>where conservative advocates maintain it suppresses free speech and censors right-wing content—something professional fact-checkers vehemently reject.
Definitely proven to be occurring.
>The quality and accuracy of AI chatbots can vary, depending on how they are trained and programmed, prompting concerns that their output may be subject to political influence or control.
The political bias by AI seems rather left wing. Bing copilot is shockingly far left. I conclude the misinformation will be thus biased.
Grok is mentioned 10 times in the article. Yet it's one of the least biased. Perhaps the article should have been done against bing copilot?
All of these AI combined probably produce a billion fact checks a day, they will get some wrong.
mdp2021 · 2d ago
The most shocking idea is that of people asking the "artificial moron without even a proper mental lab to verify ideas" for fact-checking. The very thing that has the most difficulties in discerning facts is said to be increasingly asked about controversial information.
In front of that, which is the point, that LLMs move poorly navigating corpuses full of bias is very much not a question about whether the biases have more affinity towards those with one hormonal imbalance vs another - it is a question about when the will finally we get the Real Thing - Intelligence - in an amount sufficient to fight its overwhelming deficiency.
incomingpain · 2d ago
I think that's what makes Grok so much better than the rest. The early controversy around chatgpt's bias was built into grok. It does have that perhaps call it metadata or however they handle it. So it's able to cut through the bias.
The problem is that sometimes it comes to fact check conclusions which arent politically correct enough.
Ultimately, fact checks by AI are going to be far more reliable than reading journalist opinions.
mdp2021 · 2d ago
> it comes to fact check conclusions which arent
There exists an LLM that does "fact checking"?!
> fact checks by AI
We have to get to said "AI" first! At this stage, I do not even know of a project for that.
incomingpain · 2d ago
>There exists an LLM that does "fact checking"?!
One of my favourites is perplexity due to sourcing being the first thing provided.
For the prompt: fact check white genocide in south africa
It immediately provides reuters, bbc, pbs, and yahoo sources to fact check.
Well laid out, whereas say chatgpt hides the sources down the page and doesnt link specific pages. Grok sources 15 web pages at the very end; but doesnt say which or what.
mdp2021 · 1d ago
I don't call that "fact checking", I call that a "search engine" (possibly with a natural language interface).
If a NN then wanted to express its opinion on that, it would be replied the same as the son who wants to get married: "When you will be older and wiser".
Definitely proven to be occurring.
>The quality and accuracy of AI chatbots can vary, depending on how they are trained and programmed, prompting concerns that their output may be subject to political influence or control.
https://www.trackingai.org/political-test
The political bias by AI seems rather left wing. Bing copilot is shockingly far left. I conclude the misinformation will be thus biased.
Grok is mentioned 10 times in the article. Yet it's one of the least biased. Perhaps the article should have been done against bing copilot?
All of these AI combined probably produce a billion fact checks a day, they will get some wrong.
In front of that, which is the point, that LLMs move poorly navigating corpuses full of bias is very much not a question about whether the biases have more affinity towards those with one hormonal imbalance vs another - it is a question about when the will finally we get the Real Thing - Intelligence - in an amount sufficient to fight its overwhelming deficiency.
The problem is that sometimes it comes to fact check conclusions which arent politically correct enough.
Ultimately, fact checks by AI are going to be far more reliable than reading journalist opinions.
There exists an LLM that does "fact checking"?!
> fact checks by AI
We have to get to said "AI" first! At this stage, I do not even know of a project for that.
One of my favourites is perplexity due to sourcing being the first thing provided.
For the prompt: fact check white genocide in south africa
It immediately provides reuters, bbc, pbs, and yahoo sources to fact check.
Well laid out, whereas say chatgpt hides the sources down the page and doesnt link specific pages. Grok sources 15 web pages at the very end; but doesnt say which or what.
If a NN then wanted to express its opinion on that, it would be replied the same as the son who wants to get married: "When you will be older and wiser".