Google tests A.I. tool that is able to write news articles
nytimes.com
nytimes.com
State some facts as reported by others (never link to supporting documents only summarize them)
Quote from expert or eyewitness on why this is important
Quote from other expert why this might not be important (optional as this might make the story not truly news worthy)
State that confusing aspects of the story are "unclear"
Fill the rest of the story from articles on the same subject even if not related (link to your publication only)
* don't forget the children
* emotions are very important
* strip all declarations out of context. The article is too long anyway.
* don't make exact statements. Be as vague as possible without compromising the spirit of the article.
* last but not least: construct a beautiful story. It might not get read now, (some people have work to do) but it will make the basis for your future Pulizer prize.
My guess is that the argument will be that the “tools will simply help augment rather than replace the writer.”
We know, however, that some small segment of implementations will actually in fact, cause someone to be fired or replaced, and things will get slightly worse for the consumer.
An alternative explanation will be “well The writing wasn’t very good anyway, it was all a regurgitation of the same thing over and over anyways”, so “nothing of value has been lost.”
I’d love to hear people saying hey maybe let’s not do this in order to maintain a human system run by humans for the benefit of humans and that includes paying humans for labor (when you could actually get equivalent labor from a machine) because it’s important to keep humans alive and creating communities and supporting each other.
I imagine that the new AI-enabled jobs are going to have so much more autonomy and that we'll see a switch from capital intensive businesses employing lots of people (eg. Hollywood studios) to lots of tiny indie houses with fewer than a dozen folks, building exactly what they want with freedom.
This gets said about every wave of technological driven job loss and it ignores that it is quite difficult for people to change careers in their mid life. Laid off auto workers in the 80s and 90s didn't switch to other manufacturing jobs or "learn to code". They fell down the ladder into service work and became destitute. And the difference with LLMs is that they threaten essentially every human not working in manual labour with the same fate.
> I imagine that the new AI-enabled jobs are going to have so much more autonomy and that we'll see a switch from capital intensive businesses employing lots of people (eg. Hollywood studios) to lots of tiny indie houses with fewer than a dozen folks, building exactly what they want with freedom.
I'm sorry but I can't reconcile this with my view of what is actually going on. The entire point of LLMs being introduced in the film industry is to remove humans from the equation to increase profit margins and avoid having to pay actor/writer residuals. That is where all of the investment money is going, not some utopia where the investment money is being split between AI studios churning out "content" and some small human, indie studios making "art".
Also, one of the ways the indie film scene has survived is as a proving ground for up and coming actors/actresses (it being part of the career ladder for the acting profession provides a supply of actors/actresses despite low pay) who can then go on to more mainstream films and make their living. If we cede the mainstream to AI then that career progression breaks down and indie films become wholly personal passion projects by folks with enough money to have an extremely expensive hobby.
More than exist today probably.
We should not be regularly eliminating slightly less efficient processes that use human workers, for completely alienated processes that are worse, that use fewer workers.
Self checkout is a perfect example of this. It’s worse for everyone except for grocery store owners.
As for lost jobs, should we really be propping up unnecessary jobs that really don't do much except poorly substituting a machine? Whether it's checkout or journalism or coding or politics... if a computer can do it better, why not let it and free up the human for other things?
Technology leading to job losses and changes has been happening for a couple centuries now. It has real social impacts, but the answer can't just be "pretend it didn't happen and keep employing them". Maybe our economic and educational systems need to evolve to keep pace, rather than keeping people stuck in assembly line monotony just to keep them meaninglessly employed.
Cause this has never been the majority outcome in the history of technology
The outcome in 90%? of mechanization/automation is not upskilling - it’s displacement.
Work gets outsourced to cheaper labor areas or you’re now a baby sitter instead of a tradesperson that is more easily replaced because you have few skills.
All the “meritocracy” goes out the window when there’s no need for human expertise - so everyone is the same faceless commodity that simply enforces the rules of whatever mechanical system they are now monitoring.
At the end of the day, where once a person went to be a cashier for 8-10 hours a day having conversations and fixing problems and helping people - now you’re on your own waiting for the one person who is running the self-checkout to come over and fix the weight error because you thought you typed in the UPC for a pear but you accidentally typed the UPC for a watermelon and the system required human assistance
Everything is worse and you’re doing someone else’s labor and they now have fewer job options.
> Cause this has never been the majority outcome in the history of technology
Work isn't disappearing.
The US has great employment numbers that have been stable except for a few systemic shocks. With globalization, more people worldwide have been employed and lifted out of poverty than ever before.
Would you rather pull a random job out of a hat from today or fifty to a hundred years ago?
Self-employment is also a rising labor trend. What's more fulfilling than being your own boss?
About the jobs, yes, we are in agreement about the displacement. I'm saying we need better systems to deal with that displacement at the social/national level, not at the micro level of an individual job going away one at a time.
There's just no way capitalist companies are going to keep unnecessary labor employed. Staff are a cost that eat away at profits. I'm not saying that's a nice way to think about it, but that's how they work, and unless we can directly tackle that part of the mindset (that humans are there just as an input to some owner's dollars out), we won't really solve the issue. Bandaiding them through forced employment laws won't really work as our companies just become less efficient and companies outsource. You can't really have a protected economy like that unless you're also willing to limit outsourcing (which I'd be for, but capitalists who control the government aren't).
I would love to get paid a living wage doing something less brain-wracking than coding (like journalism or barista-ing). But our economic system doesn't make that easy, or even possible. It's going to get worse over time without deep structural change...
Why not both? You can decide whether you will support or boycott organizations and systems that actively pursue these things, while also lobbying and voting and building organizations that pursue mutual voluntary organization. Might mean you have to change your lifestyle though and most people don't want to do that.
>Bandaiding them through forced employment laws won't really work as our companies just become less efficient and companies outsource. You can't really have a protected economy like that unless you're also willing to limit outsourcing (which I'd be for, but capitalists who control the government aren't).
Literally nobody suggested this in this thread, me or otherwise.
>There's just no way capitalist companies are going to keep unnecessary labor employed.
I would challenge you to imagine some solutions that are neither capitalist nor governmental. If you can't think of anything beyond these two options then I suggest reading more about mutual-voluntary organization.
Sorry I wasn't clear! I can see how that looks like a strawman... what I meant is that I would like to see a more protectionist economy (by the government). Apologies for the unclear phrasing.
> I would challenge you to imagine some solutions that are neither capitalist nor governmental. If you can't think of anything beyond these two options then I suggest reading more about mutual-voluntary organization.
> Why not both? You can decide whether you will support or boycott organizations and systems that actively pursue these things, while also lobbying and voting and building organizations that pursue mutual voluntary organization. Might mean you have to change your lifestyle though and most people don't want to do that.
I do try to support community organizations of various forms (State Farm, REI, 501s, ESOPs, food co-ops, solidarity economies, church-charities for disaster relief and homelessness, etc.).
But I think there are some problems at scale that are uniquely suited for government-like entities with national policymaking reach, and technology-driven job displacement is one of them. I'm not enough of an anarchist to see a well-functioning government (which, to be clear, the USA does not have) as terribly different from other forms of democratic decision-making. Rather than having a bunch of small shadow governments parallel to the federal and state ones, I really wish we could transform those actual governments into something with more direct democracy, especially at the local levels. That may or may not happen in my lifetime, but I think that's a better long-term solution than a bunch of small scrappy NGOs and alt-societies popping in and out of existence, almost always with very limited reach and impact. Sometimes they feel more like performance theater than actual social forces. I think one can only witness so many Occupy-like uprisings before starting to seriously doubt their effectiveness.
I believe it takes a certain economy of scale that only government (and huge multinational corps) have access to, that's needed to be truly transformative. Some unions have reskilling programs, for example, but it's not really enough to save an entire profession/industry.
But... I would love to be proven wrong. Do you have any examples of non-governmental cooperative/democratic organizations that can effect large-scale socioeconomic change?
I'm doing my own labor. Personally I often don't want to have conversations while doing my grocery shopping, especially not the fake corporate kind. Having to have such conversations as an employee of a supermarket was a very unpleasant experience, lacking in dignity.
You make a lot of good points, but others are extremely half baked.
I have regular ongoing relationships with the employees at the Giant (Grocery store) near my house and know their names and who they are as people and they know me and my kids and it's great.
To the extent where together, myself, Sandi (Night Manager), the Asst Manager and some of the staff all worked together to help get a wonderful man back on his feet that was struggling and begging and living outside nearby. (edit:In this case he was begging, and I sat and talked with him about his situation and getting him what he needed for that day. Then I recruited the staff to watch for him and I would leave a bag of food every week but they needed to give it to him, so that's what they did. He ended up moving in with family and I saw him again (at that same Giant) but instead he was shopping with his family and he told me that we all collectively saved his life.)
So yes building community is important and you can do so with the individual people that work within these corporate systems to great effect and for the tangible and measurable benefit of human flourishing today right now.
edit:I can't reply to comments on this for some reason
Do you mean you and a few employees there just got together and started, what, a shared pot of money for emergencies? Can you please share more about how that happened (or maybe even a blog post)? I'd love to try something similar.
Basically, I just took it upon myself to help this guy but I needed the staff to help. Because I already knew them over years of interaction, it was an easy conversation/ask that they agreed to and we were successful in giving this guy a hand up. No govt or coercion or anything needed.
The specifically half-baked thing is the idea that by doing something for yourself you're stealing someone else's labor opportunity. I bag my own groceries (at any checkout, self-serve or not), for 3 reasons: I want it done in a particular way, I don't like being waited on in general, and I get exercise from humping a heavy bag of groceries home in a backpack.
By your logic I'm forcing the bagging assistant to move to another register and depriving autoworkers of their economic future. Your heart is definitely in the right place, but I don't want to be part of your church.
This is the mindset of "rugged individualism"
Which has nothing to offer society (in which you live) for a way to live. You are mortally bound to need and care for others. Choosing not to does not absolve you of the responsibility, it's simply lacking the character necessary to provide for your own care via supporting, fixing, building a functional cooperative society.
This is exactly what I meant when I said I don't care to join your church. I can see why you'd object to the sort of individualism that says 'I won't help others because people should take care of themselves.' However, you've now extended this to the idea that people who do practice various kinds of self-sufficiency should not be allowed to, but must be be in a position of needing others whether they want to or not.
This is an absurdity. Is a person who cuts their own hair to be forced to go to a barber or hairdresser? Should programmers have to obtain a permission slip before writing code? Should hardware stores refuse service to DIYers? Your values as stated deny anyone the possibility of self-sufficiency, solitude, or privacy.
Not everyone thinks so. I know quite a lot of people (including myself) who hate self-checkout.
> As for lost jobs, should we really be propping up unnecessary jobs that really don't do much except poorly substituting a machine?
No, but equally, should we be kicking people out on the street with no means to support themselves?
I think that's the wrong question. The better question is how can be ensure that people aren't just out of work and can still be able to afford to live?
The butter churners didn't just die, they became professionals doing something else.
This is more problematic than it seems, though. What if there isn't enough work in other industries to absorb the people? If AI works out as proponents want, this seems likely because lots of jobs, across a wide swath of industries, will be eliminated.
There's also the issue that not everyone is suited for every kind of job.
People often adapt when big changes come to their industries. They don’t adapt when their entire livelihoods disappear. LLMs look to have the potential to do this not just to a segment of one part of the economy in some regions, but to the entire knowledge economy the world over.
They lost their homes and health care and died earlier than they had to, leaving nothing but broken families and communities.
Yet they always end up augmenting the scammer, grifter and undermine those that want to remain articulate.
A potential upside is: lowering the marginal cost of page filling drivel so much in conjunction with declining advertisement revenue, the whole industry collapses in on itself, taking opinion pieces with it. The baseline for getting out of journalism school would be "write better than an AI", and the reporter jobs themselves become high status / high difficulty again.
A potential downside is: nothing written ever matters again, all information is a rehashed AP cutting with prompt engineering, and AP / Reuters become a set of controlled bottlenecks.
The scary thing is: this will be decided by whether the paying customer wants to be informed, or feel the warm glow of confirmation bias?
That's throwing the baby out with the bathwater. Opinion pieces will be the last to go, after everything of value has been destroyed.
And honestly, not all opinion pieces are bad. Most are crap partisan hackery and outrage-bait, but some are very valuable (e.g. those unconnected with ideological bickering, those struggling to explain a point of view to hostile audiences, those "swimming against the tide," etc.).
Wouldn't opinion pieces be the first to go?
We're already in a world where most opinion pieces are part of carefully orchestrated PR campaigns (sometimes even wholly ghostwritten) meant to drive policy discussion in one way or another. If you can AI write this and ascribe it to a reputable source, why wouldn't the industry adopt this as standard?
No, because they're relatively cheap to produce. Everyone has an opinion and many people compelled to share and argue about them.
The hard, expensive thing is gathering facts, doing deep investigation, or cultivating relationships so the facts come to you. That's why the intersection between blogs and journalists has always been opinion and punditry, since no one has the time to do anything else when they have a day job.
> We're already in a world where most opinion pieces are part of carefully orchestrated PR campaigns (sometimes even wholly ghostwritten) meant to drive policy discussion in one way or another. If you can AI write this and ascribe it to a reputable source, why wouldn't the industry adopt this as standard?
Isn't that an argument for the idea that "opinion pieces will be the last to go"?
No because people are interested in the opinions of influencers.
There's unlimited ventures for investigation for anybody who wants to be a journalist, but it takes much more work than rewrites. I read a lot of newspapers, and the news part consist of about 50% rewrites from other sources, 40% mindlessly repeating press releases from companies and government, and at most 10% investigating and actually reporting new knowledge. Even without machines we already have enough tech that re-writes shouldn't be needed.
“ An alternative explanation will be “well The writing wasn’t very good anyway, it was all a regurgitation of the same thing over and over anyways”, so “nothing of value has been lost.””
In fact something has been lost, which is a human being is now less able to feed themselves and their family and take a vacation and create something valuable.
The fact that you and others don’t care about that outcome is the problem.
Is it going to produce Pulutzwr-winning articles right off the bat? Probably not. But I could easily see it replacing the more mundane everyday articles, especially when it's just regurgitating some press release or new study or product launch.
"when it's just regurgitating some press release or new study or product launch."
Maybe news media should start moving away from this model? They can link to the press release and be done with it. The valuable thing they do is their own reporting and investigating.
But yeah agreed, wish papers just republished wire stories or press releases and spent what little staff they still have on original (ideally local) reporting.
I certainly hope we would force them to identify. I don't have problems talking to an AI, but when there is a reasonable expectation that I should be talking to a human...it should be a human.
Essentially what we know is that Google (with a financial incentive to make exaggerated claims) has demoed something to media execs (with little to no background required to critically evaluate these claims).
I think it’s safe to say the intention of every party here seems to be aligned with the notable exclusion of an actual writer
Writers and journalists aren’t quoted in the article and I see no reference to them being included in the product pitch (article only discusses executives being given demos)
So in terms of incentive alignment, the person being impacted seems to be the least involved.
1/ Technology has always disrupted labor market so I mean that's not even a question this will happen.
2/ It's not that sad: I've worked in journaism and the reality is that 80% of news article are already most of the time a sad copypasta of a single main source of info such as reuters / AFP in france. Most of journalism is already noise.
Maybe that will help recenter journalism on real investigative missions rather than this current low value state.
Doesn't matter to who?
I'm pretty sure it matters a lot to people working in a newsroom
So yes it’s extremely important that employment changes not be thrust on people if we can at all avoid it as a society.
You should care more about the plight of others.
I an in Europe and wealth is a lot more redistributed. AI will just amplify wealth accumulation but it's the policies who have to step up to counterbalance it with coporate taxes. Blaming AI is like blaming rain for being wet.
The fact that "AI" makes that a shibboleth is correlated, but by no means is AI the only process for which that happens
It just so happens that's the current trajectory and is expected to impact the labor market that to an extent that has been unseen since the industrial revolution - with billions more employees
How is the LLM getting the latest news? It doesn't (yet) really have eyes and ears etc.
And news travels fast, or at least people would like it to.
Also opinion pieces. I have 0 interest in the "opinion" of an LLM, but I am interested in the opinion of at least some human beings.
According to the OP: "[t]he tool...can take in information — details of current events, for example — and generate news content." IMHO, that's not really journalism at all, which I would define as gathering that information and deciding what bits are worthy of being reported (e.g. what's new and interesting vs what's already well-known and understood).
Honestly, this tool seems like another example of techies redefining the problem to what their program can do, instead of writing a tool to solve the problem. I would imagine that the "writing" part of a news article happens nearly automatically as a side effect of all the other activities a journalist has to do. Sort of like how the "writing a program" part of developing software will nearly automatically fall out of the task of writing a sufficiently precise specification. A tool that can translate such a specification into code is a nothingburger (despite its maker's inevitable hype about rendering software engineers obsolete), because that's not where the work is.
> I’d love to hear people saying hey maybe let’s not do this in order to maintain a human system run by humans for the benefit of humans and that includes paying humans for labor (when you could actually get equivalent labor from a machine) because it’s important to keep humans alive and creating communities and supporting each other.
I totally agree on this.
I might be fruitful to develop an AI tasked with hunting down and exterminating software engineers and tech executives who seek develop AI to replace human labor. /s
I have no idea though. Predictions are hard to make, especially when they're about the future.
Looking at the trend in global newsrooms and the way they've already been gutted in the name of dividends how likely do you think that's going to be?
More like this will be used as an excuse to cut even more journalists and, worse, to run local and regional news without any journalists at all, with the few who remain being as under pressure and overworked as ever.
Let's imagine that, currently, a journalist spends 50% of their time conducting investigations and 50% of their time sitting at a keyboard typing up their findings. The ideal hope for a technology like this is the future then consists of 90% of a journalist's time being spent conducting investigations and 10% telling a text generator to type up the findings for them and cleaning it up. Overall, more time spent investigating and following up should result in higher-quality, more interesting work.
It may appear to consumers at our end that the work of a journalist is just writing because the writing is all we see, but this is not the case.
No royalties.
Also, no further employment.
The shame of it is, this is actually slightly better than the current state of things, where no one who contributed writing to the training sets in any form is receiving compensation. But it is the same issue as one of the sticking points in the SAG-AFTRA strike.
News media today are nothing more but a vehicle to bring attention to advertisement. The majority of people do not read lengthy articles anyway, only headlines. Investigative journalism is wildly expensive and not cost-effective given said objective.
News media are anachronistic anyway. People are stuck in childish and dumbed down variants of "my parents told me so".
What is necessary (and severely lacking) is two-way (or rather, public) communication. Enabling OSINT. Enabling people to form well-founded opinions. To discuss and receive constructive criticism of their views and ideas. Facilitating self-reflection on a society-level.
Can't agree more, but how? Any insight?
You need a functional format for group discussions (existed, killed off). You need functional standards for civil discussion (exist, see science, people somehow forgot). And you need to realize, anonymity is anathema to having a civil sphere (You don't need to give your real name, but for social self-regulation to work, an identity is necessary).
AI quite possibly could help a great deal with this. Sensible automated (and consistently reasoned/reviewable) moderation of internet forums is wildly underexplored.
Editing other people's stuff is even worse, because many writers aren't burdened by editorial concerns and don't bother to put in the effort to make their story coherent and interesting in the first draft. Also the money is terrible unless you are lucky enough to work in a corporate newsroom.
Won't help you when you have to edit the AI output because…
> Editing other people's stuff is even worse, because many writers aren't burdened by editorial concerns and don't bother to put in the effort to make their story coherent and interesting
So in the end, what do you gain?
Personally I think this is going to put a lot of journalists out of work. Many of them aren’t investigative journalists and do mostly write, and their days are numbered.
We need to have AI monitor the news and rate their level of factuality by consistency analysis and reputation scoring. Journalists in the loop, working with AI to find disinformation. Then we can feel safer about propaganda-LLMs, they are inevitable.
You are aware that, prior to refrigeration, people would harvest ice from the Arctic? Honestly, I don't care they all lost their jobs if it means millions of people don't die during heat waves.
Umm, are you sure about that? Source?
well, yes. you clearly understand the problem, but stating it dismissively doesn't refute it.
paying people to do minimal rewrites of press releases or wire service articles is not productive work. at all. people would be better served by reading the original source.
a job that requires no skill and provides no benefit isn't really a job, it's just a selectively-applied welfare program, disproportionately provided to the people with the best connections. putting a stop to those sorts of "jobs" is not a bad thing.
As long as you have somewhere else to put them...or else they will go on another type of welfare program where they don't have to do any work.
That is the issue here...
If this requires AI, so be it. I will pay for it.
The AI is not able to construct a lie so credible like some journalists.
This conflation is a critical flaw of reasoning. It will increase the tendency of so-called journalists hitting the "write copy using AI" button to spew out content. It will not, cannot, and should not replace critical thinking skills, interviews, or research. I'm curious how this was pitched by Google and the thinking of newsroom management. Tools should augment and automate mechanic work such as an enhanced LexisNexis search, not substitute or delegate human expertise.
At some point, without necessarily donning a tin foil hat, we ought to reflect on the possibility of computing power and automation misusing ability beyond the filter bubble of selective "resharing" for user persuasion to the realm of generative faux organizations, personalities, and content. Elaborate digital hallucinations intent on gaslighting and arguing with users individually.
Doesn't this already exist in some form? I remember headlines from a few years ago about how stuff like wire reports about sports games could be generated from a box score.
We seem to be reaching peak generative AI hype.
Some things that I've learned so far:
- The AI is too gullible. If I ask it to write a short summary of an article, if the source article is trying to shill a product or service, the AI will replicate the salesman discourse. I tried adjusting the prompt to see if I could make it more critical of content that it is analyzing (by telling it ahead of time that the post might be trying to push a product/service).
- Costs are ludricously low. It costs like 1 cent per 15 articles.
- My next experiment will be with local news. I'm building some feeds with public information from my town (the town's hall official news, the legislators weekly meeting notes, weather reports, waze, etc), and based on that make it generate news items. The thing about it is that its sources will be (nearly) primary - it will not copy content from other journalists (apart from the official prefecture news, which I will need to tell the AI that will be biased towards the current administration). When analyzing the local records, it might be able to catch shady stuff that regular journalists would not notice. Imagine feeding some buying orders from the town and asking the AI ("is something illegal going on here" or "are any of these items overpriced?")
- I see a risk/opportunity for infinite content generation. Example: generate 50 headlines for articles about the Kardashians. The next day, ask for more and provide the last 200 headlines, to make sure that no repetitions occur. It would flood search engines with almost random content. I think that something like that could be useful to fill "holes" in wikipedia, though.
The site that I built is https://dev-radar.com/
These days I have to scroll through a lot of fillers in articles just to get into the real story.
(extremely, heavily italicized and bolded: /s)
[edit: typo]
This contractor shift paired with the revenue pressure from the internet’s impact has made journalism a real tough industry.
I wonder how LLMs writing news would impact the contractor/libel situation. I could see it freeing up resources to then pay w2 journalists worth protecting for bigger stories. Or maybe LLMs writing the controversial stories and seeing who gets sued in that situation.
truthiness.ai
edit: wow, it's actually registered.
Google engineers: I'm I so out of touch?
[ beat ]
Ge: No, surely it is the users who are wrong.
Google's modus operandi has always been to automate people out of the equation. This lets Google have its own news operation, without paying for all those pesky journalists.
I wonder if the AI will eventually write stories about any of the bad things that Google does, or just the good things.
There is no fingerprint or smoking gun in textual prose to distinguish between bullshit human writers and generative AI.
If we wished to prove humans are behind content, then we should all be posting pictures of today's newspaper with the handwritten message we intended to convey. Imagine how much less shit social media would contain if that were the barrier to hitting the "post" button.