ChatGPT’s most popular uses so far are as an artist, assistant, brainstorming partner, and researcher. However, when engineers at Nvidia, the company making specialized AI chips, opened the hood and began tinkering with its engine, they found a much more sophisticated and capable entity than they had expected.

As a demonstration, they taught the chatbot to play Minecraft, one of the world’s most popular video games. Before long, the bot became an active citizen of the Minecraft realm, building a house, mining gold, hunting pigs, and living a full digital life.

In the Minecraft world, “it can explore by itself and get better and better at all kinds of skills,” Nvidia senior research scientist Linxi Fan told The Wall Street Journal. 

These surprising capabilities are “an early sign that the world’s leading artificial intelligence researchers are transforming chatbots into a new kind of autonomous system called an AI agent,” according to The Wall Street Journal.

AI agents don’t merely respond to prompts but can make decisions, initiate actions, and use a variety of online tools, often without humans telling them to do so.

“This is a huge commercial opportunity, perhaps trillions of dollars,” Jeff Clune, a former OpenAI engineer and now a computer scientist at the University of British Columbia, told the WSJ. “This has a huge upside and huge consequences for society.”

The reason: AI agents could evolve in a relatively few years to do most white-collar jobs, experts such as Clune are saying.

One skill that could help it do that is ChatGPT’s ability to write computer code that will enable the bot to interact with other software.

With an available add-on called a code interpreter, the bot can write code and then run the program it has created. In current examples, ChatGPT has used the interpreter, combined with its skill in writing code, to edit spreadsheets and turn still images into videos.

“These are projects where AIs [are] working with other AIs on your behalf,” Ashley Llorens, a Microsoft vice president, said to the WSJ. 

A next-generation AI named AutoGPT is moving closer to autonomy. Researchers are giving it broadly defined tasks such as “create a company” or “earn some money”. The bot then seeks to fulfill the instructions by asking questions—“what is a company?” or “how do I earn money?,” for example—then collecting answers from across the Internet and pondering those answers to decide what to do next.

That closely mirrors the kind of work that many white-collar workers do.

So far, AutoGPT has a tendency to get stuck in loops but researchers such as Fan at Nvidia are constantly refining it.

Eventually, most people will have access to an AI assistant that will do almost anything that can be done online, Clune predicted.

“If AI can do anything we can do, it doesn’t just replace the boring tasks,” he said. “It replaces all the tasks.”

TRENDPOST: With an almost limitless market, AI developers are rushing to make AIs ever more capable in ever more ways. 

AIs’ growing sophistication already has outstripped humans’ ability to anticipate its consequences. 

As a result, a period of turmoil lies ahead as companies, policy makers, and individuals confront and grapple with the domino-like fallout from widespread AI adoption.

As we’ve already noted in several previous articles, AI’s proliferation across our economic and personal lives will sharpen discussions around guaranteed incomes, job retraining, revising public education’s structure and goals, and the full range of human activities.