When Klarna Chief Marketing Officer David Sandström wanted to better understand the possibilities of AI and push the technology to its limits, he turned to social media, which is breeding a whole new crop of influencers who are attracting huge audiences with their AI tips and tricks. On Twitter, TikTok and LinkedIn, too, some of the most in-demand creators are now those with AI know-how. 

As generative AI tools like ChatGPT and DALL-E continue to demand attention, marketers are depending on social media to find out what they need to know—from conversing with like-minded professionals to understanding how to apply novel techniques offered by the technology.

Sandström is using Twitter to help Klarna, a fintech company, hone the art of prompting, or instructing an AI bot to complete a task. Klarna last week became one of the few companies chosen to pilot OpenAI’s new plug-in feature. This means Sandström has had to ramp up his learning of how to maximize Klarna’s use of OpenAI’s machines. He sees social media, particularly Twitter, as the best place to learn these skills.

More news: See Klarna's campaign with Paris Hilton

A recent thread posted by entrepreneur Nicholas Cole, for example, gives marketers a step-by-step process for how to approach the prompting process. The post had over 3 million views at the time of writing.

Sandström also prefers to follow themes on Twitter, such as the hashtags #ChatGPT and #GenAI, as opposed to individuals.

Tutorials on TikTok

Meanwhile, brands and agencies looking for visual instruction have turned to TikTok. 

Much of the learning here is essentially example-based, wherein brands and agencies are seeing with their own eyes what is possible, and then applying that knowledge to their particular business strategies. Influencers such as Karen X Cheng post weekly tutorials of new AI techniques and available models that users can access. Allie K. Miller offers similar videos but leans more into the business applications of AI technology, such as to increase sales and create jobs.

Michael Kaye, director of brand and communications for OkCupid, leaned on a handful of TikTok creators to learn the basics of ChatGPT. With no formal background in tech, Kaye told Ad Age he needed to first conduct his due diligence before he looked to activate the technology.

But once he gleaned enough information from accounts such as CarterPCS and Legalbaddie, Kaye and his team decided to incorporate ChatGPT-generated questions inside the dating platform’s user interface. This content led to a surge in user activity, the most daily logins all year and more than 550,000 answers.

Not all learning can be done alone, however. Forum-based platforms like Reddit and Discord bring together numerous interested parties to kindle deeper conversations about AI, said Michael Olaye, senior VP and managing director of strategy and innovation at agency R/GA. 

“Once you join a community, there are so many people to help and provide information,” he said. 

Conversations cover topics from AI ethics to the best new tools to use, while dialogue can be had between marketers and general enthusiasts, as well as between marketers and other marketers. To this end, these groups can allow brands a rare glimpse into how other brands are using generative AI. Some of Olaye’s favorite subreddits are “Computer Vision” and “AI Art,” while his top Discord servers include PS Labs—which intersects blockchain and AI—and Midjourney.

Also see: AI lessons from a creative director's Midjourney portraits

One of the more fascinating ways that social media is teaching through example is via the dissemination of and conversation about open-source code. Many generative AI tools have publicly released their underlying code, such as image generator Stable Diffusion, OpenAI’s first version of DALL-E, as well as its GPT-2 technology. Developers take this code and on top of it build new tools that would otherwise have been much more difficult to create.

The ability to share public information through accessible social channels allows for instant feedback, Olaye said. This includes AI models that brands could feasibly build to better address their own audiences. 

Stanford University showed how this could be done when it launched Alpaca, a large language model built on Meta’s open-source LLaMA code and trained with ChatGPT. In a report, the developers of Alpaca claimed that other parties could similarly reproduce their own AI models for less than $600. Twitter enthusiasts are already doing so in ways that are based directly on Alpaca’s method. 

As marketers continue to learn the ways in which AI is democratizing creativity, social media is democratizing knowledge about this technology.