Anti-scale: a response to AI in journalism
tylerjfisher.com
tylerjfisher.com
It's clear that folks have found new formats for their sources of truth. Whether those new sources of truth are better or worse is irrelevant. You have to meet people where they are. And guess what? They aren't on unusably ad-laden news sites with inscrutable layouts and useless search tools. They're on TikTok.
Of course, I've also read a lot of garbage from those respectable outlets. A lot of slant and a lot of poorly-researched "journalism". This is basically a respectability politics problem; the NPR class is offended more that people have abandoned the trappings of "proper information" than over the information itself.
TikTok is not a place to get news. The culture isn't there, and the UI purposefully doesn't lend to things like sharing sources.
Joe Rogan vs CNN is a good example - that you can apply all the same doubts you have for Tiktok also to conventional news outlets.
This is not true at all. Podcasters have no consequences for deceiving, and can easily recover from it by saying they're not journalists and that ends the subject there.
Try that with a News Outlet, then it would be Joe Rogan saying they're liars and are compromised.
There should be way more accountability for those platforms that have no code of ethics to abide by. Even the self-proclaimed "intellectual dark web" is extremely opinionated about a wide range of subjects way beyond their domain of expertise.
I'm not saying all news outlets are pristine, but comparing journalism, as an institution, and putting it at the same level of accountability as podcasters is just wrong.
And I'm still highly suspect of print media, I try to stick to individual journalists and editors who haven't gotten on my shitlist. Grassroots journalism is so good these days for many issues that it's getting easier to avoid conventional outlets altogether.
> Whether those new sources of truth are better or worse is irrelevant.
Just because the platforms people are using aren't ideal for journalism doesn't mean we can sit down, cross our arms, and say "well I guess we just won't give you news then." TikTok doesn't give special treatment to large or well-established news organizations, but that doesn't preclude such an organization from using the platform to reach an audience.
Social networks are the most democratized ways of spreading information to date. Some of that information is lies, but a lot of people spread fact.
“A [Rutgers Univerity] study […] asserts TikTok’s algorithms promote Chinese Communist Party narratives and suppress content critical of those narrative“
https://www.kqed.org/news/11999273/tiktok-stacking-algorithm...
And how do you know which is which, hmmm?
Yellow Journalism is a term for over 100 years ago. People lying with a broad audience is nothing new.
addendum: and how many times have you read a news article from the New York Times or the Wall Street Journal that's in your area of expertise that's made you go "wow, everything they've said here is wrong, said with an agenda, misinterpreted or other propaganda"?
But social networks have increased that problem by an order of magnitude. And we're seeing the results - the rise of dictatorships all around the world, that use the lie spreading capability of social networks.
We as a collection of societies, worldwide, have not grown our broad-range antibodies/protections against such things, yet.
For example,
huh. turns out theres another nationwide conglomerate running the sinclair playbook
I'm not saying don't ever be critical of the media you consume rather be nuanced in doing so. Do realize that the decision which to produce/what to report on/what counts as "news-worthy" is an opinion in itself. Deciding which headline to use is an opinion in itself. Choosing your lede down to your last sentence is an opinion. You can't ever "just report the facts"
"US Olympian ends Moscow campaign with silver" could be the same news item as "Russia takes gold at home vs US rival" but you can see they are expressing very different opinions already.
To be clear I can't agree more with "less but better" but that act is inherently---you guessed it---very opinionated. I can already read the "Hey BiasBroadCast, why you report on this war but not this other war?" all the way from here.
This is where any discussion about “the media” falls down because it’s too broad. There are news organisations that work hard and reporting the news. That costs a lot of money, though, and a lot of people don’t pay. But someone spouting off their opinion on that reported content on YouTube? Basically free to make and gets the clicks.
The media is far from blameless but I think we give consumers a pass. They’re the ones choosing the vacuous, empty opinion coverage.
It’s hardly surprising, look at who is graduating from journalism schools, what they are taught and what their agenda is. The consumers did not create this problem, they’re simply powerless to resist it - except that they can take their attention elsewhere, and that is exactly what they’re doing.
Information being bottlenecked through a handful of institutions has been increasingly harmful to society and is what has led us to this very moment. Let the people speak.
That business model has failed over and over in the past few decades.
This is a bit of a confusing comment, and it's almost like an algorithmic reply - which anecdotally is the status quo.
Maybe you can develop this idea a little bit?
Because the "Alternative Media" is for the vast majority opinion and entertainment-driven, people have no problem consuming that information and taking it for facts.
This is done with ZERO accountability or responsibility, because once in a while it's thrown "Hey folks, I'm just an actor/comedian/hobbyist/retired X" and then spreads misinformation.
Unlike News outlets, if they get something wrong and acknowledge it, all hell breaks loose. It is extremely unbalanced to demand perfection from News while at the same time having a loose leash for alternative media — many of alternative media celebrities have way more resources than news outlets, and no standards.
It's almost like people prefer the illusion of understanding subjects based on sound bites (like, for example, your broad generic remark), instead of trying to think a bit more deeply and look at things for themselves. No wonder, this takes time, and effort, and most of the times it ain't fun to do research.
You had an election cycle that was heavily influenced by years of repeated conspiracy theories, post-truth commentary, and attacks on institutions like science.
Most of this content has very little quality or barely any production value, being one of the most common formats just people sitting at a table with microphones - they're not Netflix productions. Yet you claim News outlets' production is of low quality.
So yeah, it would be great if you could expand more on your idea instead of making a shallow remark like it's coming from some great insight you've developed yourself. Which news organizations are you referring to? Can you give some examples? What would you do differently to improve on those examples?
For me, news is the "new" things, not some LLM generated regurgiation of old facts. I am really saddened for last few years since the emergence of genAI, because everyone(who do not understand more than hype info) somehow assumes that, genAI can start writing latest news and articles of events or somehow predict(Minority Report is not real!) future events.
But, these LLMs know only as much _as they are trained on_, and by the time they are being trained on latest news, those are no longer news at all.
I see how News can be immune to genAI because of the "new" factor of it, but the greed of automation of money-printing is too much and reality is still not sinking in.
As an alternative, news reporters can have somekind of twitter(X) like channel somewhere, where they can post latest events, bits of investigation and an omnichannel genAI can scrape these live and start combining and summarizing the events into more cohorent report, which could give the news management the greed of plugging genAI and printing money, while reduce the workload of journalists of sitting in and writing up that whole report ... I honestly don't know, why journalism even needs genAI, just because everyone is using it to generate slop does not mean I must jump in.
Journalism is competing against post-truth entertainment content, that thrives on conspiracies passed as facts. How can they compete with that, when facts are boring and go against their beliefs?
> Also, a lot of investigative journalism takes work and fearless collection of sources and other dangerous activities.
For years you've had journalists being murdered in Russia for doing their job, and you had Tucker Carlson doing the classic "Russia lives better than us - look at their grocery store, they have nice bread!" - which ones do you think had more reach and had a greater impact on people? The actual journalists who risked and paid with their lives, or Tucker Carlson's lies?
Because that's the competition of journalists, millionaires sitting in the comfort of a podcast studio, with no accountability or responsibility, and a massive reach.
This is what we're dealing with here. Yet you demand so much from journalists, isn't this unbalanced?
In the sphere of information, LLMs at the moment are more a tool for the amplification of deliberately produced disinformation at scale, rather than a way to create/develop information. That's the true problem in my opinion, not in terms of developing content.
> As an alternative, news reporters can have somekind of twitter(X) like channel somewhere, where they can post latest events, bits of investigation
Do you genuinely believe this wouldn't be hammered down by the algorithm of "it's a platform being controlled by the deep state/spreading lies/echo chamber/woke news"?
Remember people rallied for Elon Musk to buy Twitter, with the banner of free speech, only to support him converting it into a right-wing/MAGA propaganda platform - and they still support this, claiming its better than ever. Until recently we could claim "its a loud minority", yet here we are.
Now, the newspapers themselves have "adapted" to internet consumption by dumbing down and commercializing their own content to be at par with internet slop. Live news, on the internet as well as on TV, are a curse on consumers.
A single newspaper subscription is often more expensive than Netflix, offering far less production value. No wonder classical journalism is losing market share among young people to more affordable and relatable formats.
Not so for journalism. There are very few pieces people will return to years after they’re written. It requires constant publishing of new content and to make good content costs money.
Some explanation: https://blog.forth.news/a-business-model-for-21st-century-ne...
It’s a matter of opinion how that’s going I suppose.
The quality has been abysmal already.
Once upon a time we had a weekly newspaper in Sweden that did this. Watched the powers that be and reported on them. Government, corporations, individuals that wield power was exposed/illuminated for me to draw my own conclusions. And it didn't matter what color/team that happened to rule the government. Sadly they hired on of the top political pundits (that I 80% of times don't agree with) as editor-in-chief and my trust in their mission disappeared along with my (too cheap) subscription fee. I guess this is harder for you Americans to digest with your division (and we are heading your way as well) but this is journalism worth paying for and maybe ai actually is the solution here, not as text producer but as one source for the actual human journalist. A source that is good at extracting structured subset of (verifiable) information out of a large corpus. Eg how has this individual politician voted during his/her career.
Meanwhile, as a source of revenue for news organizations, it seems no worse than advertising, particularly since unlike ads, it doesn’t require them to degrade their websites.
The money might not last and they shouldn’t count on it continuing. But it seems okay to take it when it’s available?
It’s a bit like people saying they want organic cruelty free food if you interview them. Yet the cheap stuff sells too.
If you actually subscribe to the wsj/economist/whatever - adds up quick. So few do it.
No market means not much room for journalists. Doing journalism better won’t fix that unfortunately
AI can only do the second half. They need information to write about.
If the money goes away to support the whole enterprise, who actually makes the phone calls?
It cannot make things up, and it cannot be wrong, ever.
There is distrust in the media, much of that is for good reason. There have been countless examples of news stories that lie, misrepresent, obfuscate or simply neglect important information. That those stories exist places all stories in some degree of doubt. The lack of trust comes not from disinformation, that's always been there, but the awareness of disinformation.
The only thing that people will believe is a news source backed by a reputation, Reputations take ages to build and seconds to shatter. There is a huge scope for machine learning to safeguard against the incidents that shatter a news organizations reputation. Humans are too fallible. There is a chance that AI might not be.
The author bases his argument on the presumption listed under the heading
Why generative AI is not the answer
What follows is the usual tired list of criticisms, starting with the ever so tired 'designed to predict the next word in a sentence'. When people hear this what they imagine is something like a Marcov-chain. That is not what LLMs do. If you want to argue that predicting the next word in a sentence (given massive context) cannot lead to whatever you say the AI will not be able to do, devise a task that cannot be completed by performing next word prediction. Bear in mind that LLMs also add to their own context as they generate output creating complex patterns of behavior.
If you wish to deny that architectures made of many smaller, simpler, components can be more than those simple features, then you are going to have to take a serious look at brain cells.
Yes people are concerned, and yes AI's are not that great, their interfaces are not very good right now. The same could be said for Cars or any number of technologies as they emerged.
There are some optimists who see an AI analogous to Pong(which is where we are right now) and can imagine an AI analogous to Crysis. They might be over-optimistic, but denying that there is any possibility of it seems downright foolhardy.