Before generative AI, the simple algorithms optimizing newsfeeds for engagement acted as a powerful, yet misaligned, "baby AI." This narrow system, pointed at the human brain, was potent enough to create widespread anxiety, depression, and polarization by prioritizing attention over well-being.
Related Insights
Social media algorithms amplify negativity by optimizing for "revealed preference" (what you click on, e.g., car crashes). AI models, however, operate on aspirational choice (what you explicitly ask for). This fundamental difference means AI can reflect a more complex and wholesome version of humanity.
Extremist figures are not organic phenomena but are actively amplified by social media algorithms that prioritize incendiary content for engagement. This process elevates noxious ideas far beyond their natural reach, effectively manufacturing influence for profit and normalizing extremism.
A/B testing on platforms like YouTube reveals a clear trend: the more incendiary and negative the language in titles and headlines, the more clicks they generate. This profit incentive drives the proliferation of outrage-based content, with inflammatory headlines reportedly up 140%.
Pinterest reframed its AI goal from maximizing view time based on instinctual reactions (System 1) to promoting content based on deliberate user actions like saves (System 2). This resulted in self-help and DIY content surfacing over enraging material, making users feel better after using the platform.
Before ChatGPT, humanity's "first contact" with rogue AI was social media. These simple, narrow AIs optimizing solely for engagement were powerful enough to degrade mental health and democracy. This "baby AI" serves as a stark warning for the societal impact of more advanced, general AI systems.
