Axios Login

Generative AI could change many industries, but it still requires an actual person to interact with it. Enter the prompt engineers — people and firms setting themselves up as experts at getting what you want from ChatGPT and similar tools, Axios' Peter Allen Clark reports.

Why it matters: Fulfilling AI's promise of effective automation and productive brainstorming, many experts believe, will require skilled human operators.

"Writing a really great prompt for a chatbot persona is an amazingly high-leverage skill and an early example of programming in a little bit of natural language," Sam Altman, CEO of ChatGPT creator OpenAI, said on Twitter Monday.

• When prompted to define "prompt engineering," ChatGPT itself told Axios that "effective prompt engineering is critical for generating high-quality outputs from generative AI models, as it can help ensure that the model generates content that is relevant, coherent, and consistent with the desired output."

What they're saying: "It's kind of like selling jeans during the gold rush," Stephen Fraga, founder of the new chatbot-training business Prompt Yes!, told Axios, referring to the ecosystem of ventures that spring up in a boom. "It's not actually going out and digging up gold."

State of play: Many different businesses have already jumped in, some dating back to last summer's image-generating AI craze around Dall-E 2, Stable Diffusion and Midjourney.

• Platforms like PromptHero, Promptist and Krea serve as prompt search engines or inspiration for those looking for the right words to use.
• Marketplaces like PromptBase allow users to buy different prompts and sell their own.
• Trainers and educators are fanning out to help industries train workers on how best to use the new technologies, with video lecture sites like Udemy already offering many courses on them.

Meanwhile, a host of new "prompt engineer" jobs have opened and many job seekers are adding those two words to their resumés.

• A variety of locations have openings for the role, like a children's hospital in Boston and a law firm in London.
• Freelancer marketplaces like Fiverr and Upwork are already filled with people offering their AI prompt skills.

And of course there are books. David Boyle, director of brand agency Audience Strategies and author of the new book "Prompt: A Practical Guide to Brand Growth Using ChatGPT," tells Axios he and his co-author Richard Bowman have spent 20 years "trying to help people understand audiences and build brands off the back of them .... This technology changes everything about how we do that."

Fraga, of Prompt Yes!, started a software training business in the early 2000s, and compares the new AI wave to that environment. "I was like, 'Oh, this is a new boom.' Only everything's much faster."

• "Any white collar worker could benefit from the productivity boost from ChatGPT," Fraga says. "And then the other direction is productivity for specific occupations, like ChatGPT and prompt engineering for lawyers, ChatGPT and prompt engineering for CPAs."

The other side: Some experts believe that AI is evolving so quickly that the prompt engineering boom will burn out fast.

• "I have a strong suspicion that 'prompt engineering' is not going to be a big deal in the long-term & prompt engineer is not the job of the future," Ethan Mollick, an associate professor at the Wharton School who has rolled ChatGPT into his classes, said on Twitter.

What to watch: While most prompt engineering has so far focused on text or 2D art, AI-generated video tech isn't far away — and you can bet that specialized businesses and entrepreneurs will be right behind it.

The Supreme Court heard arguments Tuesday in a landmark tech case, Gonzalez v. Google, grappling for the first time with whether to make big changes in a 1996 law protecting service providers and publishers from being sued over content their users post, Axios' Ashley Gold reports.

Why it matters: The court's decision could have a far-reaching impact on any website that hosts content created by third parties.

Be smart: Tuesday’s arguments didn’t give justices who might favor changes to the law a ton to work with, and left the court with more questions than answers.

Background: The case stems from the 2015 ISIS Paris terrorists attacks. The family of a victim sued Google-owned YouTube, wanting to hold the platform partly responsible for radicalizing ISIS members.

• The plaintiffs argue that YouTube should not get Section 230 liability protection for its “recommendations,” while Google’s attorneys argued that sorting is what makes the internet useful and that there’s no proof that YouTube was purposely pushing ISIS content to people.

What they're saying: "This is a pre-algorithm statute," said justice Elena Kagan, who added that "algorithms are endemic to the internet."

• Kagan also recognized the court may be ill-equipped to parse the complexities of technology law, saying: "These are not the nine greatest experts on the internet," but also asking if Section 230's liability protection has gone too far.
• Those supporting Google's side of the case argue that taking away Section 230 protections from algorithmically-created content would create "economic dislocation," said Justice Brett Kavanaugh. "There are serious concerns ... we are not equipped to account for that."

Details: Over more than two-and-a-half hours of oral arguments, the justices appeared to be confused about the plaintiff's arguments, suggested that the conflict might better be fixed by Congress, and debated how far liability protection should go.

The bottom line: Google and Section 230 champions have reason to feel confident after Tuesday’s arguments, but pressure to change the law isn't vanishing either.

What's next: Today, the justices will hear oral arguments in Twitter vs. Taamneh, a very similar case attempting to hold tech platforms liable for aiding and abetting terrorism.

Users are putting ChatGPT to work writing stories and books — and trying to sell them.

What's happening: Reuters reports it found more than 200 books in Amazon's Kindle store that listed ChatGPT as author or co-author, while a respected science fiction magazine closed the door on unsolicited submissions after a deluge of ChatGPT-authored manuscripts.

Why it matters: As the generative AI bandwagon gathers momentum, enabling easy mass production of text and images, expect any open platform (or transom) to get clogged fast.

Details: Neil Clarke, the editor of Hugo Award-winning Clarkesworld, Monday tweeted, "Submissions are currently closed. It shouldn't be hard to guess why."

• Clarke said the influx of AI-written entries was "largely driven ... by 'side hustle' experts making claims of easy money with ChatGPT."

Between the lines: Clarke had previously noted in a blog post that, though spam submissions to his magazine had mounted from the start of the COVID pandemic, they started multiplying on a hockey-stick-like exponential curve with the arrival of ChatGPT at the end of 2022.

Our thought bubble: The war against spam is a long-running arms race, and the supply side of the fight has just gotten a giant shipment of powerful new weapons.

On Tap

• More internet arguments at the Supreme Court today.

ICYMI

• NFTs may no longer be hot, but nearly half a billion dollars' worth traded in the last 30 days. (Axios)
• The U.S. Supreme Court declined to hear a Wikipedia lawsuit challenging online surveillance by the NSA. (Reuters)

Minesweeper, the free game that came with early versions of Windows, was the addictive tech scourge of its era, according to a new book. Axios' Stephen Totilo has more.