Now in some states, you can't trust teachers not to be giving you misinformation.

Nope, we all misunderstood what they meant. Wikipedia is not an authoritative source, it is a derivative work. However, you can use the sources provided by the Wikipedia article and use the article itself to understand the topic.

Wikipedia isn't and was never a primary source of information, and that is by design. You don't declare information in encyclopedias, you inventory information.

Not to trust wiki as a format? Or did you mean Wikipedia specifically?

There's a lot of problems with Wikipedia, but in my years editing there (I'm extended protected rank), I've come to terms that it's about as good as it can be.

In all but one edit war, the better sourced team came out on top. Source quality discussion is also quite good. There's a problem with positive/negative tone in articles, and sometimes articles get away with bad sourcing before someone can correct it, but this is about as good as any information hub can get.

It's worth checking out the contribs and talk regarding articles that can be divisive. People acting with ulterior motives and inserting their own bias are fairly common. They also make regular corrections for this reason. I still place more faith and trust in Wikipedia as an info source more than most news articles.

So very much on-script though

Tankies don't think Wikipedia is the devil. You could call me a tankie from my political views, and I very much appreciate Wikipedia and use it on a daily basis. That is not to say it should be used uncritically and unaware of its biases.

Because of the way Wikipedia works, it requires sourcing claims with references, which is a good thing. The problem comes when you have an overwhelming majority of available references in one topic being heavily biased in one particular direction for whatever reason.

For example, when doing research on geopolitically charged topics, you may expect an intrinsic bias in the source availability. Say you go to China and create an open encyclopedia, Wikipedia style, and make an article about the Tiananmen Square events. You may expect that, if the encyclopedia is primarily edited by Chinese users using Chinese language sources, given the bias in the availability of said sources, the article will end up portraying the bias that the sources suffer from.

This is the criticism of tankies towards Wikipedia: in geopolitically charged topics, western sources are quick to unite. We saw it with the genocide in Palestine, where most media regardless of supposed ideological allegiance was reporting on the "both sides are bad" style at best, and outright Israeli propaganda at worst.

So, the point is not to hate on Wikipedia, Wikipedia is as good as an open encyclopedia edited by random people can get. The problem is that if you don't specifically incorporate filters to compensate for the ideological bias present in the demographic cohort of editors (white, young males of English-speaking countries) and their sources, you will end up with a similar bias in your open encyclopedia. This is why us tankies say that Wikipedia isn't really that reliable when it comes to, e.g., the eastern block or socialist history.

MAGA and tankies are pretty much the same except MAGA votes while tankies whine.

That instance is fucking bananas

They are scared of facts.

Perplexity is okay with more academic topics at the least, albeit pretty shallow (usually isn't that different to google). There might be a policy not to include encyclopedias, but it would be an improvement over SEO garbage for sure.

It's not that AI don't or cannot use Wikipedia they do actually, but AI can't properly create a reliable statement in general. It halucinates so goddamn much, and that can never, ever, be solved, because it is at the end of the day just arranging tokens based on statistical approximation of things people might say. It has been proven that modern LLMs can never approach even close to human accuracy with infinite power and resources.

That said, if an AI is blocked from using Wikipedia then that would be because the company realized Wikipedia is way more useful than their dumb chatbot.

I’ve been asking a bunch of next-to-obvious questions about things that don’t really matter and it’s been pretty good. It still confidently lies when it gives instructions but a fair amount of time it does what I asked it for.

I’d prefer to not have it, because it’s ethically putrid. But it converts currency and weights and translates things as well as expected and in half the time i’d spend doing it manually. Plus I kind of hope using it puts them out of business. It’s not like I’d pay for it.

Curious what and how you're prompting. I get solid results, but I'm only asking for hard facts, nothing that could have opinion or agenda inserted. Also, I never go past the first prompt. There be dragons that way.

asymptomatic long-COVID

The hell even is that? Asymptomatic means no symptoms. Long-COVID isn't a contagious thing, it's literally a description of the symptoms you have from having COVID and the long term effects.

God that makes my freaking blood boil.

Damn @BigBenis@lemmy.world that was a hell of a conversation you we having.

"Cool, send me the actual studies."

*crickets*

I believe DuckDuckGo is just as bad. I think they changed their search to match Google. I’m not sure if you are allowed to exclude search terms, use quotes, etc.

I had a song intermittently stuck in my head for over a decade, couldn't remember the artist, song name, or any of the lyrics. I only had the genre, language it was in, and a vague, memory-degraded description of a music video. Over the years I'd tried to find it on search engines a bunch of times to no avail, using every prompt I could think of. ChatGPT got it in one. So yeah, it's very useful for stuff like that. Was a great feeling to scratch that itch after so long. But I wouldn't trust an LLM with anything important.

I think that the people who did the research will continue doing the research.

Yes, but that number is getting smaller. Where I live, most households rarely have a full bookshelf, and instead nearly every member of the family has a "smart" phone; they'll grab the chance to use anything that would be easier than spending hours going through a lot of books. I do sincerely hope methods of doing good research are still continually being taught, including the ability to distinguish good information from bad.

So now the AI trawls other AI slop, so it's essentially getting inbred. So they literally need you to subscribe to their AI slop so they can get new data directly from you because we're still nowhere near AGI.

As an expert in my engineering field I would agree. LLMs has been a great tool for my job in being better at technical writing or getting over the hump of coding something every now and then. That’s where I see the future for ChatGPT/AI LLMs; providing a tool that can help people broaden their skills.

There is no future for the expertise in fields and the depth of understanding that would be required to make progress in any field unless specifically trained and guided. I do not trust it with anything that is highly advanced or technical as I feel I start to teach it.

FWIW, wp:Category:Internet search engines.

Lemmy also does code markup with `text`

https://www.google.com/search?q=%s&udm=14