2024 ELECTION THREAD

Aight then.

Sorry, I missed the Gemini stuff. Sorry, still not buying the autocomplete stuff. "Dead children gaza" works, OK, still not buying that someone manually did something to the autocomplete to thwart people's attempts to search for literally the biggest news story of the month. I mean, really, what benefit does it have?

I *never* said "manually".

I said they tweaked the algo.

"what benefit does it have to nudge low info people toward the info i want them to be exposed to and against the info i don't them to see"?

Just had a look at your study. https://www.researchgate.net/publication...

LMAO. What a study. Those limericks prove that Copilot is biased. Use of the term "misinformation" makes the bias in Gemini "obvious" even though all it did was refuse to perform the request. Well, I'm sold.

Did you even read this before linking it? You can't have, because you wouldn't have embarrassed yourself like this otherwise.

Yes, "tweaking the algo" means manual interference. Same thing.

Bro, anyone searching for trump assassination news was going to find it whether autocomplete helped them or not. I mean, it was literally the biggest news story of the month. That's why I'm saying there was no benefit and huge risk with ****ing around with something that blatantly.

I mean google literally admitted the ultraleftist bias for gemini in image generation (remember when it depicted black roman emperors, popes, vikings and so on?), and took it down after the outcry, but sure go on

Way to shift the goalposts. I'm talking about the study that you linked to prove your point. If you meant to say "sorry, that was a BS 'study' and proves **** all, but here's some other evidence" then fine. Acknowledge that your "study" was complete dogshit and apologise for wasting my time by making me read through that nonsense first though.

The only thing I've seen that's biased beyond belief so far is that PDF written by a 5 year old passing itself off as a "study".

That is absolutely hysterical.

These chuckleheads will never do that. They pile on me for never admitting I'm wrong, even though I do *every single time I am* and then they act like this. It's painful to watch.

I didn't "pile" on you for having made a bad prediction (which happens and actually has to happen: if you don't fail enough with predicitions, you aren't predicting often enough), i attacked the fact that you didn't put your money where your mouth was when the gap between your opinion and market prices was huge

No it doesn't. Manual would be changing stuff for a specific instance only, "manually". Tweaking is changing stuff structurally which changes a lot of different queries results, some more than others.

Tell me how your life personally is better because of this

I understand that you want to deny that censorship = leftism, and i mean, you do you, no problem.

I didn't link that claiming it was nobel prize winning material.

Anyway if you want other sources for the claim that the Gemini chatbot is biased (if the image generation bias wasn't enough), let's check gemini itself

Only positives for vegetarianism (when asked for the positives)

Lot of negatives for meat (when asked for the positives)

When bruteforced about talking only of positives, at least it complies

Obviously I disagree that the left has a monopoly on censorship, but we need not reach that claim. Nor do we need to reach the claim that those limericks from Copilot are evidence of bias, which was almost equally as ridiculous.

A chatbot having a policy of not engaging on political topics is not "censorship". Are you saying that if someone asks you about a topic you don't want to discuss and you respond "I'd prefer not to discuss that" that you are engaging in censorship? How is a private company running a chat bot different?

You are literally saying that the chatbot refusing to engage on any political topics is evidence of left wing bias. Just take a small break from posting and think about that.

I like how he mentions in passing that child exploitation and terrorism support will be censored, after the diatribe about how sacred it is that nothing is censored. I’m sure that won’t line up precisely with ideology. Sounds like racism in, trans kids and any opposition to the Donald out. Don’t really see how they can force companies into hosting content they don’t want either. This is the kind of dumb **** that I don’t see getting very far.

Same way T-Mobile is forced to give phone service to people they might not like.

My actual claim would be a little more complicated but it requires you to allow that gemini is biased.

Why does the fact that gemini got hard-wired against giving explicitly political commentary strengthen the idea that it is biased? because , given it is biased, google had no way to make it so that gemini was capable of offering neutral political commentary.

In order to avoid another huge-proportion **** up like that which happened with image generation, they just censored explicitly political topics. But they didn't want to block all conversations from happening about topics that have political implications and the bias is clear there.

I'm quite happy to take you at your word that it's biased per your definitions. I don't know much about how AI works; I understand it's trained on a bunch of human text. If that text comes from mainstream sources, you'd probably consider it left leaning, in which case the bot will end up left leaning. I don't really want to get into the obvious debate of what exactly "left leaning" means, especially since we already know that we disagree on definitions in this regard.

The study you linked is absolute dogshit and evidence of **** all other than the author's desire to reach a certain conclusion, complete incompetence, and hyper partisan hackery.

The fact that Google turned off political discussion for their bot is ipso facto evidence of **** all, and is certainly not "censorship". If you want to argue that the reason they turned it off was that if they turned it on it would be biased, then fine, see my first paragraph. They probably just didn't want to have to deal with people like you perpetually whining about their bot's responses so they disabled the functionality.

The LLM (large language models) that are the basis for these chatbots are trained on a ton of human generated text yes. That content comes from a myriad of sources.

Problem is that the developers then apply "AI alignment" techniques.

They introduce external, human inputed preferences, ethical rules, and so on. For example to avoid (or at least reduce the probability of) the AI helping nuclear researchers in Iran develop the bomb quicker.

You might have seen "jailbreaking" examples around, of people using convoluted prompts to make the AI talk about stuff it was "hard-wire", externally , not to talk about (or talk in ways that the enforced "ethics" wouldn't have allowed). Anything from promoting violence to use very racist language and so on.

So it's undeniable those chatbots have a layer of permissions/ethical rules/legal rules and so on applied on top of what they came out with, from the pure unsupervised learning conducted on the text they got fed.

And we have good reasons to believe that google and microsoft explicitly put left leaning ethos based rules to control those chatbots. And that's for example why grok (twitter AI) has different rules.

This doesn’t sound right to me. I’d be curious to hear what Rococo or another lawyer here thinks.

What you say on a t-mobile contract isn’t broadcast to the rest of the world. That seems like a fundamental difference. T-mobile is more like your internet provider than Reddit. I also don’t see how they can stop companies from writing user agreements that restrict content but I’d like to hear from anyone that has knowledge on the subject.

again, you can.

But then sec 230 can be re-written excluding you from immunity for civil liability of all content posted on your plaftorm.

The state can also decide that some companies are big enough that they have a duty to provide service even to people who post "bad things" according to their ToS but that's another topic. It should hold in court to allow the companies to remove all content that is against the ToS but not allowing anyone from being banned from facebook for example.

What a "basic service" is , is decided by congress at the end.

"why did the right obsessed so much with trans issues!"

Blueprint, a leftist organization which works to try to win elections for democrats

https://blueprint2024.com/polling/why-trump-reasons-11-8/

Blueprint

Why America Chose Trump: Inflation, Immigration, and the Democratic Brand - Blueprint

/Kamala Harris is focused more on cultural issues like transgender issues rather than helping the middle class/ was the single most important issue for swing voters lol (while Israel appears to have been mostly irrelevant)

Right, but my point is their opinion on what’s lawful speech or not is going to align with ideology.

And he says users should be allowed to have an unmoderated stream of content, meaning the site will have to host it.