This week YouTube hosted Brandcast 2025 in which it revealed how marketers could make better use of the platform to connect with customers.

A few new so-called innovations were announced at the event but one has caught the attention of the internet – Peak Points. This new product makes use of Gemini to detect “the most meaningful, or ‘peak’, moments within YouTube’s popular content to place your brand where audiences are the most engaged”.

Essentially, YouTube will use Gemini and probably the heatmap generated on YouTube videos by people skipping to popular points, to determine where to place advertising. Anybody who has grown up watching terrestrial television where adverts arrive as a way to build suspense will understand how annoying Peak Points could become.

  • GenosseFlosse@feddit.org
    link
    fedilink
    English
    arrow-up
    9
    ·
    edit-2
    9 hours ago

    Why do you need “AI” for this? That is something that can be done client side on a 30 year old phone without sweat if you already have a list of timestamps and number of engagements for a video?

    • Mniot@programming.dev
      link
      fedilink
      English
      arrow-up
      12
      ·
      8 hours ago

      I was at Google when they announced that only AI-related projects would be able to request increased budget. I don’t know if they’re still doing that specifically, but I’m sure they are still massively incentivizing teams to slap an “AI Inside” sticker on everything.

      • Ledericas@lemm.ee
        link
        fedilink
        English
        arrow-up
        2
        ·
        6 hours ago

        considering thier Pixels are heavily devoting most of its resources to AI solely, while neglecting all the other hardware.

      • GenosseFlosse@feddit.org
        link
        fedilink
        English
        arrow-up
        1
        ·
        edit-2
        5 hours ago

        This task is not that complicated. You have a list of timestamps with the seconds and the number of interactions for that second. All you need to do is to find the seconds with the most interactions. On a 1h video this would be only 3600 calculations “if currentValue is greater than maxValue”. If you store it as a Plain integer array you would need ~14KB of RAM. For comparison, a 1987 homecomputer with a 68000 CPU would do ~7Million calculations per second and have ~512kB of RAM, depending on the options.