This article investigates the evolving relationship between AI and the ubiquitous 'like' button on social media platforms. It highlights the potential of using 'like' data to improve AI models through reinforcement learning from human feedback (RLHF).
Max Levchin, PayPal co-founder and Affirm CEO, argues that the vast amount of 'like' data held by companies like Facebook is incredibly valuable for training AI to make decisions aligned with human preferences. This data offers a cost-effective alternative to traditional RLHF methods which often rely on costly human feedback.
The article also explores how AI is changing how preferences are shaped. Social media platforms are using AI to predict user preferences, potentially making the 'like' button obsolete in the future. Facebook's experiments using AI to improve video recommendations, resulting in increased watch times, are cited as an example.
YouTube co-founder Steve Chen suggests that while AI might eventually replace the 'like' button's function in predicting user preferences, it may persist due to its role in handling sudden shifts in viewing habits and its importance in connecting viewers, creators, and advertisers.