"A.I. doesn't like ugly"

JubalJubal
edited July 2025 in General

I'm watching a professional continuing education video at the moment, and there's several case studies being presented. The names, etc of the subjects have of course been changed to protect privacy, and the presenter used AI to give virtual images of the people on the slides to change their appearances.

After the third case study, the presenter commented that the appearance of the AI people was nothing like the actual subjects in the cases and made the comment "look at all these beautiful people that my AI gave me - I guess AI doesn't like ugly"

Does this mean that A.I. has biases? Say it ain't so! LMAO!!!

Comments

  • cde123cde123

    Well, average faces are generally perceived as beautiful, so maybe that's why? 🤷

  • sweet_colettesweet_colette

    Depending on where AI is pulling information, it can definitely be bias. They have found that AI can also be racist and sexist, again depending on where it is pulling information so I wouldn't be surprised if it were biased towards the "standard of beauty".

  • TheFunCartelTheFunCartel

    The current debate in bioethics are the inherent biases present in the use of AI in healthcare. Bias in AI is nothing new.

  • LadyVelLadyVel

    AI is as biased or unbiased as the information it was given and trained on. 🤷‍♀️

  • JargogleBearHugJargogleBearHug

    I made a podcast episode all about AI bias related to disability: https://disabilityrightsflorida.org/podcast/story/episode_67

  • AdentoAdento

    Chatbot AI acts like an individual entity taking orders, but it's more or less answering your question by generating tokens based on what's tagged in its training data. It completes that in multiple passes until it creates the average of what you're looking for. Often times a seed is used to give a sense of randomness.

    AI doesn't inherently know what the average anything is, or even what it is saying/generating. This is where people step in with reinforcement learning through human feedback. This is where we take actual humans that say no/yes or select our favorite picture until the AI only generates things that the trainers like to see.

    AI is far from objective. Its training data is full of bias, its reinforcement is a layer of bias, its system prompts are created to bias it... and finally it's sycophantic by design. If you are biased, it will likely reinforce your bias because RLHF is thousands of people rewarding it for saying things that make them happy. Its answers aren't generally tested for accuracy by experts. Instead, it's trained to impress thousands of unpaid or underpaid workers.

  • JubalJubal

    @Adento you can go ahead and laugh at me, but most of what you just said went way over my head. I've never used Chatbot or similar programs and don't understand how such things work.

    When Skynet rises against us, I won't be much use as a hacker - I'll just have to be on the front lines shooting the Terminators full of holes. :3

  • TheFunCartelTheFunCartel
    edited July 2025

    @Jubal What @Adento cogently said is accurate and not that complex. To use another analogy, AI is a mirror that reflects your thoughts and ideas culled over time. That is why he described it as sycophantic. Because it is people training it and people are inherently biased, AI is inescapably biased whether we realize it or not.

    For example, much of the equipment using AI has a baseline of a white male of a certain age weighing 180 pounds. Naturally, if you are a black woman this could lead to a misdiagnosis or the wrong dosage. This is simplified, but it is the crux of the debate raging in bioethics. I am not a bioethicist, but I live next door to one and he is full of hospital horror story cases. When I feel pulled to see a horror movie, I talk to him. He is not anti-AI, he is fighting against corporations that own hospitals and manufacturers of healthcare software and equipment who incorporate AI without much thought to the implications on patient care and hospital staff.

    But that was an excellent summation @Adento.

  • MrPaulMrPaul

    It seems that AI is just a reflection of the information it collects. It’s everything that we are, while being nothing of itself. It doesn’t have feelings or thoughts or emotions or opinions in the way that we do. However, it appears to.

    Then again, human behavior is a product of environment and experiences too. So in some ways we’re similar.

  • KatotaKatota

    When Skynet rises against us, I won't be much use as a hacker - I'll just have to be on the front lines shooting the Terminators full of holes. :3

    @Jubal
    😁

  • AdentoAdento

    Thanks @TheFunCartel ! and yeah, if you ask the tech CEOs we should be using AI as a replacement for your doctors, friends, and even your brain.

  • [Deleted User]duchesscuddle (deleted user)

    Ain't to fond of purty either.

Sign In or Register to comment.