Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

How A.I. chatbots become political

nytimes.com

Post reply on HN
Print view View original

41–50 of 76 posts

Re: How A.I. chatbots become political

#41

Post by talldayo »

coolhand2120 wrote: post #14

> But our A.I. systems are still largely inscrutable black boxes, which makes herding them difficult. What we get out of them broadly reflects what we have put in, but no one can predict exactly how. So we observe the results, tinker and try again. What an absurd thing to say. You don't get an abomination like Gemini without extreme and intentional tampering with the model. IIRC this was demonstrated in the HN thread…

I don't think the existence of Gemini disproves the author's statement. The model is clearly broken, not only within the definitions of what you or I would consider acceptable but also within the definition set by the prudes on-high. The wildly diverse output seems especially emblematic of a hackjob finetune not dissimilar to what OpenAI does with their instruction-tuning.

The quoted comment seems to align with how Google saw the situation. They wanted a specific desired outcome (neutered AI output), they applied a documented strategy, and got a torrential wave of "observed results" from the audience.

Re: How A.I. chatbots become political

#42

Post by Terr_ »

coolhand2120 wrote: post #14

> But our A.I. systems are still largely inscrutable black boxes, which makes herding them difficult. What we get out of them broadly reflects what we have put in, but no one can predict exactly how. So we observe the results, tinker and try again. What an absurd thing to say. You don't get an abomination like Gemini without extreme and intentional tampering with the model. IIRC this was demonstrated in the HN thread…

Isn't that confusing an ineffective treatment (prompt engineering) with the actual disease (insufficiently apolitical training data and use cases)?

Re: How A.I. chatbots become political

#43

Post by Fauntleroy »

Fauntleroy wrote: post #15
jimbokun wrote: post #13

Earlier quoted context omitted.

Studies comparing outcomes of children raised in married, two parent households vs single parent homes do not support left leaning orthodoxy. That’s just one example off the top of my head. Overton window just keeps people from voicing right leaning opinions in polite company, even when thoroughly backed by data.

What does single vs two parent households have to do with this?

Naturally, OP has abandoned this indefensible position.

Re: How A.I. chatbots become political

#44

Post by Terr_ »

FrustratedMonky wrote: post #12
kleer001 wrote: post #6

Earlier quoted context omitted.

the center of a current Overton Window, to be precise, that thing's slippery

If you take the Overton Window as a simple parabola, with an up swing on left and right side of curve. I'm finding more and more that it has become more like a Mandelbrot Set and on any given day I'm not sure if I'd be considered Left or Right by any random other person who I also can't identify if they are Left or Right. It is almost like Left and Right are so confused now, that they have lost all meaning. Except wh…

When it comes to LLMs being fed vast swathes of commentary, it's not so much the limits of the Overton window as the distribution.

Re: How A.I. chatbots become political

#45

Post by epgui »

mhuffman wrote: post #33
epgui wrote: post #5

They do say that reality has a left leaning bias... Personally I find these quadrant visualizations a bit misleading (regardless of where you sit) because the "centre" is not so much a neutral point as it is the centre of the "Overton window".

>They do say that reality has a left leaning bias I am not even sure how to interpret that. In the US, at least, politically people leaning left and leaning right are about even. I would definitely say that during my time in academia it appeared to me that more were left-leaning than people I knew in my non-academic life (ie. work, etc.) The fact that they have to spend so much time "re-aligning" these models seems t…

Your interpretation was immediately wrong, because the statement doesn't say anything about the prevalence of right-leaning people or left-leaning people.

Google the phrase "reality has a left leaning bias" and you will find a variety of explanations as to what it means.

Re: How A.I. chatbots become political

#46

Post by epgui »

algorias wrote: post #23
RangerScience wrote: post #16

Earlier quoted context omitted.

(edited) Are you claiming (1) that children from single-parent households have worse outcomes and (2) that left-leaning policies result in more single-parent households? (Not going to argue about #1, that seems pretty well established, just pulling it out as a claim for clarity. #2 is what I'd argue against, and would want to actually see your example so that it's possible to do so in a constructive way) > That’s jus…

The OP means that kids raised by married parents have better life outcomes. The policy proposals that would follow from this (e.g. restricting divorce) are right-leaning positions outside the overton window. But even beyond a specific policy proposal, there is a viewpoint in right-leaning circles that atomized individualism has caused lasting damage to the fabric of society and needs to be reigned in. (NB: I can't sp…

How can you know what OP meant? The best any of us can do is guess.

Re: How A.I. chatbots become political

#47

Post by Terr_ »

blueyes wrote: post #19

Can someone tell me how giving government officials the right to control how AI models are trained would produce less political LLMs? What wouldn't they skew toward supporting the current regime whatever it may be? And why would it not be protected by 1a?

> why would it not be protected by 1a?

One argument might be that it's not really a human's speech... but IMO that's a weak argument since the amendment is held to cover other forms of expression. Also it seems like a foot in the door towards "the mastermind of the unethical plan was a computer, therefore none of us are to blame for all the profit we made" excuses.

How about this for an analogy: Someone makes a machine that throws together scrabble tiles to make words, then chooses to take pictures of the results and sell them as art. The copyrightability may be suspect, but if a politician came in and demanded that certain words no longer be generated, that would still be a first amendment violation.

Re: How A.I. chatbots become political

#48

Post by epgui »

RangerScience wrote: post #30
algorias wrote: post #23

Earlier quoted context omitted.

The OP means that kids raised by married parents have better life outcomes. The policy proposals that would follow from this (e.g. restricting divorce) are right-leaning positions outside the overton window. But even beyond a specific policy proposal, there is a viewpoint in right-leaning circles that atomized individualism has caused lasting damage to the fabric of society and needs to be reigned in. (NB: I can't sp…

> (e.g. restricting divorce) I am so confused by the right. Aren't they usually in favor of things promoting freedoms, like freedom of association?

> Aren't they usually in favor of things promoting freedoms

Definitely not in a general sense. They place more importance on some freedoms than on others.

Re: How A.I. chatbots become political

#49

Post by epgui »

Terr_ wrote: post #44
FrustratedMonky wrote: post #12

Earlier quoted context omitted.

If you take the Overton Window as a simple parabola, with an up swing on left and right side of curve. I'm finding more and more that it has become more like a Mandelbrot Set and on any given day I'm not sure if I'd be considered Left or Right by any random other person who I also can't identify if they are Left or Right. It is almost like Left and Right are so confused now, that they have lost all meaning. Except wh…

When it comes to LLMs being fed vast swathes of commentary, it's not so much the limits of the Overton window as the distribution.

> When it comes to LLMs being fed best to swab the son of commentary

I understand all of these words individually, but I do not understand them in that sequence.

Re: How A.I. chatbots become political

#50

Post by Terr_ »

Covzire wrote: post #25

There's a cheap and simple way to de-politicize chat bots and make them 100% trusted by everyone: AI chatbots should refuse to answer moral or ethical questions unless the user specifies the precise ethical or moral framework to be evaluated against.

"Use the ethical framework that all ethical frameworks are unethical."
Post reply on HN