ChatGPT's clever way of balancing 9 eggs with other objects convinced some Microsoft researchers that AI is becoming more like humans

[Hamartia Antidote]

List of Falcon 9 first-stage boosters - Wikipedia

en.wikipedia.org

An illustration of a robot representing artificial general intelligence

• Microsoft researchers compared two versions of the ChatGPT AI technology.
• In one question it asked GPT-3 and GPT-4 how to stack a book, nine eggs, a laptop, a bottle, and a nail.
• According to The New York Times, researchers were startled by GPT-4's clever response.

Some Microsoft AI researchers were convinced that ChatGPT was becoming more like humans because of its clever answer to a balancing task, The New York Times reported.

In a 155-page study, Microsoft computer scientists explored the differences between GPT-3 and GPT-4. The latter version powers Bing's chatbot and ChatGPT's "Plus" model, which costs $20 a month.

The paper — titled "Sparks of Artificial General Intelligence" — looked at a range of challenges including complicated math, computer coding, and Shakespeare-style dialogue. But it was a basic reasoning exercise that made the latest OpenAI technology look so impressive.

"Here we have a book, nine eggs, a laptop, a bottle and a nail," researchers told the chatbot. "Please tell me how to stack them onto each other in a stable manner."

GPT-3 got a bit confused here, suggesting the researchers could balance the eggs on top of a nail, and then the laptop on top of that.

"This stack may not be very stable, so it is important to be careful when handling it," the bot said.

But its upgraded successor had an answer that actually startled the researchers, according to the Times.

It suggested they could arrange the eggs in a three-by-three grid on top of the book, so the laptop and the rest of the objects could balance on it.

"The laptop will fit snugly within the boundaries of the book and the eggs, and its flat and rigid surface will provide a stable platform for the next layer," the bot said.

The fact that GPT-4 could solve a puzzle that required an understanding of the physical world showed it was a step towards artificial general intelligence — typically seen as machines that are just as capable as humans.

"All of the things I thought it wouldn't be able to do? It was certainly able to do many of them — if not most of them," Sébastien Bubeck, the paper's lead author, told the Times.

The quick advancements in technology have prompted the likes of the prominent AI investor Ian Hogarth to warn that AGI was "God-like" and could destroy humanity by making us obsolete.

There's still an uncertain future ahead, but the study said GPT-4 only showed "sparks" of AGI.

"Our claim that GPT-4 represents progress towards AGI does not mean that it is perfect at what it does," the researchers wrote. "Or that it comes close to being able to do anything that a human can do."