• @KazuyaDarklight
    link
    English
    10
    edit-2
    1 month ago

    Ugh, It’s a little judgy and I actually like AI in some day-to-day search scenarios but instantly disappointed in the blue cheese company when this banner appeared across the top of their homepage. " Find out how AI is shaping the future of Food and how you can support the movement! " With links to join a mailing list.

    Also, just an FYI, they aren’t selling the cheese to consumers yet, it’s only available in a handful of restaurants.

    • Match!!
      link
      fedilink
      English
      51 month ago

      That is a specific day to day scenario that makes sense to me for AI - finding new combinations of flavors by following some machine learning recommendation, especially if the intended application is to use AI to discover better vegan cheeses

      • @[email protected]
        link
        fedilink
        11 month ago

        I don’t think you really even need AI (in the sense of LLMs, as that’s usually what’s referred to with AI) for a flavor pairing software whatchamacallit. I forget what the concept is called but I learned about it through the flavor matrix. Where you essentially compare the different aromatic and flavor compounds in a given food/ingredient and base recommendations off of other foods with compatible compounds in them. A large enough database and a good UI would be a gamechanger for cooking

        • Match!!
          link
          fedilink
          English
          31 month ago

          They probably don’t mean an LLM in this case, but rather some combination of statistical methods for determining under-explored flavor combinations

          • @Rob
            link
            21 month ago

            Yeah, machine learning is going to be great for the protein revolution. For Qorn they had to run thousands of experiments to find something that tasted good. Imagine if you can model millions of experiments and already weed out 98% of proteins…

    • blargerer
      link
      fedilink
      21 month ago

      I’m not sure I get the mailing list part, but this is the type of task modern Machine Learning is actually great at (much better than they are at text or art generation). You have some huge open possibility space the humans can’t possibly explore all of, and where false negatives aren’t costly. You can use the model to narrow down the possibility space to something manageable for a human to review manually. Very similar to how its used in astrophysics, for example.