Comment on I investigated millions of tweets from the Kremlin’s ‘troll factory’ and discovered classic propaganda techniques reimagined for the social media age.

2pt_perversion@lemmy.world ⁨1⁩ ⁨month⁩ ago

• sbv@sh.itjust.works ⁨1⁩ ⁨month⁩ ago

  I tried this with Gemini. Regardless of the number of rs in a word (zero to 3), it said two.

  source

  • Kraven_the_Hunter@lemmy.dbzer0.com ⁨1⁩ ⁨month⁩ ago

    So ask it about a made up or misspelled word - “how many r’s in the word strauburrry” or ask it something with no answer like “what word did I just type?”. Anything other than, “you haven’t typed anything yet” is wrong.

    source

    • sukhmel@programming.dev ⁨1⁩ ⁨month⁩ ago

      But it’s a phrase you typed, the very one that contains the question, unless you ask by voice or in a picture

      source
• the_post_of_tom_joad@sh.itjust.works ⁨1⁩ ⁨month⁩ ago

  Its 3 right? Why can’t ai guess that one?

  source

  • tee9000@lemmy.world ⁨1⁩ ⁨month⁩ ago

    Llms look for patterns in their training data. So like if you asked 2+2= it would look its training and finds high likelihood the text that follows 2+2= is 4. Its not calculating, its finding the most likely completion of the pattern based on what data it has.

    But a new model chatgpt-o1 checks against itself in ways i dont fully understand and scores like 85% on international mathematic standardized test now so they are making great improvements there.

    source
  • 2pt_perversion@lemmy.world ⁨1⁩ ⁨month⁩ ago

    Over simplification but partly it has to do with how LLMs split language into tokens and some of those tokens are multi-letter. To us when we look for R’s we split like S - T - R - A - W - B - E - R - R - Y where each character has its own token, but LLMs split it something more like STR - AW - BERRY which makes predicting the correct answer difficult without a lot of training on the specific problem. If you asked it to count how many times STR shows up in “strawberrystrawberrystrawberry” it would have a better chance.

    source

    • the_post_of_tom_joad@sh.itjust.works ⁨1⁩ ⁨month⁩ ago

      Thanks, you explained it well enough this layman kinda gets it!

      source