AI and the Future of Likes: How Algorithms Learn and Shape Our Preferences

Artificial Intelligence and the Power of the Like: How Algorithms Learn Our Preferences

The small "Like" symbol on social media seems insignificant, but it holds enormous potential. It provides valuable data about human behavior and significantly shapes the development of artificial intelligence (AI). But how exactly does AI use this data, and what impact does this have on our online experience?

The Like Button as a Data Treasure Trove

Max Levchin, co-founder of PayPal and CEO of Affirm, sees the Like button as a valuable resource for training AI models. He argues that the collected Like data from platforms like Facebook represents enormous value because it provides insights into human preferences. This data can be used to train AI systems and align their decision-making with human preferences.

A known problem in the field of machine learning is that AI systems trained on a clear reward function, while optimizing their performance, often arrive at results that deviate from human decisions. To counteract this problem, developers often use what is called Reinforcement Learning from Human Feedback (RLHF). This involves training the AI with data that reflects human preferences. However, obtaining this data is often costly, as it requires human supervisors and annotators.

Levchin sees the Like button as a solution to this problem. The vast amounts of Like data that platforms like Facebook collect could serve as a cost-effective source for RLHF and significantly influence the development of AI systems.

AI is Changing the Way Preferences are Formed

While Levchin emphasizes the use of Like data for AI training, AI is already changing the way these preferences are formed in the first place. Social media uses AI not only to analyze likes, but also to predict them. Facebook, for example, has already experimented with AI-powered algorithms to recommend Reels videos to users. The result: Longer viewing times, exactly what Facebook wanted to achieve.

Steve Chen, co-founder of YouTube, raises the question of whether the Like button is even needed anymore if AI can accurately predict what users want to see next. However, he concedes that the Like button could still be relevant for capturing short-term changes in users' viewing habits, triggered, for example, by life events. In addition, the Like button plays an important role for advertisers, as it signals user engagement and preferences.

AI-Generated Content and the Question of Authenticity

Another influence of AI is evident in the generation of content. More and more text and images on social media are created by AI. This raises the question of whether the original purpose of the Like button – the motivation for content creation – is even still relevant. Would the platforms be just as successful if users no longer created their own posts?

This development inevitably leads to the problem of authenticity. AI can subtly alter content, as the example of Alicia Keys' performance at the 2024 Super Bowl shows, where a mistake in the online version of the performance was corrected. The same technology that can correct a wrong note can also be misused for voice cloning, with potentially serious consequences.

The US Federal Communications Commission (FCC) has already taken measures against the misuse of AI for voice cloning, as this technology can be used for fraud and identity theft.

Virtual Influencers and the Future of the Like Economy

The future of social media could be populated by virtual, AI-generated influencers who gain legions of fans. "Aitana Lopez" is an example of such a virtual influencer who promotes brands and has thousands of followers, even though she doesn't really exist.

With AI-driven influencers and bots interacting with each other at a rapid pace, online interaction is fundamentally changing. If likes no longer come from real people and content is no longer created by them, what does this mean for the future of the Like economy?

It is conceivable that new tools will be developed to create transparency and verify whether a Like comes from a real person or a bot. Different platforms might use such tools to varying degrees.

The future of social media in the age of AI raises many questions. Do we want a future where our likes are transparent and unconcealable? Or do we want to retain the ability to hide our true preferences? The developments in this area will profoundly shape the way we interact online.