Hey everyone!

I recently came across mentions of a “RightWingGPT.” However, I’m curious: with the notable presence of alt-right communities online, why hasn’t there been an instance of them fine-tuning a model that reflects their perspectives? It’s surprising given how sizable these communities can be. Has anyone heard or read about any such attempts?

Thanks!

  • Vikare_Mandzukic@alien.topBEnglish
    1·
    1 year ago

    The “easy” thing about a “far-right AI” is that it doesn’t depend on large amounts of parameters to work well, any small model running locally can be converted into a right-wing AI.

    Since AI can hallucinate anything from conspiracy theories, racism, climate denial and any other bizarre theories, AI does not need to be connected to reality.

    Being essentially a fake news machine, it is much easier to build a model that hallucinates non-stop than a model based on reality.

    I think that even a model with 1 billion parameters can be adjusted to be a far-right AI, it is enough to simply exacerbate the hallucinations, and will easily convince millions of right-wingers around the world.

  • the_unknown_coder@alien.topBEnglish
    1·
    1 year ago

    I’m alt-right working on LLMs. I’m looking to a generation of products beyond ChatGPT “stochastic parrots” and looking for something with more capability. Work is in progress.

  • the_unknown_coder@alien.topBEnglish
    1·
    1 year ago

    Any way, if you research the topic, you’ll see that there have been several attempts at non-leftist ChatGPTs. In some cases, the leftists interfered with their release because they claimed violation of their licenses (I think that happened for awhile at tusksearch, below).

    Yes, there was RightWingGPT by David Rozado [ https://davidrozado.substack.com/p/rightwinggpt ].

    But, there’s also Gipper AI:

    https://tusksearch.com/search?tab=Chat

    As I pointed out earlier, just about all GPT chatbots have been trained on extensive corpuses and often include alt-right/right wing viewpoints. Just about any model can be made to elicit alt-right or right talking points. It’s not hard to do. In fact, I’ve tried to get good alt-right philosophical dialog and sometimes get out right quotes from Stormfront.

    Of course, leftist corporatocracy has done their best to silence viewpoints that they don’t like, but there’s really nothing they can do.

  • arthurwolf@alien.topBEnglish
    1·
    1 year ago

    Oh boy! Let’s train a right-wing and a left-wing model, and have them talk to each other on Twitter forever!

    … Profit!

  • Illustrious-Lake2603@alien.topBEnglish
    1·
    1 year ago

    It’s wild that they all have alt right data in them. I made one of the uncensored models to become a bigot. It chose the personality of an alt right white nationalist.

  • Monkey_1505@alien.topBEnglish
    1·
    1 year ago

    I don’t think there’s a single human being who identifies as ‘alt-right’.

    Probably the people who get classified by other as such have a diverse set of beliefs that contradict and disagree with each other. For example you might get minarchist libertarians, socialist ethnonationalists, religious social conservatives etc.

    I don’t think it’s a useful label.