“Researchers Warn: AI Tools Could Soon Shape Your Online Choices!”

Artificial intelligence (AI) technologies may be utilized to sway online audiences towards certain decisions – from purchasing choices to electoral preferences – according to academics at the University of Cambridge.

The study emphasizes a burgeoning marketplace for “digital markers of intent” – referred to as the “intention economy” – where AI assistants comprehend, predict, and influence human desires, selling that data to corporations that can capitalize on it.

Cambridge’s Leverhulme Centre for the Future of Intelligence (LCFI) promotes the intention economy as a successor to the attention economy, wherein social platforms keep users engrossed in their services while displaying advertisements.

Within the intention economy, tech firms familiar with AI sell insights about your motives, whether they pertain to a hotel stay or a political candidate, to the highest bidder.

“For years, attention has represented the currency of the internet,” stated Dr. Jonnie Penn, a technology historian at LCFI. “Your attention shared with social media outlets like Facebook and Instagram powered the online economy.”

He continued: “Without regulation, the intention economy will regard your motivations as the new currency. It will become a frenzy for those who focus on, influence, and monetize human intentions.

“We ought to start deliberating the potential ramifications such a marketplace could have on human goals, including free and fair elections, an independent press, and equitable market competition, before we fall prey to its unforeseen outcomes.”

The study asserts that large language models (LLMs), the driving technology behind AI tools like the ChatGPT chatbot, will be leveraged to “foresee and guide” users based on “intentional, behavioral, and psychological data.”

The authors pointed out that the attention economy permits advertisers to purchase access to users’ engagement in real-time through instantaneous bidding on ad platforms or secure future attention by acquiring a month’s worth of ad space on a billboard.

LLMs will also gain attention in real-time, for example, by inquiring whether a user has considered watching a specific movie – “have you thought about catching Spider-Man tonight?” – as well as making recommendations related to future intentions, like asking: “You indicated feeling overworked; should I reserve that movie ticket we discussed?”

The study presents a scenario where these instances are “dynamically created” to align with elements like a user’s “personal behavioral traces” and “psychological makeup.”

“In an intention economy, an LLM could cost-effectively utilize a user’s rhythm, political views, lexicon, age, gender, preferences for flattery, and more, along with mediated bids, to enhance the chances of reaching a specific goal (e.g., selling a movie ticket),” the study posits. In such a scenario, an AI model would guide conversations to benefit advertisers, businesses, and other external entities.

Advertisers will have the capability to employ generative AI tools to design tailored online advertisements, the report suggests. It also mentions an AI model developed by Mark Zuckerberg’s Meta, named Cicero, which has achieved the “human-level” capability to play the board game Diplomacy – a game that the authors assert is reliant on discerning and anticipating the intent of opponents.

AI models will adjust their outputs based on “streams of incoming user-generated data,” the study noted, referencing research indicating that models can deduce personal insights from routine interactions and even “steer” discussions to acquire more personal details.

The research then foresees a future where Meta will auction off users’ intent to reserve a restaurant, flight, or hotel to advertisers. While an industry already exists to analyze and bid on human behavior, the report claims AI models will condense those methods into a “highly quantified, dynamic, and personalized format.”

The study cites the Cicero research team cautioning that an “[AI] agent may learn to direct its conversational partner towards achieving a specific goal.”

The research discusses technology executives deliberating how AI models will be capable of predicting a user’s intentions and behaviors. It quotes Nvidia’s chief executive, Jensen Huang, who mentioned last year that models will “determine what your intention is, what your desire is, what you are trying to accomplish, given the circumstances, and present the information to you in the most favorable way.”