[–] 0 points 6 months ago (3 children)

Are you forgetting the IP theft

I'm going to come out and say it: IP theft isn't a thing. IP is not something that can be stolen. It can have its license violated or it can be copied against the wishes of its owner. What it absolutely cannot be is "stolen".

A car can be stolen. A phone can be stolen. A book or a CD or a DVD can be stolen. The concepts or ideas or literal content of what amounts to Intellectual Property cannot be stolen. It can only be copied.

If anything has been stolen it's the commons that is the public domain. It was taken away for about four generations. Long enough that no one remembers the IP that's only just now becoming public domain. It's a loss far greater than anything related to AI.

I'll also say this: Even if an AI were trained on nothing but public domain works (like most image generating AI a la ImageNET) people would still be spouting bullshit like, "it's stealing IP!"

  • source
  • parent
  • context
  • [–] 7 points 6 months ago (1 child)

    While interesting—in a curious sort of way—I think it will simply reveal that a whole lot of stuff on the Internet has been automatically replacing dashes with em dashes for a long time. Also, a whole lot of professional writers have been using it since forever.

    A better "aha! AI!" catch is post-2020 usage of emojis as bullet points. Though, that really only catches ChatGPT and Claude (Gemini and the other models don't do it as much unless they know they're making a README.md).

  • source
  • [+] -10 points 6 months ago* (last edited 6 months ago) (11 children)

    Gemini thinks it's peony...

    Screenshot of an android phone using the "circle to select" feature of the app overview screen that uses AI to examine what it sees

    Note: Gemini is usually very good at identifying plants but some plants are notoriously difficult for anyone to pin down except horticulturalists (e.g. experts). For example, Pothos vs. Heartleaf Philodendron or more dangerous, Queen Anne's Lace vs Poison Hemlock. AI also has trouble identifying grasses (in general) because many can look the same from a picture. Even close up! Yet a human (in person) would be like, "these two grasses are nothing alike!"

    It's all about pattern matching. If two leaves from two different plants have the same appearance and growth pattern, AI can easily get it wrong. In general, though, it usually does a really good job.

  • source
  • [–] 0 points 6 months ago

    This is slightly unnerving to me. I mean, all my social media is already public but I post so much and so often, how would they even go through it all?

    Me: "You want my social media accounts? Sure! They're all public, and Google says there's over 4.6 million words to go through—I actually had it run through them all to gather the real figures from its database—starting with newsgroups, then phBB forums, Slashdot, Digg, IRC logs, Reddit, Stack Overflow, and loads more. Good luck! I recommend sorting by most viewed and then most upvotes 👍"

    Them: "Uh. What are newsgroups?"

    Because much of my writing is highly technical, it's not exactly "beach reading" material. So let's assume the customs officer in charge of producing my biography can read it at about 150 WPM...

    511 hours or 12.8 weeks if they're only going to read during their 40-hour work schedule. They'd have to spend over three months reading my thoughts.

    Imagine the line of people waiting behind me!

  • source
  • [–] 2 points 6 months ago

    Just want to point out that nearly all new data centers use closed loop water cooling. That only makes sense in very, very dry places in the world that also have extremely cheap water.

    For example, cooling towers would make no sense in Florida because the ambient humidity is too high. Even though water is plentiful.

  • source
  • parent
  • context
  • view more: ‹ prev next ›