Advances in Artificial Intelligence Raise Ethics Concerns

In the coming weeks and months, we’re going to be exploring the newest developments in artificial intelligence and how it’s changing the way we live and work. Stephanie [inaudible 00:00:10] kicks off our periodic series, The AI Frontier with a look at some new tools that are getting attention and sparking concern over their ability to produce original work, ranging from college level essays to art.

A chat box that can mimic human intelligence and create poetry? I had to try it. Write a haku about a tabby cat. Here’s what came back. “”Soft fur brushed by breeze. Tabby cat basks in the sun. Purrs of contentment. Not bad. There’s also an image generator that can compose and manipulate pictures with a few keywords. A tabby cat sleeping on the beach, wearing a fedora, as an oil pastel. A site that writes in debugs programming code for you or a video generating tool.

Hi, I’m Anna. I’m an AI avatar created entirely by artificial intelligence. Today we’re going to-

It’s the rise of generative AI, a branch of artificial intelligence that enables computer programs to create original content.

Let’s see how good of a college essay ChatGPT can write.

This tool from the San Francisco based company Open AI is ChatGPT and can write essays.

Take a look at a couple of these sentences starting from the top. “Computer science has always been a fascinating field to me from the moment I first learned about computers.”

And children’s books. Last month, Design Manager, Ammar Rashi, used ChatGPT and text-to-image program MidJourney to create his book, Alice and Sparkle. Here’s how it works. To create new content, these programs are trained on data sets of existing content that hold text, images, video files, or even code scraped from the internet. Some artists say this amounts to appropriating their work without permission.

I do hyper realistic drawings.

Including South African artists, Jonah Dry.

This scares me. It makes a huge part of my practice somewhat redundant.

Hey, artists, have you been wondering if AI is using your art to train itself? I figured out where to look, you go to-

Illustrator Carlianne Tipsy, suggested artists go to the site haveibeentrained.com to see if their artwork was used in training popular text to image programs.

I personally found my book cover on there and a lot of stuff from some of my favorite artists.

And last month artists protested the portfolio site Art Station for featuring AI generated art on its page. This new wave of sophisticated AI tools is raising some tricky ethical questions, as well as some big concerns about topics such as the future of human labor. For more on some of these ramifications, I’m joined by Kelsey Piper. She’s a senior reporter at Vox News who covers artificial intelligence. Kelsey Piper, thank you so much for joining the News Hour. I know this is a tricky topic, but what strikes me as the most new about some of these AI platforms is that we’re no longer just talking about artificial intelligence. We seem to be talking about artificial imagination. Is that how you see this new breakthrough?

Yeah, I think a few years ago there was something a lot of people would say about AI, “Which is maybe they will replace some machines and automation and learn how to do lots of manual work that humans can do, but they’ll never write poetry, they’ll never play great chess. They’ll never tell good stories.” And so, one of the things that’s been astonishing about the recent breakthroughs in AIs is that long before AI has learned how to clean out a toilet, or put together a fast food order, it has figured out how to write poetry, how to draw beautiful artwork, how to riff off Shakespeare. It’s not the direction many people expected AI to take.

And I think it’s created a lot of anxiety. Is that what you sense when you talk to artists and people who feel that their work is somehow being not only appropriated but threatened, even replaced?

So I think that any industry that sees itself start to be automated is going to be sort of appalled and frustrated at people losing good jobs that were able to support them being replaced with computers. But I think a couple things have made it worse. One of those is just that it’s happened so fast. We went from having no AIs that could do meaningful beautiful images to having a ton of them from different companies, including some open source ones that anybody could go online and play with at any time, for free or for very small price. So that’s just an overnight shift.

(04:50)
And then I think the other thing is that artists take pride in their craft. They don’t think of it as just a job. They think of it as like an expression of their individuality, and their style and who they are as people. And so of course, it’s a little galling to have an AI system that can just do it all and copy your style without any need for you.

This application, this tool that’s been created ChatGPT, as you say, it’s gotten incredibly popular, incredibly fast. Who’s using it? Is it just people having fun or are there real world ramifications? I mean, for example, does it become irrelevant to be able to write a college essay now?

Yeah, so I think the college essay is certainly in trouble because ChatGPT can put together, I would say it is not quite at the peak of human ability yet. It is not yet writing something that is as great as a talented human who’s dedicated and working hard, but it can write something as good as your average 18 year old who’s kind of phoning it in a little bit. And that means that it is writing things that are indistinguishable from many college student essays, and that has a lot of professors sort of reeling and asking themselves, “How do we do assignments that aren’t going to be trivially cheated on?”

What strike me when I tried ChatGPT is that it can do the, what I thought was a uniquely human thing of BSing. And I mean a serious question though, what does its sophistication and its ability to do that say about the future of AI and what we’re looking at?

Yeah, I think that’s a great question because the key thing here is that four or five years ago, people were first coming out with the earliest language models using these generative techniques that could produce text. And they weren’t very good. They were kind of stilted. They were maybe writing it like a middle school level, not a college level. They would make up a lot of stuff, and they didn’t seem to have a good way of telling whether the things they were saying were true. And over the course of just a couple of years, we have vastly improved the quality of these programs, and now they’re writing at a college level, and they’re clever, and they bluff, and they make things up, but they can also be pretty accurate when they’re prompted to be pretty accurate.

(07:02)
So where is that taking us? In a few years, maybe we do have something that replaces my job as a journalist? Maybe we do have systems that can tell you what you want to hear in incredibly convincing length at any time. And then, there’s the question of who decides what those AI systems tell everybody? And how do we train those AI systems such that they are on humanity’s side helping us understand the world better instead of right now, I would argue they’re kind of trained to appease us. They’re trained to say what they think we want to hear, and there’s something scary about the idea of introducing these immensely powerful, immensely persuasive systems to the world with a mandate that’s as limited as, “Just say what will make people happy with you.”

Absolutely fascinating. Kelsey Piper, who covers artificial intelligence for Vox News, thanks so much for joining the News Hour.

Yeah, thanks so much.