了书籍的内容以及它们发生的世界。有了这个实验工具，我终于打破了创作瓶颈。
When I wrote a book the old-fashioned way, if I got 2,000 words out of a day, I’d feel really good about myself. With this tool, I’ve had days where I can write 6,000 words in a day, which for me feels like a lot. It feels like a qualitatively more energizing process than what I was doing before. This is the “copilot for everything” dream—that you would have a copilot that could sit alongside you as you’re doing any kind of cognitive work, helping you not just get more done, but also enhancing your creativity in new and exciting ways. 当我用老式方法写书时，如果一天能写出 2000 字，我会感觉非常好。有了这个工具，我有时一天能写 6000 字，这对我来说感觉很多。这感觉比我以前所做的过程在质量上更具活力。这就是“万物副驾驶”（copilot for everything）的梦想——当你进行任何认知工作时，身边都有一个副驾驶，它不仅能帮你完成更多工作，还能以新颖且令人兴奋的方式增强你的创造力。
This increase in productivity is clearly a boost to your satisfaction. Why do these tools bring more joy to work? All of us use tools to do our work. Some of us really enjoy acquiring the tools and mastering them and figuring out how to deploy them in a super effective way to do the thing that we’re trying to do. I think that is part of what’s going on here. In many cases, people now have new and interesting and fundamentally more effective tools than they’ve had before. We did a study that found… 这种生产力的提高显然提升了你的满意度。为什么这些工具能给工作带来更多乐趣？我们所有人都使用工具来完成工作。有些人非常喜欢获取工具、掌握它们，并研究如何以超高效的方式部署它们来完成我们想要做的事情。我认为这就是原因的一部分。在许多情况下，人们现在拥有了比以往更新颖、更有趣且从根本上更有效的工具。我们做了一项研究，发现……