3 ms·

My (cynic) take: using LLMs to police speech, down to the minutest nuance. There are bad things that we humans say and do. But let's stop using AIs to do our p

• persnickety 2y ago
  What's the difference what tool is used to police speech? I fail to see any difference in goodness between this and any conceivable alternative..
  • blackbear_ 2y ago
    Efficiency in terms of cost and time makes a big the difference between something becoming widely used or remaining a curiosity.
    • brobdingnagians 2y ago
      But the presumption that we should be policing speech is not necessarily correct. Making it cheaper to do a thing which ought not be done is no real benefit to humanity.
  • wakawaka28 2y ago
    Presumably this would be used in an unsupervised way, to illegally censor speech. Even in cases where it is legal to censor people (such as ostensibly independent actions by private entities), it goes against the core Western values of freedom of expression.
• weinzierl 2y ago
  Models are biased, but it is a bias we control through the selection of training data. The act of inference reflects this bias but is free from arbitrary rule otherwise. I think this is a big step forward compared to policing by moody humans[1]. LLMs are a knife, let's use them for good instead of demonizing them. [1] https://en.m.wikipedia.org/wiki/Hungry_judge_effect https://en.m.wikipedia.org/wiki/Hungry_judge_effect
  • protomolecule 2y ago
    [flagged]
  • ch4s3 2y ago
    > but it is a bias we control through the selection of training data I don't find this to be very satisfying. It implies that the "bias of the model" is something we can quantify, understand, and control for during training. I'm sure you can come up with canned pat answers about coarse grained biases that have a lot of social valiance, but how do you know the model is weighted towards or against subtle word choices in millions of different scenarios. People seem to have enough trouble understanding the nuances of texts that are coming from another cultural perspective or form merely 20 years ago. Its a common enough experience to see young people cringing in response to the popular media of 30 years ago, because cultural expectations change, words shift in meaning, the "biases" of society move. LLMs are trained on a mountain of old books, media, internet forms, twitter and facebook posts, etc. How could you ever get a handle on just the "biases" of the training data from the contemporary internet sources, let alone trying to pick apart the backed in cultural assumptions of a bunch of books spanning decades?
    • weinzierl 2y ago
      When I said, we control the bias I meant we can control for it when we try to make a model as unbiased as possible. I think we can do that well even if the result will never be completely unbiased. When it comes to giving a model a deliberate bias I agree with you.
      • ch4s3 2y ago
        How do you propose that could practically be done though?
  • aiforecastthway 2y ago
    > Models are biased, but it is a bias we control through the selection of training data. But what if controlling bias is impossible because the very format of the data demands its emergence? I always thought both structuralism and poststructuralism were the worst types of in(s/n)ane navel gazing. I think I even wrote on a course feedback section in a final many, many years ago that these topics were massive wastes of my time. (We didn't have "evals" as such, but some faculty would elicit feedback on a separate sheet of paper you could leave with the final exam.) Observation: the more time I spend just letting the algorithms converge under a huge variety of treatments, the more convinced I become that the most basic structures of our language cast a spell on human understanding of the world, and that the interactions between the primitives of our language and human meaning-making is FAR more deterministic than we would like. Especially in terms of how they direct certain emergent phenomena in human groups. Implication: maybe it's reasonable to hypothesize that bias isn't "in the data". Maybe bias emerges, nearly deterministically, from the underlying grammatical structure of the data, at least in certain semantic contexts. Maybe certain utterances, and the persuasive power of those utterances, correspond to a sort of default hard-wired path for a majority subset of the population. A co-occurrence of certain neural pathways and certain grammar/semantics that eventually must arise in a given language and socially situated use of that language. FWIW: I have almost no social science training, the training I did get was received by me with extreme skepticism, and this sounds bunk to me even as I type it. So, for me, this hypothesis would have sounded stupid -- or at least inane/pointless -- before 2019. But I'm becoming increasingly convinced that the languages we've converged on encode certain types of group preferene/bias at an extremely fundamental level. ...either that, or I'm far FAR worse at filtering/generating data than I give myself credit for. But in at least a few cases I have literally reviewed every token within a certain delta of influencing the models behavior on an input sequence and see the same results.
• libertine 2y ago
  You're missing when LLMs are used to misinform, manipulate, and create chaos. The way you're framing it is "let's use limited humans to moderate an infinite number of LLM bots" - how do you go about it?
  • dsign 2y ago
    >> how do you go about it Let's not allow LLM bots to misinform, manipulate, and create chaos? This is the argument that we should allow everybody to have guns so that they can defend from other gun owners. In the case of LLMs, the solution is to leave them at the "shooting range"; which sadly, it's most of the Internet now. We need spaces purely for humans and the authenticity of their fleshy needs.