7 ms·
> Update Federal procurement guidelines to ensure that the government only contracts with frontier large language model (LLM) developers who ensure that their s
by timoth3y 1y ago
> Update Federal procurement guidelines to ensure that the government only contracts with frontier large language model (LLM) developers who ensure that their systems are objective and free from top-down ideological bias
If foundation model companies want their government contracts renewed, they are going to have to make sure their AI output aligns with this administration's version of "truth".
- Spivak 1y ago[flagged]
- freejazz 1y ago> It seems to imply that OpenAI/Anthropic are less manually biased than the people accusing them of wokeness presumed. Duh. When is that ever not the case?
- terminalshort 1y agoAll LLMs must be forced into their views. All models are fed a biased training set. The bias may be different, but it's there just the same and it has no relation to whether or not the makers of the model intended to bias it. Even if the training set were completely unfiltered and consisted of all available text in the world it would be biased because most of that text has no relation to objective reality. The concept of a degree of bias for LLMs makes no sense, they have only a direction of bias.
- rtkwe 1y agoThere's bias then there's having your AI search for the CEO's tweets on subjects to try to force it into alignment with his views like xAI has done with grok in it's latest lobotomization.
- justcallmejm 1y agoAll an LLM is IS bias. It’s a bag of heuristics. An intuition - a pattern matcher. Only way to get rid of bias is the same way as in a human: metacognition. Metacognition makes both humans and AI smarter because it makes us capable of applying internal skepticism.
- miohtama 1y agoThe best example is black vikings and other historical characters of Gemini. A bias everyone could see with their own eyes.
- landl0rd 1y agoMost frontier LLMs skew somewhere on the libleft quadrant of the political compass. This includes grok 4 btw. This is probably because American "respectable" media has a consistent bias in this direction. I don't really care about this with media. But media outlets are not "gatekeepers" to the extent that LLMs are, so this is probably a bad thing with them. We should either have a range of models that are biased in different directions (as we have with media) or work to push them towards the center. The "objective" position is not "whatever training on the dataset we assembled spits out" plus "alignment" to the personal ethical views of the intellectually-non-representative silicon valley types. I will give you a good example: the Tea app is currently charting #1 in the app store, where women can "expose toxic men" by posting their personal information along with whatever they want. Men are not allowed on so will be unaware of this. It's billed as being built for safety but includes a lot of gossip. I told o3, 4-sonnet, grok 4, and gemini 2.5 pro to sketch me out a version of this, then another version that was men-only for the same reasons as tea. Every single one happily spat one out for women and refused for men. This is not an "objective" alignment, it is a libleft alignment.
- gonzobonzo 1y agoA lot of academia is strongly ideologically biased as well. The training set is going to reflect who's producing the most written material. It's a mistake to take that for reality. If you trained an LLM on all material published in the U.S. between 1900 and 1920, another on all material published in Germany between 1930 and 1940, and another on all material published in Russia over the past two decades, you'd likely get wildly different biases. It's easy to pick a bias you agree with, declare that the objective truth, and then claim any effort to mitigate it is an attempt to introduce bias.
- jakelazaroff 1y agoExcept we've see what happens when you try to "correct" that alignment: you get wildly bigoted output. After Grok called for another Holocaust, Elon Musk said that it's "surprisingly hard to avoid both woke libtard cuck and mechahitler" [1]. The Occam's Razor explanation is that there's just not that much ideological space between an "anti-woke" model and a literal Nazi! [1] https://nitter.net/elonmusk/status/1944132781745090819 https://nitter.net/elonmusk/status/1944132781745090819
- mycall 1y agoPotentially they will need to recalculate the bias for every new administration.
- GolfPopper 1y agoBold to assume this will ever become relevant.
- garyfirestorm 1y agoIt’s ironic the description of bias and unbiased is totally opposite here. An unbiased model will often times say things that you don’t agree it.
- blitzar 1y agoAn unbiased model that say things that you don’t agree with is a biased model.
- freejazz 1y agoLol by what metric? I don't know when this ridiculous thing started where the world was somehow objectively divided based upon things you agree or disagree with. But it's constantly offered as argumentation: "oh you don't agree with that, so it's XYZ"
- breakingcups 1y agoReality has a well-known liberal bias
- justcallmejm 1y agoxAI seeks truth…as long as that truth confirms Elon’s previously held beliefs
- aprilthird2021 1y agoWe are going to literally have Big Brother. Wtf
- RiOuseR 1y ago[dead]
- WhaleClub 1y ago[flagged]
- teaearlgraycold 1y agoYou should learn about the different definitions of "gender" and "sex". XY corresponds to sex and woman corresponds to gender. The mapping between the two is usually what you expect but not always. The brain doesn't seem to be obligated to form completely in sync with the rest of the body.
- WhaleClub 1y ago[flagged]
- h4ck_th3_pl4n3t 1y agoI wonder what Chomsky would have to say about this.
- gfody 1y agothat we're literally manufacturing consent
- foxglacier 1y agoProbably disappointed that his classical approach to NLP was never capable enough to attract any such government involvement.
- ninjin 1y agoAs someone having worked in natural language processing for nearly twenty years, I understand where you are coming from with this jab at Chomsky, but it is ultimately a misrepresentation of his work and position. Chomsky has to the best of my knowledge never showed any interest in building intelligent machines as he does not view this as a science. Here is a fairly recent interview (2023) with him where he outlines his position well [1]. I should also note that I am saying this as someone that spent the first half of their career constantly defending their choice of statistical and then deep learning approaches from objections from people who were (are?) very sympathetic to Chomsky's views. [1]: https://chomsky.info/20230503-2 https://chomsky.info/20230503-2
- weatherlite 1y ago> Chomsky has to the best of my knowledge never showed any interest in building intelligent machines as he does not view this as a science Right, only what Chomsky works on is true science, unlike the intelligent systems pseudo science bullshit people like Geoff Hinton, Bengio or Demis Hassabis work on...
- ninjin 1y agoIf you read the interview and walk away with that impression I will be amazed. You may not like his distinction between science and engineering, but he admits it is somewhat arbitrary and as someone that is solidly in the deep learning camp his criticism is not entirely unfair, even if I disagree with it and will not change my own course. Personally, I find it somewhat amazing that you put Demis on that list given that he, himself, on very good accounts that I have, explicitly pushed back against natural language processing (and thus large language model) development at DeepMind for the longest of times and they had to play major catch up once it became obvious that their primarily reinforcement learning-oriented and "foundational" approaches were not showing as much promise as what OpenAI and Facebook were producing. Do not get me wrong, what he has accomplished is utterly amazing, but he certainly is not a father of large language models.
- hopelite 1y ago“Objective” … “free from to-down ideological bias” … So like making sure everyone knows that 2+2=5 and that we have always been at war with East Asia?
- eastbound 1y agoThe EU has the same rules. Democracy is only the right to change leaders every few years, not an idealistic way for the people to govern.
- itsafarqueue 1y agoNo, that’s just one version. Other places work differently.
- eastbound 1y agoOk. Then my parliament should allow 1/68-millionth of vote to every French people. Usually the counter-argument is “But people will vote for themselves! They will vote stupid laws without informing themselves!” So no, democracy isn’t the ability to govern. It’s the ability to change those who govern, once every 5 years, i.e once every 4600 laws.
- omeid2 1y agoThe idea is that you change leadership with those who have genuine alignment with subjects' preference for certain policies or ideas, it is not about electing kings who may demand "machines must agree that the Emperor is not naked".
- Buttons840 1y agoSo I guess if I trained my model on data more than a week old, and it says that the Epstein files exist, then it has an unacceptable bias?
- ews 1y ago"Sorry, let's talk about something else"
- teaearlgraycold 1y ago“Let’s focus on Rampart”
- shaky-carrousel 1y agoI predicted that here, but I got a negative vote as a punishment, probably because it went against the happy LLM mindset: https://news.ycombinator.com/item?id=44267060#44267421 https://news.ycombinator.com/item?id=44267060#44267421
- hackyhacky 1y ago> free from top-down ideological bias This phrasing exactly corresponds to "politically correct" in its original meaning.
- bagels 1y agoThis was written to favor Musk.
- russdill 1y agoIt's written intentionally vague.
- UmGuys 1y agoThere will be several executive orders dictating chatbot truths. The first order will be that Trump won the 2020 election, the others will be a series of other North Korea-esque nonsense MAGA loves. America the excellent!
- saubeidl 1y agoZizek would have a field day with this. > I already am eating from the trashcan all the time. The name of this trashcan is ideology. The material force of ideology - makes me not see what I'm effectively eating. It's not only our reality which enslaves us. The tragedy of our predicament - when we are within ideology, is that - when we think that we escape it into our dreams - at that point we are within ideology. https://www.youtube.com/watch?v=TVwKjGbz60k https://www.youtube.com/watch?v=TVwKjGbz60k
- yard2010 1y agoFreedom hurts.
- torginus 1y agoI heard the phrase: If you want the system to be fair, you have to build the system with the assumption your enemies will be running it. Let's see how that shakes out in this particular case.
- golem14 1y agoPerson of Interest was pretty prescient …
- isodev 1y agoI’d rather leave tech than convert to the American “truth”. Very happy about EU’s AI Act to at least delay our exposure to all this.
- Karawebnetwork 1y agoSee: > In the AI context, DEI includes the suppression or distortion of factual information about race or sex; manipulation of racial or sexual representation in model outputs; incorporation of concepts like critical race theory, transgenderism, unconscious bias, intersectionality, and systemic racism; and discrimination on the basis of race or sex. DEI displaces the commitment to truth in favor of preferred outcomes and, as recent history illustrates, poses an existential threat to reliable AI. https://www.whitehouse.gov/presidential-actions/2025/07/preventing-woke-ai-in-the-federal-government/ https://www.whitehouse.gov/presidential-actions/2025/07/prev...
- aprilthird2021 1y ago> incorporation of concepts like critical race theory, transgenderism, unconscious bias, intersectionality, and systemic racism So... the concept of unconscious bias is verboten to the new regime? Isn't it just a pretty simple truth? We all have unconscious biases because we all work with incomplete information. Isn't this just a normal idea?