Topic: Trust And Safety

Marc Andreessen explains "Reinforcement Learning by Human Feedback" (RLHF) to Jordan Peterson, describing it as the process of "socializing" raw AI models. Andreessen reveals that many of the people hired for these "trust and safety" roles at AI companies are the same individuals Elon Musk fired from Twitter, leading to concerns about embedded ideological bias.

Mark Zuckerberg announces that Meta will reduce censorship, move its trust and safety team to Texas, and add Dana White to its board. The move is seen as a political evolution following Trump's election victory, though NBC News reports on the continued presence of controversial AI chatbots on the platform.

Mainstream media outlets blame the Magdeburg attack on a lack of content monitoring on X, following Elon Musk's staff reductions. Critics argue that if the suspect's "conspiratorial narratives" had been removed, the violence might have been prevented. The hosts dismiss this as an attempt to hold Musk responsible for the actions of a disturbed individual.

Del Harvey, the former head of Trust and Safety at Twitter, is profiled regarding her unconventional background and use of a pseudonym. Before joining Twitter as an early employee, she volunteered for "Perverted Justice," a group that posed as minors to catch online predators. Her history has led to speculation about her ties to law enforcement and her influence over public speech on the platform.

Elon Musk dissolved Twitter's Trust and Safety Council, an advisory group of human rights organizations, stating it was not the best structure for external insights. Musk also faced media backlash for a tweet targeting Dr. Anthony Fauci, which prompted criticism from financial analysts like Andrew Ross Sorkin. Observers suggest Musk is focused on transforming Twitter into a payment and banking platform.

Pinterest implemented a new policy to block search results for "vaccine" and "cancer cures" to prevent the spread of what it deems "health misinformation." The company stated it relies on experts from the WHO and CDC to determine harmful content. This move is part of a broader trend of Silicon Valley "Trust and Safety" departments regulating user-generated information.

Twitter establishes a "Trust and Safety Council" to implement new community guidelines based on the "Dangerous Speech Project." The criteria for banning users include the speaker's influence and the "propitious" social context for violence. Critics argue these variables are subjective and could be used to silence controversial political speech.