G/O Media, an online media company that owns Gizmodo and Kotaku has announced that it will begin a “modest test” of AI content on its sites.

  • Chozo
    link
    fedilink
    931 year ago

    The fact that I can open up ChatGPT right now and say “Write a Kotaku article about why Tetris is racist” and get a 100% believable result out of it should be a sign that they’ve been replaceable for a while now.

      • HidingCat
        link
        fedilink
        421 year ago

        WTF that is a whole load of baloney, it’s hilarious. Also a good reminder for us who lean left to remember to be critical when discussing such things too.

        • Nepenthe
          link
          fedilink
          33
          edit-2
          1 year ago

          Say what you will, I need myself a racially diverse tetris now. We should all aspire to be L shaped block.

          It’s so hard to even make fun of that without sounding like I’m on the extreme right, but I am CACKLING

      • thingsiplay
        link
        fedilink
        191 year ago

        @Helldiver_M That reads exactly like a typical dumb shit Kotaku article. No wonder, because it was trained from human data. I don’t know what’s more shocking, that our News outlets by human is so bad we think a robot wrote it, or if the AI is that good that we think a human wrote it. Both perspectives are frightening.

        • ArugulaZ
          link
          fedilink
          141 year ago

          Video game journalism has been crappy for a long, loooong time. You ever read pre-Ziff Davis EGM or GamePro? It’s like a lobotomy in print form.

          • Colombo
            link
            fedilink
            41 year ago

            Yet, there always has been a good journalism, either very quality reviews describing well the game in question, or very funny articles making fun of a game that is otherwise boring or bad.

            • ArugulaZ
              link
              fedilink
              41 year ago

              I’m not saying the magazines of the time were totally devoid of good gaming coverage. Video Games and Computer Entertainment was a solid, substantial read… the layout was nothing special, but the writing was pure quality. Kind of helped that they were writing articles for adults and not edgy fourteen year olds, or nine year olds hopped up on sugar.

              • Colombo
                link
                fedilink
                31 year ago

                Agree. And yet again, there was reason why gaming bloggers and YouTubers like TotalBiscuit got so much popular. Gaming Journalism crashed.

                I am not disagreeing with you, I grew up on Level and Score personally.

          • phosphorik
            link
            fedilink
            31 year ago

            Yo I loved EGM back in the N64 era and I’m pretty sure that was Ziff Davis. Then again maybe I was lobotomized with a magazine.

            • ArugulaZ
              link
              fedilink
              21 year ago

              Ziff-Davis EGM was pretty good. Pre Ziff-Davis EGM (the Sendai era, from 1989 to 1997) was not.

      • osarusan
        link
        fedilink
        181 year ago

        That’s hilarious, and about on par with most of the writing on so-called journalism sites these days. Glorious!

      • AnonymousLlama
        link
        fedilink
        61 year ago

        Holy shit it’s spot on for the type of outrage baiting I expected. I’m really impressed

      • CrowAlethe
        link
        fedilink
        61 year ago

        That’s so much fluff.

        Fun to read but honestly felt like watching an Ancient Aliens episodes.

      • May
        link
        fedilink
        21 year ago

        Why is every point marked ‘1.’ lol

        • chaogomu
          link
          fedilink
          101 year ago

          It’s the training data. Every list the bot has ever seen has had the number 1 in it. All other numbers come up less than that.

          The bot also has no memory of what it wrote before, and no clue what it will write next. It simply guesses what the next word will be based on what the last word was.

          Another failure of these bots, they’re Pre-Trained. It’s the “p” in the name. So anything they generate will be based on the training data, with no changes to the algorithm based on interacting with users. You can “convince” the bot of anything and the second you close that browser window, the bot basically resets to factory defaults.

          • FaceDeer
            link
            fedilink
            121 year ago

            Maybe it’s trying to use markdown to create a numbered list, in many flavours of markdown that would do it.

        • m3adow
          link
          fedilink
          01 year ago

          Probably botched Markdown formatting. Ordered Markdown lists will automatically be ordered properly, so starting each point with ‘1.’ doesn’t matter.

          • exscape
            link
            fedilink
            11 year ago

            I also think it’s botched Markdown but from the opposite perspective. When you have many points that end up parsed as several separate 1-item lists, you can write 1. 2. 3. and it will make it 1. 1. 1. since each list only has one point.

      • @pulaskiwasright@lemmy.ml
        link
        fedilink
        11 year ago

        This sounds exactly like something Kotaku would write:

        One of the key criticisms leveled at Tetris is the lack of diversity in its visual representation. The game predominantly features blocks of different shapes and colors, but the absence of any explicitly diverse or racially inclusive elements raises questions. In a world that is culturally diverse, the omission of representation within the game can be seen as a missed opportunity to promote inclusivity.

      • victron
        link
        fedilink
        11 year ago

        Lmao it’s awesome, I haven’t read Kotaku in ages, so I don’t know if it sounds Kotaku-y enough, but holy damn I can totally imagine a human nutjob writing that shit

    • yaomtc
      link
      fedilink
      191 year ago

      Believable as in imitates the writing style, sure, but what’s the point if it’s factually incorrect?

      From the article:

      The company joins a growing number of media entities experimenting with the technology […] These trials have already led to a flood of error-laden, plagiarized, and poorly written content due to badly implemented — and, some would argue, inherently unsuited AI models — that still have a strong tendency to make up facts.

      Hate on Kotaku all you want but they don’t make shit up as often as AI does

      • Helldiver_M
        link
        fedilink
        11
        edit-2
        1 year ago

        To be fair, Kotaku does sometimes make shit up. Like the Persona 5-Smash crossover lyrics being ableist thing. And they tried to double down on it for awhile if I recall right.

        And yes, ChatGPT makes shit up all the time. More often then Kotaku.

        I know in a post gamergate world, we need to be diligent for things like dog-whistles. And hating on Kotaku is arguably in dog-whistle territory. I guess in my opinion Kotaku is so bad, that we should be able to safely mock the crap out of them. I’m even more happy to mock any chuds that want to keep non cis-white-males out of games. They just weren’t relevant for this occasion.

      • Colombo
        link
        fedilink
        81 year ago

        Hate on Kotaku all you want but they don’t make shit up as often as AI does

        Oh they do. Take the whole kerfuffle around Kingdom Come.

      • FaceDeer
        link
        fedilink
        31 year ago

        I have found that one of the more effective ways to use ChatGPT for writing is to not just tell it “write me an article about…” But to give it a list of all the facts and basic arguments you want to include in the article and then tell it to use those. Takes more work to gather those bits of information ahead of time, but not a lot more work - you could basically do a bunch of Googling and copy and paste bits and pieces of what you find to use as your starting data.