There is No Algorithm for Truth - with Tom Scott

20 insights saved from this video by @technology

@technology
Giving a video to outside experts can improve its substance while hurting retention, sharing, and algorithmic performance.
@technology
AI read retinal photographs to infer sex with 97% accuracy and age within four years.
@technology
Apparently, Google serves more advertisements to users who tolerate them, so skipping ads can reduce future ad frequency.
@technology
YouTube’s advertiser-suitability system linked LGBT-related content with explicit sex, making videos about being gay more likely unsafe.
@technology
An ideal social algorithm would suppress conspiracy theories and fake news while preserving enough nonsense to keep people engaged.
@technology
Opaque algorithms work like Skinner boxes: unpredictable rewards keep creators experimenting and breed superstitions about what the system favors.
@technology
Isaac Asimov’s 1950 story about machines quietly manipulating economies to protect humanity makes benevolent optimization deeply unsettling.
@technology
Watch-time optimization rewarded longer videos that buried important material at the end, pushing lower-quality videos into recommendations.
@technology
When platforms reward videos for keeping viewers on-platform, conspiracy theories and clickbait gain an attention advantage.
@technology
YouTube receives 500 hours of video every minute. Fully human-first moderation cannot operate at that scale.
@technology
In a 2015 Reddit survey, users’ biggest reason for not recommending Reddit was exposing friends to hateful content.
@technology
Policies protecting vulnerable people from coordinated abuse can also suppress public pushback against dangerous claims and sellers.
@technology
Parasocial relationships connect viewers emotionally to performers who do not know them, creating a power imbalance, not reciprocal friendship.
@technology
A mostly unedited ten-minute monologue to camera became a popular recurring series despite television conventions demanding movement, cuts, and spectacle.
@technology
A platform can elevate authoritative voices without establishing that those voices are true. Authority and truth are distinct.
@technology
Poor or biased training data make machine-learning systems reproduce distortions in their classifications.
@technology
Online authority often comes from audience size, not deep research, broad expertise, or superior qualifications.
@technology
When a creator shifts from personal videos to science communication, life-focused viewers leave and must be replaced.
@technology
YouTube retention graphs reveal opening drops, boredom-driven decline, and an end-card plunge. The algorithm favors gentler declines.
@technology
A successful platform cannot be all clickbait or contain none: either extreme drives away users, advertisers, or both.

Post on X

Keep what you found in There is No Algorithm for Truth - with Tom Scott

A free account saves any of these cards to your Collection, and Korva resurfaces them so you actually remember.