Duolingo will “gradually stop using contractors to do work that AI can handle,” according to an all-hands email sent by cofounder and CEO Luis von Ahn announcing that the company will be “AI-first.” The email was posted on Duolingo’s LinkedIn account.

According to von Ahn, being “AI-first” means the company will “need to rethink much of how we work” and that “making minor tweaks to systems designed for humans won’t get us there.” As part of the shift, the company will roll out “a few constructive constraints,” including the changes to how it works with contractors, looking for AI use in hiring and in performance reviews, and that “headcount will only be given if a team cannot automate more of their work.”

von Ahn says that “Duolingo will remain a company that cares deeply about its employees” and that “this isn’t about replacing Duos with AI.” Instead, he says that the changes are “about removing bottlenecks” so that employees can “focus on creative work and real problems, not repetitive tasks.”

  • millie@beehaw.org
    link
    fedilink
    English
    arrow-up
    14
    ·
    12 hours ago

    So… if Duolingo is just AI now, why would anyone pay for Duolingo? Just ask the AI for language lessons yourself.

    • Pete Hahnloser@beehaw.org
      link
      fedilink
      English
      arrow-up
      20
      ·
      17 hours ago

      One of the first things drilled into me in journalism was “Smith thinks” should be recast to “Smith said he thinks.”

      The C-suite is likely well aware of limitations, but shareholders like to hear about the hot new thing.

      The thing is, the idea isn’t wrong. Automating complex tasks is a bitch, but the repetitive tasks that turn any job into a grind are prime candidates. The larger issue is instead of letting employees spend more time doing fulfilling activities because of increased efficiency, companies tend to do layoffs.

      • t3rmit3@beehaw.org
        link
        fedilink
        arrow-up
        2
        ·
        edit-2
        1 hour ago

        the repetitive tasks that turn any job into a grind are prime candidates

        The problem is, this varies from person to person. My team divvies (or did, I quit not too long ago) up tasks based on what different people enjoy doing more, and no executive would have any clue which repeating tasks are repetitive (in a derogatory way), and which ones are just us doing our job. I like doing network traffic analysis. My coworker likes container hardening. Both of those could be automated, but that would remove something we enjoy from each of our respective jobs.

        A big move in recent AI company rhetoric is that AI will “do analyses”, and people will “make decisions”, but how on earth are you going to keep up the technical understanding needed to make a decision, without doing the analyses?

        An AI saying, “I think this is malicious, what do you want to do?” isn’t a real decision if the person answering can’t verify or repudiate the analysis.

  • orca@orcas.enjoying.yachts
    link
    fedilink
    arrow-up
    17
    ·
    edit-2
    14 hours ago

    >claims to care deeply about its employees

    >replaces some employees with bots

    Duolingo is talking out of both sides of their mouth.

    • ByteSorcerer@beehaw.org
      link
      fedilink
      arrow-up
      2
      ·
      3 hours ago

      The enshittification of Duolingo has already been going on for quite a while. It has really gone downhill in the last few years.

  • themurphy@lemmy.ml
    link
    fedilink
    arrow-up
    9
    ·
    18 hours ago

    Okay, real talk guys. If there’s one thing LLM’s actually CAN do, its language and translation. The best ones are also great at context.

    You sound a bit like people yelling at companies for laying off typewriter manufactures when the computer came along.

    And we are in the technology community…

    • orca@orcas.enjoying.yachts
      link
      fedilink
      arrow-up
      11
      ·
      edit-2
      10 hours ago

      It can’t self-correct, write new curriculum, etc. It’s an abyss for errors and it does not understand when you correct it. I’ve tried and watched it give me the same alternating 2 answers that I told it were both wrong. If it’s wrong, you basically have no idea. You’re putting trust in a navigator that is simply reading instructions and understanding none of them.

      I’ve been in tech professionally for around 18 years and have lived through trends and changes. This is by far the worst one and the most haphazard. I’m not anti-AI (I use Copilot at work and it’s constantly wrong), but what is happening now with this AI wave is just reckless. They’re throwing it at everything because shareholders love shiny new things that make line go up.

    • Umbrias@beehaw.org
      link
      fedilink
      arrow-up
      19
      ·
      17 hours ago

      they cant actually but it’s convincing enough that you’ll think it’s the same, and in the process make it financially impossible for improvements to be made by actual translators.

      • themurphy@lemmy.ml
        link
        fedilink
        arrow-up
        1
        ·
        16 hours ago

        Im curious as how it makes it finacially impossible for improvements by actual translators?

        And what improvements do you mean?

        • Umbrias@beehaw.org
          link
          fedilink
          arrow-up
          14
          ·
          edit-2
          16 hours ago

          if you put the people making translation possible out of work, you will run out of sources for useful translations.

          LLM are not magic. They function off of human effort for thir training data. High quality data is thus, sourced from (in this case) human translators. Some can be done without them by nonprofessional texts, but it is not enough.

    • noodlejetski@lemm.ee
      link
      fedilink
      arrow-up
      12
      ·
      edit-2
      6 hours ago

      4 months ago I thought I’d cut some corners and instead of looking for an online Cyrillic keyboard to transliterate a simple three word phrase, I’d ask ChatGPT. it transcribed “zvezda” (звєзда, a star) as жірка (“zhirka”, a frying pan). that was the first and the last time I tried using LLM to do language stuff.

    • projectmoon@lemm.ee
      link
      fedilink
      arrow-up
      8
      ·
      18 hours ago

      The problem is that while LLMs can translate, it’s still machine translation and isn’t always accurate. It’s also not going to just be for that. It’ll be applying “AI” to everything that looks like it might vaguely fit, and it’ll stifle productivity.

    • TurboHarbinger@feddit.cl
      link
      fedilink
      arrow-up
      1
      ·
      12 hours ago

      Bro every post jn Lemmy about AI carries a bandwagon full of hateful morons who don’t even know why they’re so angry.

    • witx@lemmy.sdf.org
      link
      fedilink
      arrow-up
      4
      ·
      edit-2
      18 hours ago

      Not really, back then you replaced people with more people that knew how to operate “insert tech advancement of the time”. Sure there were still people losing their jobs, but not so much like we’re seeing with this AI bs

  • phorq@lemmy.ml
    link
    fedilink
    arrow-up
    8
    ·
    19 hours ago

    That’s weird, because their servers can’t even seem to stay up long enough for any of the ai features in their app. It’s why I gave up on their max subscription because it would require doing conversations and 9 times out of 10 it would just stop listening/processing halfway through.

  • hddsx@lemmy.ca
    link
    fedilink
    arrow-up
    3
    ·
    18 hours ago

    Darn I was about to start using it to learn French. Any suggestions?