华盛顿邮报 仍在进化的ChatGPT变得更强大了吗?(1)（在线收听）

    I don't mean to sound dramatic, but when a service called ChatGPT first went live to the public, it kind of felt like an era-defining moment.

    我并不想讲得太夸张，但当ChatGPT首次向公众开放服务时，感觉就像是一个定义时代的时刻。

    All you had to do was go to a website and type a question or a request into a box, and you would get a response that sounded a lot like it came from a human.

    你只需要登上网站，在方框里输入问题或要求，就会得到与人类相仿的回答。

    That was three months ago. Ever since then, that service called ChatGPT has churned out song lyrics and half decent academic essays, computer code that actually works, and for my benefit, explanations of how ChatGPT functions in terms a 10-year-old can understand.

    那是三个月前的事了。自那时起，这个名为ChatGPT的服务就开始大量产出歌词、还算不错的学术论文、可以运行的计算机代码，以及以10岁孩子都能理解的方式为我解释它自身是如何运作的。

    And now, according to the company that made it, the tech that makes ChatGPT as dazzling and as unnerving as it is just got an upgrade. It's called GPT-4.

    现在，根据ChatGPT研发公司的说法，使ChatGPT变得既耀眼夺目又令人不安的技术升级了，这项技术就是 GPT-4。

    The new GPT is actually better in like a thousand different hard-to-tell ways.

    新版GPT实际上在难以言说的方方面面都比旧版更好。

    Drew Harwell is a tech reporter at The Post and covers A.I.

    德鲁·哈维尔是《华盛顿邮报》的科技记者，负责人工智能报道。

    But I think the real cool thing in GPT-4 that's like the killer app is that it deals with images.

    我认为GPT-4真正厉害的地方在于图像处理这个杀手锏应用。

    You can upload an image and you can talk to GPT-4 and ask it to do things.

    你可以上传一张图片，可以与GPT-4对话，要求它做事。

    Like, you can upload a like jokey meme and say like, "Why is this meme funny?"比如，你可以上传一个搞笑的表情包，然后问它，“为什么这个表情包很有趣?”

    This is a pretty big step forward for a bot that you could only really type at before.

    对于一个以前只能打字交流的机器人来说，这是一个相当大的进步。

    But like any new product launch, Drew says this version of GPT has flaws, just like the old version.

    不过，就像任何新产品发布一样，德鲁表示这个版本的GPT，与旧版本一样，也有缺陷。

    And GPT-4 is so advanced that OpenAI is not releasing that new image description feature to the public yet, since they don't know exactly how it will be used.

    GPT-4非常先进，以至于OpenAI还没有向公众发布新的图像描述功能，因为他们也不知道使用方法。

    We don't know how this is going to be abused.

    我们不知道人们将如何滥用这项技术。

    We don't know how this is going to break and affect people's lives and be used to, you know, potentially throw people in jail or cost people their benefits.

    我们不知道这将如何影响甚至破坏人们的生活，可能会把人们送进监狱，也可能牺牲人们的利益。

    You know, all the other ways we've seen even more sort of simple A.I. systems like facial recognition be abused today.

    我们看到各种滥用技术的方式，甚至更简单的人工智能系统，比如今天被滥用的面部识别技术。

    So I think, you know, it's a moment that is really exciting, but it's also unnerving because we just don't know how this is going to be used and how it could be affecting real people in the real world.

    我想，这是一个非常激动人心的时刻，但也伴随着不安，因为我们不知道这项技术将如何被使用，以及它将如何影响现实世界中的人们。

    Today, how GPT-4 is pushing the boundaries of artificial intelligence and the new concerns this chatbot is already generating.

    今天，我们将讨论GPT-4如何拓宽人工智能的边界，以及这个聊天机器人已经产生的新问题。