There is No Algorithm for Truth - with Tom Scott

The Royal Institution・2 minutes read

Google developed a machine learning system to filter search results based on scientific consensus, challenging personal beliefs like a no-deal Brexit. Algorithms in platforms like YouTube can exhibit biases, leading to unfair demonetization and perpetuating inequalities.

Insights

  • Google developed a machine learning system to filter search results based on scientific consensus, potentially challenging personal beliefs and highlighting the importance of unbiased information dissemination.
  • The lecture emphasizes the complexities of algorithmic bias, showcasing how systemic societal biases can influence machine learning systems, leading to inequalities in content promotion and the importance of addressing these issues for fairness in artificial intelligence.

Get key ideas from YouTube videos. It’s free

Recent questions

  • How does Google distinguish fact from fiction?

    By using a machine learning system, Google ensures only scientific consensus appears in search results and recommendations. This system may challenge personal beliefs by suggesting controversial topics like a no-deal Brexit or dissolving the UK into Europe.

  • What are the challenges of reaching diverse audiences in science communication?

    The speaker, a successful science communicator, discusses the challenges of reaching diverse audiences and the impact of algorithms on media consumption. He acknowledges the complexities of modern science communication and the influence of algorithms on shaping content consumption patterns.

  • How do machine learning systems categorize content?

    Machine learning systems categorize content based on human-curated examples and learn from feedback. These systems can exhibit biases, like YouTube's algorithm associating LGBT content with explicit material, leading to unfair demonetization.

  • Why is algorithmic bias a significant concern?

    Algorithmic bias is a significant concern because systemic biases in society can influence machine learning systems, perpetuating inequalities. The speaker emphasizes the difficulty of eliminating biases in artificial intelligence and acknowledges the work of experts in addressing these issues.

  • What is the goal of YouTube's recommendation engine?

    YouTube's recommendation engine prioritizes increasing watch time to retain viewers on the platform, potentially perpetuating biases and systemic inequalities in content promotion. The algorithm rewards videos that attract viewers and keep them engaged, particularly with advertisements.

Related videos

Summary

00:00

"Google's AI filters facts, challenges beliefs"

  • Google invents a machine learning system to distinguish fact from fiction, ensuring only scientific consensus appears in search results and recommendations.
  • The system may challenge personal beliefs, like suggesting a no-deal Brexit or dissolving the UK into Europe.
  • The speaker, a successful science communicator, discusses the challenges of reaching diverse audiences and the impact of algorithms on media consumption.
  • The algorithm, a collection of machine learning systems, categorizes content based on human-curated examples and learns from feedback.
  • Machine learning systems can exhibit biases, like YouTube's algorithm associating LGBT content with explicit material, leading to unfair demonetization.
  • Algorithmic bias is a significant concern, as systemic biases in society can influence machine learning systems and perpetuate inequalities.
  • The speaker acknowledges his privilege as a successful communicator and the challenges of ensuring fairness in artificial intelligence.
  • YouTube's recommendation engine prioritizes increasing watch time, potentially perpetuating biases and systemic inequalities in content promotion.
  • The speaker emphasizes the difficulty of eliminating biases in artificial intelligence and acknowledges the work of experts like Hannah Fry in addressing these issues.
  • The talk highlights the complexities of modern science communication, the influence of algorithms on media consumption, and the ongoing efforts to ensure fairness and accuracy in artificial intelligence.

13:04

YouTube Algorithm: Balancing Engagement and Revenue

  • Videos that were 20 minutes long were considered good by the system, leading to the phenomenon known as Goodhart's law.
  • To manipulate the system, people started making longer videos and placing important content at the end to force viewers to watch the entire video.
  • YouTube now prioritizes high-quality videos that retain viewers on the platform, regardless of whether they are from the same channel or genre.
  • The new engine for YouTube recommendations, developed by Google Brain, focuses on generalizing recommendations and identifying similar but not identical content.
  • The algorithm rewards videos that attract viewers to YouTube and keep them engaged, particularly with advertisements.
  • Google serves more ads to viewers who are more tolerant of them, encouraging viewers to skip ads less frequently.
  • Platforms like YouTube, Twitter, and Facebook are primarily advertising companies, aiming to target ads perfectly to each user.
  • The algorithm's goal is to balance engaging content with educational and verified content to maintain viewer interest and advertiser revenue.
  • Investigations have shown how easily viewers can be led from apolitical content to extreme and misleading political content through algorithms.
  • Authority online often stems from having an audience, with gatekeepers playing a role in determining what content is authoritative.

25:35

"Balancing credibility and audience engagement in science"

  • The lecture discusses the importance of qualifications and expertise in science communication, highlighting the value of credibility over popularity.
  • Sister Wendy, an English major, gained popularity as an art historian due to her engaging commentary, emphasizing the impact of audience connection.
  • Neuroscientist Danny Beck stresses the ethical duty to defer to more qualified individuals on topics beyond one's expertise.
  • The lecture delves into audience preferences, noting that adding personal commentary to shared content increases engagement and reach.
  • YouTube creators analyze retention statistics to gauge audience interest, with a focus on maintaining viewer engagement throughout videos.
  • The lecture questions the balance between pandering to audience preferences and maintaining authenticity in content creation.
  • The text references The KLF's satirical guide on achieving a number one hit in the music industry, cautioning against the potential pitfalls of early success.
  • The lecture draws parallels between music industry success and science communication, highlighting the need for sustained effort and gradual audience growth.
  • The concept of parasocial relationships is explored, emphasizing the emotional investment of viewers in creators who may not reciprocate the connection.
  • The text discusses the evolution of fan interactions, from celebrity phone lines to social media engagement, and the blurred lines between genuine connection and performative fandom.

37:56

"Online Platforms: Echo Chambers and Extremism"

  • Twitch streamers engage in constant interaction with their audience, reading messages, replying, and acknowledging subscribers who pay monthly fees ranging from $5 to $25.
  • Subscribers receive perks based on their payment level, such as shoutouts and special animations indicating their subscription length.
  • Patreon advises creators to share personal content to establish an emotional connection with fans, potentially converting them into patrons.
  • The blurred line between admiring someone's work and admiring the person themselves is highlighted, with parasocial relationships blurring this distinction.
  • Effective science communication involves creating an emotional connection with the audience, emphasizing the importance of relatability and personal storytelling.
  • The significance of having a relatable figure in science communication is stressed, with personal stories often resonating more than well-cited research.
  • Online platforms like Reddit have faced challenges with free speech, leading to the infiltration of extremist groups like neo-Nazis due to lax moderation.
  • The concept of echo chambers, where dissenting opinions are silenced, is contrasted with the free speech approach of platforms like Reddit, highlighting the dangers of both extremes.
  • The negative impact of echo chambers is discussed, where group conformity stifles dissenting voices and perpetuates misinformation.
  • The analogy of a "Nazi bar" is used to illustrate the consequences of allowing extremist views to dominate online spaces, leading to the exclusion of more moderate voices.

49:37

Extreme beliefs drive out dissent, impacting discussions.

  • Extreme radical believers drive out those who are less certain, leading to discussions becoming more extreme and less open to dissent.
  • The discussion on extreme policies, not political alignments, is crucial for companies facilitating discussions to decide where they stand.
  • A Florida-based natural medicine clinic selling homeopathic vaccines faced significant backlash on Twitter, with many replies criticizing the practice.
  • The concept of "ratio" on Twitter determines the type of content shared based on the number of replies, likes, and retweets.
  • Policy decisions to reduce abuse on platforms like Twitter can inadvertently impact legitimate pushback against controversial practices like homeopathic vaccines.
  • Centralization of online platforms has led to standardized community standards, limiting diverse discussions and differing rules seen in the early 2000s.
  • Attempts to create alternative platforms like Mastodon or Discord face challenges due to the ease and accessibility of centralized platforms like Twitter.
  • The ideal algorithm for social media platforms would balance suppressing harmful content like fake news and conspiracy theories while still allowing for entertainment and engagement.
Channel avatarChannel avatarChannel avatarChannel avatarChannel avatar

Try it yourself — It’s free.