This image was created using the generative AI platform DALL-E based on the text of this blog. Generative AI tools can create a variety of content, including stories, poems, podcasts, images, and video. Copyright: Gabriel Demombynes

Misconceptions about artificial intelligence and what it means for people

Gabriel Demombynes
Manager of the Human Capital Project at the World Bank

In the last few weeks, several people have asked me “what’s the big deal with AI?” I tell them to read the book The Coming Wave by AI pioneer Mustafa Suleyman, which argues that we are at the cusp of an epic technological transformation that will rival those wrought by the steam engine, electricity, and the digital transistor. The ripple effects of AI will be as far reaching as those of earlier technologies and are impossible to confidently forecast.
The reason so many people are blindsided by AI is that it’s now coming at a lightning pace. One branch of the technology, generative AI, has taken off since a cornerstone 2017 paper enabled the creation of ChatGPT and a suite of tools for image, voice, and video creation. Even people who have been working on AI for years have been stunned by the recent velocity of progress, and a torrent of new advances will hit the public in the coming months.
Here are three misconceptions I have heard about AI and why I think they are wrong:

“AI is not capable of being creative”

Pick your definition of “creative” and AI is already there. AI systems exceed the large majority of humans in formal tests of creativity. A recent study found that AI produced higher quality ideas than students on average in a faceoff in a business innovation course. If you still doubt AI is creative, try listening to this 2-minute podcast I generated with AI using just the title, “How Investing in Human Capital Can Help Us Face Climate Change.” Or play around yourself with public AI tools to generate a story, poem, or image. See this post for more research on this issue.

“AI is out of reach for people in developing countries”

In more than 160 regions worldwide, an internet connection is all one needs to access some of the most powerful publicly available tools for free (ChatGPT and DALL-E image generation, both via Bing Chat). Yes, access remains a critical obstacle in more remote places. Basic internet network coverage (at least 3G) is still not available to 5 percent of people in lower-middle-income countries and 24 percent of those in low-income countries. And the cost of cellular data puts the internet out of reach for many who have network coverage—in the poorest countries, just one-quarter of people use the internet. These digital divides are shrinking, in part because of the efforts of the World Bank across the globe. And once people are online, they have many of the same AI tools at their fingertips as I do.

“This is like previous technology waves”

Economists instinctively see technology as a never-ending story: when new tech comes along, it increases the productivity of people who use it, makes some jobs redundant, creates new jobs, increases incomes and overall demand, and everything works out in the end. AI may follow this script, but there are reasons to believe this time is different.
As David Autor, the leading labor economist who has studied technological change, has said, AI presents both a promise and peril for jobs. The promise is that like previous tools AI will augment human expertise. See, for example, this real-world experiment showing that business consultants using AI finished 12 percent more tasks, completed tasks 25 percent more rapidly, and increased quality by 40 percent. In the near term there are productivity gains to be had for all sorts of jobs—likely including yours—from incorporating AI.
The peril is that AI could replace human expertise across whole classes of tasks.  Given the creative capabilities of already existing AI, a future where most of our cultural products are generated by machine does not seem far away. AI can learn autonomously, master functions that for humans require tacit knowledge, and make decisions by itself. Over the long-term horizon it’s difficult to envision what tasks AI will not be able to do.
Next week I’ll publish a follow-up post with reflections on what artificial intelligence may mean for human capital.

Previously posted at :