The headline said it was right-wing, not a MAGA cultist.

With different prompts it might do the opposite...

why do so many people think that what they see with their eyes must be the truth

I agree how these conclusions were developed is trash; however, there is real value to understanding the impact alignments have on a model.

There is a reason public llms don't disclose how to make illegal or patented drugs. Llms shy away from difficult topics like genocide, etc.

It isnt by accident, they were aligned by corps to respect certain views of reality. All the llm does is barf out a statically viable response to a prompt. If they are weighted you deserve to know how

What would neutrality be? An equal representation of views from all positions, including those people consider "extreme"? A representation that focuses on centrism, to which many are opposed? Or a conservative's idea of neutrality where there's "normal" and there's "political" and normal just happens to be conservative? Even picking an interpretation of "neutral" is a political choice which will be opposed by someone somewhere, so they could claim you're not being neutral towards them. I don't know that we even have a very clear idea of what "unbiased" would be. This is not to deny that there are some ways of presenting information that are obviously biased and others that are less so. But this expectation that we can find a position or a presentation that is simply unbiased may not even make much sense.

That is what the big AI companies do, though they are actually just packaging up American corporate norms as "neutral".

Moreat a 11

Here's why America's economy is great, but we can't afford basic healthcare and education systems even third world countries have by now.

Woke is just another meaningless label to them. It's the same as liberal, or BLM, or antifa, as in they don't understand what it is, but they hate it. Then they just call whatever they don't like by one of those labels and their dogs all go rabid. Bunch of fucking sheep...

Yeah, what they're calling AI can't create, they're still just chatbots.

They get "trained" by humans telling them if what they responded was good or bad.

If the humans tell the AI birds aren't real, it's going to tell humans later that birds aren't real. And it'll label everything that disagrees as misinformation or propaganda by the CIA.

Tell an AI that 2+2= banana, and the same thing will happen.

So if conservatives tell it what to say, you'll get an AI that agrees with them.

It's actual a topical concern with musk wanting an AI and likely crowdsourcing trainers for free off Twitter. When every decent human being has left Twitter. If he's able to stick around trumps government long enough and grift the funds to fast track it...

This is a legitimate concern.

As always it's projection, when musk tweeted:

Imagine an all-powerful woke AI

Like it was a bad thing, he was already seeing dollar signs from government contracts to make one based on what Twitter thinks.

It's really annoying actually. Sometimes you're asking something completely benign, and it's likes sorry I'm an AI and I can't do anything.

Example: I asked it what a commonplace book was. Apparently language models don't have the capacity to help with that!

The training set can tend toward expressing an ideology.

People always say things like this, but the fact of the matter is that there are a lot of extremely lonely people desperate to express themselves, and some of them think that a machine is their only hope.

https://www.hhs.gov/about/news/2023/05/03/new-surgeon-general-advisory-raises-alarm-about-devastating-impact-epidemic-loneliness-isolation-united-states.html

Some of those people talking to LLMs are fools who think they'll get some sort of wise response, sure. The rest of them are just looking for someone to talk to. Unfortunately, if politics gets brought up, that LLM might lead them down a dark path without them realizing it.

And honestly, the pathetic fallacy is an easy trap to fall into, especially with computers. Back in the 80s when I talked to ELIZA, I knew rationally that it wasn't alive, but there was still a tiny emotional part of me that would think of it as a human on the other end that I was talking to. And plenty of people let their emotions take over their reasoning abilities.