NewsNotFound: An open-source, unbiased news company
81–90 of 176 posts
Re: NewsNotFound: An open-source, unbiased news company
#82(meaning media that doesn't fact check their lies and doesn't tell facts they dislike)
If you tell a chatbot trained on the internet to be unbiased how do you know which definition it will apply?
Re: NewsNotFound: An open-source, unbiased news company
#83The two sentences in the article that tells me this will be a failure: “ Once we have a list of related headlines, we then use GPT-3 to generate a list of facts, not opinions, that are present within each of those headline’s respective articles.” “ We then run the article through a “bias checker,” which uses sentiment analysis to rate the sentiment polarity of the generated article.”
Re: NewsNotFound: An open-source, unbiased news company
#84Re: NewsNotFound: An open-source, unbiased news company
#85But it's an admirable experiment. Impressive that Josh Waller (that's the name on the github repo) went all the way with this.
My initial impression was that it is some kind of satire but he seems to be playing it straight. If that's the case then I just don't get the preoccupation with "unbiased".
Journalism always has a bias at some level. Whether it becomes a problem that needs to be addressed is a matter of degree and intent. I would even say that streams of "just facts" without context, without some point of view are uninteresting to a reader unless they're performing certain kinds of tasks which benefit from such a stream-- but that is very different from "news".
Re: NewsNotFound: An open-source, unbiased news company
#86The two sentences in the article that tells me this will be a failure: “ Once we have a list of related headlines, we then use GPT-3 to generate a list of facts, not opinions, that are present within each of those headline’s respective articles.” “ We then run the article through a “bias checker,” which uses sentiment analysis to rate the sentiment polarity of the generated article.”
Re: NewsNotFound: An open-source, unbiased news company
#87Similarly, a few days later security forces went to arrest the perpetrators of the attack, and were themselves attacked with Molotov cocktails and large stones thrown from a high floor. They returned fire, and one of the attackers, a 15 year old boy, was killed. The BBC headline read "SpecificCountry soldiers murder 15 year old Foobar boy". No mention was made of the attack on the security forces, nor of the attack on the car to which the security forces were responding. At least, not in the first few paragraphs that I read, which mostly concerned themselves with other incidents in which Foobar citizens were hurt or killed by SpecificCountry forces, with absolutely no context.
Nothing published was an outright lie. But the omission paints a bias by itself. How could this AI-powered "unbiased news company" form an unbiased version of events, when the relevant information is not even provided by the sources?
Re: NewsNotFound: An open-source, unbiased news company
#88As someone who has worked in the media industry, I can tell you that the problem with unbiased news goes much deeper than simply presenting facts objectively. The very first step mentioned in your breakdown - scraping the latest headlines from publicly available news sites - is inherently biased, because it relies on a small selection of sources to determine what is considered newsworthy. But by whom? In reality, edi…
However, I realized that’s not really true.
I mean, of course you cannot have unbiased news, and reevaluating the definition of “news” is not going to help you.
It’s just that there is actually a very good mechanism for combatting this—and it’s called “be aware of your bias and upfront about it”. There is any other way around it.
Re: NewsNotFound: An open-source, unbiased news company
#89As someone who has worked in the media industry, I can tell you that the problem with unbiased news goes much deeper than simply presenting facts objectively. The very first step mentioned in your breakdown - scraping the latest headlines from publicly available news sites - is inherently biased, because it relies on a small selection of sources to determine what is considered newsworthy. But by whom? In reality, edi…
I used to think that unbiased news is an intractable problem purely based on information theory: to have an unbiased picture of the world, you need to have all the facts, which is physically impossible; and any layer that simplifies those facts (be it a journalist or an LLM) will add bias (either by journalist, journalist’s employer, LLM’s training data or operator preferences, and so on). In short, when your selecti…
1. Identify all clusters of power and opinion, eg. Republicans, Democrats, Chinese Communist Party, UK Labour Party, Green Party, etc.
2. Associate each cluster with news organizations that tightly align with their policies.
3. Evaluate stories - headlines in only one cluster (under reported elsewhere) represent a view. Headlines that have different takes, likewise, give you a spectrum of thought on an issue or topic.
4. Collect fine points and build a report. Bonus for providing contrasting opinions and associating them with the party/power responsible for that thinking.
I want this tool. It'd be okay if it wasn't perfect. Just getting 80% of the way there would be incredibly valuable. I'd pay for it.
Re: NewsNotFound: An open-source, unbiased news company
#90Earlier quoted context omitted.
I used to think that unbiased news is an intractable problem purely based on information theory: to have an unbiased picture of the world, you need to have all the facts, which is physically impossible; and any layer that simplifies those facts (be it a journalist or an LLM) will add bias (either by journalist, journalist’s employer, LLM’s training data or operator preferences, and so on). In short, when your selecti…
Unbiased news algorithm: 1. Identify all clusters of power and opinion, eg. Republicans, Democrats, Chinese Communist Party, UK Labour Party, Green Party, etc. 2. Associate each cluster with news organizations that tightly align with their policies. 3. Evaluate stories - headlines in only one cluster (under reported elsewhere) represent a view. Headlines that have different takes, likewise, give you a spectrum of tho…