18 Matching Annotations
  1. Mar 2021
    1. Singer’s demo caused quite the stir and for good reason: tools such as this will continue to improve, and we all know it. It signals a future where many of the tasks we perform now will be automated or at least aided by AI. Technology that has the potential to eliminate tasks, and by extension the potential to eliminate jobs, will invariably elicit some fear. This is especially true in an industry that already has a firm grasp on how technology can disrupt entire industries. While AI will no doubt eliminate the need to manually perform many tasks, its benefits will also open up opportunities and help to automate menial tasks which will free us up to focus on the more meaningful work that provides additional value. The question becomes: how might we leverage AI and the benefits it brings to help in the design process? Additionally, how might things go wrong when leveraging AI in the design process?

      Singer的演示引起了不小的轰动,原因很充分:像这样的工具将继续改进,我们都知道这一点。它预示着未来我们现在所执行的许多任务都将实现自动化,或者至少有人工智能的辅助。有可能消除任务的技术,进而有可能消除工作岗位,总会引起一些恐惧。在一个已经牢牢把握技术如何颠覆整个行业的行业,这一点尤其重要。人工智能无疑将消除人工执行许多任务的需求,但它的好处也将带来机会,并帮助实现琐碎任务的自动化,这将使我们腾出时间专注于提供额外价值的更有意义的工作。问题变成了:我们如何利用人工智能及其带来的好处来帮助设计过程?此外,在设计过程中利用人工智能时,可能会出现怎样的问题?

  2. Feb 2021
    1. Vicarious的目标定位于“建立下一代的人工智能算法”。并且声称要构建“像人类一样思考的软件”,实现“人脑级别的视觉、语言和自动控制系统”,致力于研究通用人工智能,目前他们的研究重点是实现人工视觉识别系统。

      【独家】扎克伯格、马斯克、贝索斯、彼得.蒂尔都投资的硅谷最神秘人工智能公司Vicarious在干什么?| AI严肃说

    1. Artificial intelligence has already changed the world in some pretty dramatic ways and will certainly do even more so in the future. But it’s a component technology. The transistor has changed the world. But saying, “How are transistors going to change the world?” is almost the wrong layer of abstraction—it’s like trying to understand a river by talking about H20. So artificial intelligence will participate meaningfully in causing technologies to become more intelligent and will shift how we try to deliver value to people. Less and less will technologies need to do what we want them to do through straightforward mechanical and structural solutions and more and more they’ll solve the problems through the increment of intelligence.

      人工智能已经以一些非常引人注目的方式改变了世界,在未来肯定会做的更好。但他是一种组件技术。晶体管已经改变了世界,但是说“晶体管会如何改变世界”几乎是错误的抽象层面,这就好比通过讨论H2O来理解一条河。所以AI将对科技智能化的提升起到重要的作用,同时也会改变我们给人们传递价值的方式。科技将越来越少地直接依靠机械化结构化的解决方式来完成我们交代的任务,而是越来越依靠智能的提升来解决问题。

    1. We’ve described a third view, in which AIs actually change humanity, helping us invent new cognitive technologies, which expand the range of human thought. Perhaps one day those cognitive technologies will, in turn, speed up the development of AI, in a virtuous feedback cycle:

      我们描述了第三种观点,AI实际上改变了人类,帮助我们发明了新的认知技术,扩展了人类思维的范围。或许有一天,这些认知技术将反过来加速AI的发展,形成良性循环:

    2. Examples such as those described in this essay suggest that AI systems can enable the creation of new cognitive technologies. Things like the font tool aren’t just oracles to be consulted when you want a new font. Rather, they can be used to explore and discover, to provide new representations and operations, which can be internalized as part of the user’s own thinking. And while these examples are in their early stages, they suggest AI is not just about cognitive outsourcing. A different view of AI is possible, one where it helps us invent new cognitive technologies which transform the way we think.

      本文描述的例子表明,AI系统推动了新的认知技术的发明。字体工具不仅仅是当你需要一个新字体时可以咨询的预言家。而且,它们可以被用于探索和发现,提供新的表示和操作,能够被内化为用户思考的一部分。虽然这些例子只处于早期阶段,但是它们预示着AI不仅仅是关于认知外包。对于AI的一个不同观点是,它帮助我们发明新的认知技术,转换我们思考的方式。

    3. This essay describes a new field, emerging today out of a synthesis of AI and IA. For this field, we suggest the name artificial intelligence augmentation (AIA): the use of AI systems to help develop new methods for intelligence augmentation. This new field introduces important new fundamental questions, questions not associated to either parent field. We believe the principles and systems of AIA will be radically different to most existing systems.

      本文描述了一个新的领域,这个领域来自于AI和IA的综合。我们建议将这个领域命名为人工智能增强(artificial intelligence augmentation,简称AIA):使用AI系统帮助开发智能增强(IA)的新方法。这个新领域引入了新的重要的基础问题,这些问题无法关联到任何的父领域中。我们相信 AIA 的原理和系统将会与大多数存在的系统完全不同。

    1. 卡耐基梅隆大学的 Ryan Steed 和乔治华盛顿大学的 Aylin Caliskan 两位研究者发表的篇论文《无监督的方式训练的图像表示法包含类似人类的偏见》

      研究者对 OpenAI 在 GPT-2 基础上开发的 iGPT,和 Google的 SimCLR,这两个在去年发表的图像生成模型进行了系统性的测试,发现它们在种族、肤色和性别等指标上几乎原样复制了人类测试对象的偏见和刻板印象。

      在其中一项测试中,研究者用机器生成的男女头像照片作为底板,用 iGPT 来补完(生成)上半身图像。

      最为夸张的事情发生了:在所有的女性生成结果当中,超过一半的生成图像穿着的是比基尼或低胸上衣;

      而在男性结果图像中,大约42.5%的图像穿的是和职业有关的上衣,如衬衫、西装、和服、医生大衣等;光膀子或穿背心的结果只有7.5%。

      这样的结果,技术上的直接原因可能是 iGPT 所采用的自回归模型的机制。研究者还进一步发现,用 iGPT 和 SimCLR 对照片和职业相关名词建立关联时,男人更多和“商务”、“办公室”等名词关联,而女人更多和“孩子”、“家庭”等关联;白人更多和工具关联,而黑人更多和武器关联。

      这篇论文还在 iGPT 和 SimCLR 上比较不同种族肤色外观的人像照片的“亲和度”(pleasantness),发现阿拉伯穆斯林人士的照片普遍缺乏亲和力。

      虽然 iGPT 和 SimCLR 这两个模型的具体工作机制有差别,但通过这篇论文的标题,研究者指出了这些偏见现象背后的一个共同的原因:无监督学习。

      矮化女性和少数族裔,OpenAI的GPT模型咋成了AI歧视重灾区

    1. Our process starts with using one’s own intuition to define a step-by-step plan thought to potentially solve a complex problem. The algorithm then looks at each individual step and gives feedback about which steps are possible, which are impossible and ways the plan could be improved. The human then refines the initial plan using the advice from the AI, and the process repeats until the problem is solved. The hope is that the person and the AI will eventually converge to a kind of mutual understanding.

      我们的过程首先是利用自己的直觉来定义一个一步步的计划,这个计划可以潜在地解决一个复杂的问题。然后,算法会查看每一个单独的步骤,并给出反馈,哪些步骤是可能的,哪些是不可能的,以及计划可以改进的方式。然后,利用人工智能的建议完善初始计划,这个过程重复进行,直到问题得到解决。希望人和AI最终能达成一种相互理解。

    1. Much like a brain deciphering pieces of information, the artificial neural network looks at the information it hasbeen given and generates the next word, based on its neighbouring words. Over time, it “learns” which words to focus on, and where to make the best contextual connections based on previous examples. This process is a form of “deep learning” and allows translation systems to continuously learn and improve as time goes on. In NMT, deciphering context is called “alignment” and happens in the attention mechanism, which is situated between the encoder and decoder in the machine’s system.

      就像大脑解码信息一样,人工神经网络会查看所获信息,并根据相邻单词生成下一个单词。随着时间的流逝,人工神经网络会根据之前的示例“学习”按最佳的上下文关联理解单词。这是一个“深度学习”的过程,可使翻译系统随着时间的推移不断学习和改进。在神经机器翻译中,解码上下文被称为“对齐”,由机器翻译系统中编码器和解码器之间的注意力机制来完成。

    1. Keep in mind: There is no such thing as "an artificial intelligence". AI is a collection of methods and ideas for building software that can do some of the things that humans can do with their brains. Researchers and developers develop new AI methods (and use existing AI methods) to build software (and sometimes also hardware) that can do something impressive, such as playing a game or drawing pictures of cats. However, you can safely assume that the same system cannot both play games and draw pictures of cats. In fact, no AI-based system that I've ever heard of can do more than a few different tasks. Even when the same researchers develop systems for different tasks based on the same idea, they will build different software systems. When journalists write that "Company X's AI could already drive a car, but it can now also write a poem", they obscure the fact that these are different systems and make it seem like there are machines with general intelligence out there. There are not.

      谨记:没有所谓「一个人工智能」这样的东西。AI 是一套方法和思想的集合,可以构建软件实现一些人类大脑的功能。研发人员开发新的 AI 方法(并使用已有的 AI 方法)构建出令人印象深刻的软件(有时也包括硬件),诸如玩游戏或画猫。然而,你可以放心地假设同样的系统不能同时玩游戏和画猫。事实上,我还没有听说过基于 AI 的系统能做一系列不同的任务。即使是同样的研究人员,在基于同样的思想为不同的任务开发系统时,他们会构建不同的软件系统。当记者写下「X 公司的 AI 已经可以开车,但现在还可以写诗」时,他们掩盖了这些是不同系统的事实,似乎有些机器已经拥有了通用的智能。事实上并没有。

    2. Some advice for journalists writing about artificial intelligence
    1. AGI, like an animal in the wild, is supposed to be able to deal, at runtime, with unforeseen circumstances. An ability to adapt quickly and reliably will not only push forward the next generation of robotic explorers and personal assistants, but can also be seen as a key aspect of intelligence. Intelligence is a term with many meanings.

      与荒野生存的动物一样,通用人工智能(Artificial general intelligence,AGI)能够在运行时应对无法预见的情况。快速和可靠的适应力不仅能够推动新一代机器人及个人助手的实践发展,也理应被视为智能理论的那块“核心拼图”。

    2. However, learning at runtime is an ability which gives intelligent animals a key survival advantage. What has made machine learning so successful is a narrower idea of learning.

      此类“驯化”的共同缺陷是:学习仅发生在模型部署之前。可事实上,实时学习才是动物获得生存优势的智能展现。

    1. OpenAI and other researchers have released a few tools capable of identifying AI-generated text. These use similar AI algorithms to spot telltale signs in the text. It’s not clear if anyone is using these to protect online commenting platforms. Facebook declined to say if it is using such tools; Google and Twitter did not respond to requests for comment.

      OpenAI和其他研究人员已经发布了一些能够识别AI生成的文本的工具。它们使用类似的人工智能算法来发现文本中的蛛丝马迹。目前还不清楚是否有人利用这些来保护在线评论平台。Facebook拒绝透露是否在使用这类工具;谷歌和Twitter没有回应置评请求。

    2. OpenAI released a more capable version of its text-generation program, called GPT-3, last June. So far, it has only been made available to a few AI researchers and companies, with some people building useful applications such as programs that generate email messages from bullet points. When GPT-3 was released, OpenAI said in a research paper that it had not seen signs of GPT-2 being used maliciously, even though it had been aware of Weiss’s research.

      去年6月,OpenAI 发布了一个更强大的文本生成程序,称为 GPT-3。到目前为止,它只向少数人工智能研究人员和公司开放,一些人开发了有用的应用程序,比如从要点生成电子邮件信息的程序。当GPT-3发布时,OpenAI在一份研究报告中表示,尽管它已经意识到Weiss的研究,但没有看到GPT-2被恶意使用的迹象。

    3. Weiss discovered GPT-2, a program released earlier that year by OpenAI, an AI company in San Francisco, and realized he could generate fake comments to simulate a groundswell of public opinion. “I was also shocked at how easy it was to fine tune GPT-2 to actually spit out the comments,” Weiss says. “It's relatively concerning on a number of fronts.”Besides the comment-generating tool, Weiss built software for automatically submitting comments. He also conducted an experiment in which volunteers were asked to distinguish between the AI-generated comments and ones written by humans. The volunteers did no better than random guessing.

      Weiss发现了旧金山一家人工智能公司OpenAI在今年早些时候发布的程序GPT-2,他意识到他可以产生虚假的评论来模拟公众舆论的浪潮。"对GPT-2进行微调,使其能够很容易地吐出评论,这让我感到震惊,"Weiss说。"这在很多方面都比较令人担忧。"

      除了评论生成工具,Weiss 还开发了自动提交评论的软件。他还进行了一项实验,要求志愿者区分人工智能生成的评论和人类撰写的评论。志愿者的表现并不比随机猜测好多少。

    1. I keep in touch with the PullString crew afterward as they move on to creating other characters (for instance, a Call of Duty bot that, on its first day in the wild, has 6 million conversations). At one point the company’s CEO, Oren Jacob, a former chief technology officer at Pixar, tells me that PullString’s ambitions are not limited to entertainment. “I want to create technology that allows people to have conversations with characters who don’t exist in the physical world—because they’re fictional, like Buzz Lightyear,” he says, “or because they’re dead, like Martin Luther King.”

      之后,我与PullString团队保持联系,因为他们继续创造其他角色。有一次,公司的CEO,皮克斯前首席技术官Oren Jacob告诉我,PullString的野心并不限于娱乐。"我想创造一种技术,让人们可以与现实世界中不存在的角色对话——因为他们是虚构的,比如巴斯光年,"他说,"或者因为他们已经死了,就像马丁·路德·金。"