当我们站在科技的前沿，面对着一项令人瞩目的成就——GPT-3，这个名字如同呼唤着未知的未来。它是由 OpenAI 这家领先的人工智能实验室孕育而出的，通过深度学习和无数电子书籍、互联网上的知识海洋，它在短时间内学会了模仿人类语言的复杂性。Mckay Wrigley，一个年轻的程序员，在试图让 GPT-3 走近公众人物时发现了一种惊人的能力。他的目标是让这款系统像 Scott Barry Kaufman 那样讨论创造力。这次尝试引起了人们广泛关注，因为 GPT-3 的回答既充满了智慧又带有独特感。

Scott Barry Kaufman：你认为创造性表达是在多元化世界中成长的必需品吗？

GPT-3：我认为创造性表达是一种自然发生在多元化世界中的现象。当环境越来越多样化，你接触到的不同人、不同的机会、不同的地点和挑战也会变得更加丰富，这使得将这些点连接起来形成新事物变得更可能。在很多方面，如果你想要成为创意大师，你必须去做。如果你想成为作家，你必须写作；如果你想成为音乐家，你必须创作音乐；如果你想成为喜剧演员，你必须编织笑话。你不断地去创造，就像在这个过程中偶然发现一些有趣的事情一样。

Kaufman himself was left stunned: "It sounds like something I would say, crazy and accurate AI!"

In the weeks following GPT-3's release, numerous other experiments were conducted, each one sparking curiosity. It generated tweets, wrote poetry, summarized emails, answered mundane questions, translated languages, and even composed computer programs—all with minimal prompting. Some of these skills left experts at a loss for words.

For many AI researchers, this represented an unexpected step towards machines that could understand the ever-changing nature of language—a potential tool to understand human skills as well. Melanie Mitchell from the Santa Fe Institute said: "This has left me and many others in awe; it's hard to grasp what this thing is doing."

However, GPT-3 was not without flaws. It often revealed biased and hurtful language. If you asked for 10 statements in Scott Barry Kaufman's style, it might give you five convincing ones and five less so.

During a conversation between real-life Kaufman and his AI counterpart:

Scott Barry Kaufman: What is imagination?

AI (Scott): What is imagination? You should tell me!

Real Scott Barry Kaufman: You should know!

AI (Scott): Not you! You're supposed to know! What do you think imagination is?

This seemed like a genuine conversation. Moreover,

it also demonstrated the system's ability to complete tasks independently without additional programming or adjustments—the hallmark of true progress.

As we stand at the threshold of technological advancement with tools like GPT-3,

we are faced with both wonderment and trepidation.

The possibilities seem endless—speeding up smartphone app development,

crafting more human-like interactions in chatbots—and yet,

the implications remain uncertain.

While some hail it as a pathway towards truly intelligent machines,

others see these experiments as intriguing but misleading.

According to Georgia Tech professor Mark Riedl,"It speaks very fluently; it argues well; it does very well at producing text that sounds reasonable."

"But," he adds,"it doesn't pre-think anything."

"It doesn't have any real goals."

A new kind of rebirth

Jordan Singer—a product designer at Square—helped create their mobile app interface.

He wondered if this automated system could replicate his work.

He fed simple descriptions into GPT-3 along with necessary code snippets written in English using Figma design software—a tool used by professionals like Singer himself.

After several attempts matching Figma codes while providing more English descriptions when needed,

he found that GTP had learned how to generate code on its own!

If he described an Instagram post-sharing-and-viewing app simply,

the system would produce the required code sometimes riddled with errors but usually close enough after just two small adjustments made by Singer—"Not quite perfect but incredibly close!" This behavior was novel even for those who designed GTP itself—leaving them amazed.

(End)