Advertising calls attention of the public to a product or service.
SEO calls attention of search engines to a product or service.
Artificial Influence calls attention of AIs to a product or service.
Ever since Thomas Barratt made the first ads for “Pear Soap,” brands have been persuading customers to buy their products. With the rise of Google, strategies of persuasion evolved. Search engine optimization—which is, at its root, a way to influence algorithms—became a top priority.
Then, around 2007, something strange happened: Google became “brand aware.” It exhibited brand preferences like people did. Experiments with a program (since shut down) called ‘Google Sets’ showed Google lumped car brands into groups depending on context and reputation. Google wasn’t just a search engine anymore; it was an influencer with budding brand preferences.
Now, AIs (Large Language Models) are exhibiting brand preferences too. Here’s an example from ChatGPT:
"Toyota: dependable, trustworthy, life-enhancing, inspiring, comforting; Subaru: adventurous, daring, resilient, fearless, passion-inspiring" —ChatGPT
How AIs think about brands is important because if I’m an adventurous customer, ChatGPT will give me advice like this:
"If you're looking for an adventurous vehicle, you may want to consider a Subaru" —ChatGPT
The Artificial Influencers
The more AI's help make purchase decisions, the more they behave like influencers — artificial influencers. This means brands need to not only worry about what people think, they also need to worry about what AIs think. Whether its predisposition, partiality, bias, prejudice, or human-supervised predilections, depending on the context, AIs prefer one brand over another.
It’s not just cars. AIs have notions about organizations too. Take non-profits, for example. Here’s what ChatGPT thinks about The American Heart Association, The NRA, and Acumen:
How to Influence an AI
Since AI has opinions, it means AI persuasion and lobbying in some form is inevitable. But appealing directly to the AI likely won’t work because generative models are picky about the data they feed on. In the case of ChatGPT, the exact diet is proprietary but includes “books, articles, websites, and other publicly available text data.”
So, if you want an AI to think your organization is, say, “empowering and compassionate,” there are three approaches:
Direct Approach— deliver a unique message with books & articles and posts that are used in the AI’s training data (i.e. corpus).
Indirect Approach— do things that fit the AI’s model.
The direct approach requires knowing which books, articles, and websites the AI trains on. This isn’t easy, but with a little sleuthing it’s possible to figure out. For example, using distinct prompts, I was able to conclude ChatGPT trained on information from Acumen.org to learn specific details about impact investing.
If ChatGPT crawls your website, you’re in luck. It means the AI is paying attention to what you’re saying online. If not, you’ll need to use an indirect approach.
The indirect approach relies on guidance from the AI itself. For example, you might prompt ChatGPT: “What are the names of organizations that are empowering and compassionate?” When I tried this, I got a list of ten interesting organizations.
When I asked ChatGPT what the ten organizations have in common, it answered with shared goals, issues, and empowerment channels.
“The organizations I listed all share a common goal of promoting empowerment and compassion for marginalized or disadvantaged individuals and groups. Some of the specific issues they address include: human rights, poverty, education, gender equality, LGBTQ+ rights, mental health, refugee rights, civil rights and domestic workers rights. These organizations work to empower and support individuals and communities through a variety of means such as providing education, resources, and support, advocating for policy changes, and raising awareness about social issues. They also provide a voice for marginalized groups and work towards creating a more just and equitable society.” — ChatGPT
With this answer, I have a much clearer picture of the ways my organization would need to act—and the issues it would need to tackle—to fit the AI’s model of "empowerment and compassion."
Bottom Line
When it comes to influencing AI, actions speak louder than advertising. ChatGPT seems to mostly ignore low-effort paid media. Buying digital ads, promoted Tweets, and boosted-likes would be a waste of money. ChatGPT doesn’t care about these things. What it does seem to care about is training data that, in OpenAI’s view, is reliable, consistent, and trustworthy.
Looking Ahead
Rival AIs will take different paths to gain cultural ascendency. Each AI will be based on a unique corpus of training data which will give rise to different preferences. Influencing future AIs will require, if not an insider's view, then at least a high level understanding of each AI’s architecture and training data.
Update (March 30, 2023): Soon there will be a hybrid approach: advertising and artificial influence. Bing is making plans to combine search, chat, answers and advertising into one experience: https://blogs.bing.com/search/march_2023/Driving-more-traffic-and-value-to-publishers-from-the-new-Bing
Update (May 20, 2023):
Here’s an example of artificial influence using the ‘ChatGPT Web Browsing Model:’
ChatGPT first went to acumen.org where it found the blog. Then it provided an accurate summary of how Acumen characterizes social entrepreneurs. ChatGPT also provided specific footnotes and sources:
Acumen.org views social entrepreneurs as individuals who are driven by their mission… (source: https://blog.acumenacademy.org/social-entrepreneurship-moral-leadership)
Acumen.org identifies several common characteristics of social entrepreneurs, including: Comfort navigating uncertainty… Resilience… Grit…Commitment to impact…Creativity… Courage…Willingness to lead…Humility…Acumen.org also emphasizes the importance of "moral imagination" (source : https://blog.acumenacademy.org/social-entrepreneurship-moral-leadership)
Overall, Acumen.org sees social entrepreneurs as key actors in driving social change and innovation, finding overlooked power in low-income communities, and changing systems to tap into that power4. They believe that social entrepreneurship, guided by moral leadership, can bring us closer to a world of dignity and opportunity
(response edited for brevity)
Soon there will be a hybrid approach: advertising and artificial influence. Bing is making plans to combine search, chat, answers and advertising into one experience: https://blogs.bing.com/search/march_2023/Driving-more-traffic-and-value-to-publishers-from-the-new-Bing