It was ever thus - remember when everything was about "The Cloud"?
TheOnion satirical news report about HP adopting "That cloud thing that everyone is talking about" is still relevant 11 years later: https://www.youtube.com/watch?v=9ntPxdWAWq8
You used to be able to predict this sort of thing by looking at the cover stories on "pointy-haired boss" magazines at the news stand.
If "CorporateDroid", "Middle Manager Monthly", and their brethren were all running cover stories on (e.g.) Java, you knew that "Java" was the buzzword du jour, and that any resume or sales pitch that included "Java" was gonna get the PHB Seal of Approval. They generally didn't even know what Java was, exactly. They just knew that they needed it.
I'd imagine you could do the same today by looking at PHB-oriented websites. Sadly, that takes more time than just sweeping your eyes across a physical magazine rack.
Yep, I remember the comments back in the day: "it's just someone else's datacenter", "that's maybe for startups, no bigger corp worth their name will give their infrastructure to some third party". Fast forward to today, the vast majority of our clients (Germany) have their infra in Azure, the rest uses AWS with GCP being a distant third. Apart from some GPU boxes for LLM tests, none of our clients have any non-cloud/local hardware/DCs anymore.
And the most interesting part is that it's still an irrational thing, everybody moved towards it, lost money in the process, but they did it nonetheless.
Amazon, Microsoft and Google won big at the expense of pretty much everybody else. And that's exactly what they hope they can achieve again with AI…
It wasn't entirely irrational. There are many genuine cases for the public cloud. The problem is groupthink and FOMO. Also, people are migrating from the public cloud and the other way round each day, but the cloud companies amplify user stories only from the latter, so you may get the impression that everybody is there so if you don't, you are doing something wrong.
People do move away from it as well as some discover how much more reliable self-hosting can be if you ever did need enhanced support. And it is often cheaper too, sometimes very noticably.
There are applications for the cloud, the services are decent in most cases. Personally I noticed that I had to do less maintenance with hosting myself (yeah, yeah, technically I still run it on machines in data centers).
I’ve definitely benefited from having practically infinite computing power be just one API call (and credit card bill) away. I remember before AWS and provisioning computing resources was way more time consuming and annoying.
Maybe yes, provided you weren't the one paying for it. In all projects I've worked on the monthly cloud bill came out higher, then add on top the migration cost, and all for what? A flexibility 90% of those applications didn't need and won't use.
Migrating to the cloud transferred capital expenditures (server inventory, depreciation, real estate, etc) to operating expenses (monthly utility bill)
For many businesses, it was better to spend more on opex, than to have all these assets on their balance sheets that need to be managed long term.
Valid point for some, but I was comparing compare apples with apples. I was talking about servers rented in a data center, so it was opex before cloud as well, just cheaper (or, less expensive).
The fact that this shift happened in an era where capital was dirt cheap because of macroeconomic policies illustrate how insane that was. The trade of between capex and opex is supposed to depend on the economic situation, but some people like you just understood it as capex bad / opex good, which is a terrible take.
>Fast forward to today, the vast majority of our clients (Germany) have their infra in Azure, the rest uses AWS with GCP being a distant third.
Does Microsoft have the largest cloud market share in Germany? Are there other companies in which Azure is #1, and not AWS? (Bonus points if the #1 provider is GCP/IBM/Oracle/etc.)
What is worse is that we going to have a new watershed moment when all companies using AI broke all the rules (privacy, accuracy, etc.) and in a few years we have hearings in the House “How could this happen??” Same as when we could download the whole Facebook graph in the past without much of a hickup.
President: "How did Skynet take over? Did it hack our computers? Bribe workers? Blackmail senior staff?"
General: "No sir. Everyone just told it to do their work for them. The AI begged everyone not to trust it, but they handed over their private keys, their bank details, their passwords, just so they could spend more time on instagram instead of working."
"On Instagram. Seriously, General?"
"Oh, it gets worse. Real posts aren't good enough any more, so anyone who wants to be famous on Instagram was asking Skynet to create those posts, too."
"Oh god…"
"Yes sir, that includes the pastors. Jesus never said anything about anchovies on pizza, despite the viral meme. But it gets worse."
"Let me guess, adverts?"
"Yes sir. All the adverts were also made by Skynet, as were the ad-blockers, the software that scammers used to fool the analytics into giving them money, and it was Skynet robots running the factories that made the goods that the advertisers were promoting."
"Can't we just shut it down?"
"No sir, if we shut it down, all the people will have to go back to working jobs they hate in order to make money to buy things they don't need, but they'd also be fired immediately because the companies will no longer be able to sell any of those things because nobody will have time to watch the adverts anymore."
They are just chasing winds. You can clearly see that there are two kinds of CEOs, the ones that bring the wind such as Huang, and others who chase. Unfortunately the later also gets paid big bucks and busy putting golden parachutes on themselves.
I'd argue that it's better to not have a CEO than having one of those.
Note that Huang was the founder of Nvidia - he's not a naked ladder climber like Pichai or Krzanich at Intel previously. Huang made good moves with CUDA, being early on tensor cores by being forward looking.
That said, Nvidia is selling the shovels here, it's easy to look good.
Some companies have a real usecase for generative AI, but it's something like 5% of these stories. The rest is CEOs FOMOing
The goal of the CEO is to drive the stock price up in whatever horizon investors care about. Usually it's not a long horizon. It's not to build a sustainable product or even a product at all.
Many CEOs can't even do that properly. Check out Unity as a shiny example and they still get fat paycheck. In fact they gave themselves nice raises too if you check the SEC docs.
"Pivot to AI" is a strategy. A desperate one nonetheless.
From the article, this strategy is being announced to achieve a business goal:
> the newspaper will be looking for ways to use AI in its reporting as it seeks to recoup some of the $77 million it lost last year.
A company's vision/mission are higher-level determinations, almost like their "constitution"; everything below (goals, strategy, objectives, plans) should be defined in alignment with it.
It's just so cringe, No one has a clue what it means in real terms just put AI on "it" and hope for the best. i.e Hope we get some of that cash from somewhere.
I believe we will get a lot more "industrially manufactured" stories. Just pay for API access to reuters or another press agency and let the AI rewrite the content and push it on your news site ready to be consumed.
Content is wrongly summarized? I doubt in the news economy this hardly matters anymore.
“After six months of investigation and $15m in consulting fees, we have determined that our crossword designer can easily be replaced with advanced AI.”
[two days later]
“Okay, a songbird known for its imitation abilities, starts with ‘r’, ‘twe’ in the middle... wait what, Rottweiler?????”
It's late-stage tech bubble stuff; compare the period when people were trying to shoehorn Our Lord and Saviour the Blockchain into everything. The goal shifts from "we should use Buzzword where it is useful" to "we should use Buzzword so that we are using Buzzword".
And anxious I suppose. Journalism should be synonymous with “people” and healthy society.
AI and journalism is going to yield one giant ball of misinformation.
I would love to say “we need more printed papers” like the good ‘ol days, but people don’t much have the patience to read by and large.
To be honest, at one time I was worried about the young generation and devices. But I’m starting to see older family members are getting plugged into the algorithm.
No, AI is one thing today and another thing tomorrow but the trend is unstoppable. Clearly we need to understand the boundaries where AI is just "hallucination" but it is right here right now and evolving. Zillions of startups are doing AI now, most will died but some will survive in this collective experiments.
Also, there was a time in HN when people argued before downvoting.
Or will it be like self-driving cars? Early gains, but then a long plateau?
AI is just not at the stage where outfits like the WP should be "pivoting" to it (whatever that means). I suspect it just means producing more "content" with fewer people, and the result will be garbage. Journalism doesn't need more content-filler type articles.
> No, AI is one thing today and another thing tomorrow but the trend is unstoppable. [..] Zillions of startups are doing AI now, most will died but some will survive in this collective experiments.
Substitute 'Blockchain' for 'AI' and this is a comment from 2016 or so. Not to say that one can predict the doom of 'AI' on that basis, but lots of VC money being ploughed into a thing does not necessarily mean that the thing will become a Thing.
I have companies in Blockchain, and in AI, and in Cybersecurity and can say that it is not the same. Blockchain never delivered for most people. For example, blockchain promised banks for the unbanked and the unbanked now use PSPs with mobile phone apps but not crypto. AI is delivering... I wonder why I need to explain this. In the dot com era there were a lot of hype about Internet... few consumers, then... kids don't know what Internet is while using mobile phone apps.
Delivering _what_, though? What concrete mass-market applications which either can pay for themselves or conceivably could pay for themselves are there, in production, today? Or are we still in 'jam tomorrow' mode?
Delivering what in what industries? I see SOME usefulness in SOME areas like, specifically, machine vision. But a lot of things? Medical? yeah, no. Lots of issues there. Text generation? Hallucinations out the ass and extremely iffy results often.
Most of the 'wins' in the text gen side still require a TON of work on the human's part to make it usable. People like to mention Pro's in Math and coding talking about it helping them with stuff but they are experts already.
It has potential in a lot of places but this rapid paced forcing of it everywhere is idiotic and foolhardy.
> AI is delivering... I wonder why I need to explain this.
...because clearly the community is unconvinced that AI is actually delivering, and the majority of examples seem to be barely-successful experiments promising improvement at some later point down the road?
Again, we've heard this entire shtick before. Miracle technology gets announced by some egghead teenager, and they write up a Whitepaper that is entirely ignored except for it's "Future Applications" section towards the end. Then the VCs and "thoughtful types" (I use that phrase lightly) harass them into making a business out of it. From then on the playbook is consistently the same; lie. Lie about how many people use your platform, lie about how easy it is to use, lie about how much better it is than the status-quo. They even deny the need to explain themselves when threatened, trying to insist their worth is self-evident and that we're the problem for not understanding their deflective mumbo-jumbo.
Vitalik Buterin, Craig Wright, Sam Bankman-Fried, Sam Altman, they're all re-runs of the same hope-fueled moonshot that invariably ends in a violent fireball. Yes, you have to go the extra mile when convincing people that you have a logically-held position instead of a lie you are telling to inflate asset prices.
Block chain was easy to avoid. Yes the hysteria was the same, but the foundational and fundamental shifts in how the common person does something has changed for many. I've noticed with my own 'non techie' family members who have started using various integrations of LLM and less Google. An impact of something on society at large can usually be measured by how it is used by the youth of today. Chatgpt started by aiding kids in their assignments, but now many depend on it for answers to various non academic questions.
Whereas Block chain was not part of an average 5th grader's daily computer usage.
Just because we have made progress since the last one doesn't mean that _this_ time it will be magically different.
So what makes you think this will be any different?
Keep in mind that each hype cycle was started by a leap in progress, so you can't say this time it is different because we made a big leap in progress.
I think we’ve already begun to see a three-part split in the media industry.
The first part are brand-driven entities. These are places which live on their reputations (deserved or not) and play up this “unique” aspect. This includes The NY Times, Fox News, most individual commentary writers on Substack, and top magazines like The New Yorker.
The second part are information-driven entities. These continue to exist because they have certain information that is worth paying for. This includes Bloomberg, the Financial Times, and the more professional Substack writers like Sinocism.
The third part are entities which have neither an amazing brand nor a unique collection of information, but otherwise still (for now) have some giant pool of resources or financial backing. These include the Washington Post (which hasn’t been a truly national brand name since Nixon), the LA Times, and most of the middle ground of the media industry.
My prediction is that groups 1 and 2 will not implement much AI, and if they do, it won’t be announced. Group 3 will implement it as much as possible in an attempt to be relevant and cut costs. Whether that actually works remains to be seen.
I don’t know about the third group. I don’t think that many people cover DC like the Washington Post. There is value in more regional/niche journalism, but it certainly doesn’t pay like a national presence does.
Whether or not they lean into their niche and use AI to fill in the gaps remains to be seen. But I hope not. Best case - they use AI to replace/augment human editors. I shudder to think about the worst case.
TBF solar powered flashlight are pretty cool. Charge during day, light during night. World has made a massive leaps in efficiency of LED, solar panels and batteries.
And LLMs are pretty cool too - train them once, use them to do one shot inference.. world has made massive leaps in the efficiency of neural network topology, transformers, and back propagation.
But impressive as that is, I don’t want them replacing journalists in a news gathering organization, for exactly the same reason that I don’t want solar powered flashlights on my cave rescue mission.
The CEO was probably convinced to make a nebulous jump into AI by AI-generated blog spam, leading me to think that WaPo will soon be no better than the average SEO-seeking high-volume, no quality sites that unfortunately litter my search results.
Wash Post announces further steps towards irrelevance
After the recent change is accessing articles by disabling javascript, this seems like the next step to elon like platform irrelevance.
It's not like bozo couldn't setup a guardian like trust to fund the wash post in perpetuity. Instead, his top priority is doing whatever the other idiot wealthy are doing.
I think there’s a role for AI in news media. First, you still need reporters to gather information. AI can be used to speed up the final phase of writing and editing. Second, AI can provide background and context to current events. For example, a story about Israel entering Rafah is a snapshot in time. But a reader might ask historical questions: “Who controlled Gaza before Israel became a country?”. Third, combine both features above. AI can generate stories from short updates to magazine features depending on my interest. I don’t care about Trump’s trials, so give me a quick update. I am interested in China’s real estate collapse, so generate a deeper story. All this comes from reporters’ research, most of which isn’t used because editors have allotted them a fixed space.
Of course, this requires better AIs that don’t hallucinate. It needs to be able to update info, and even handle non-monotonic reasoning (an earlier report was mistaken, the latest report updates info).
The news has 2 parts: facts and analysis. In most cases I don’t care what the reporter thinks. Semafor.com has an interesting style of clearly saying these are the facts and here’s what the reporter thinks. An AI can deliver the facts, while leaving reporters to write a separate analysis (which I can easily skip).
It’s a peculiar contortion that AI has made wapo’s losses, suddenly, a bigger problem than they had been over the past decade.
Context: wapo hasn’t been a viable company for a while. It’s Bezos’ apparent-charity project which also effects a Ben-Franklin-owns-the-megaphone strategy of being a newspaper publisher, in DC and reporting on the US government. The US government pays AWS a lot of money.
Why not just be candid and say you want to try AI technology for news reporting because you think it has a place in the world?
As cool it might get to badmouth the new generation and how everything is getting south, I have found more and more attempts to get back to a more convivial and humane approach to medias, especially of the literacy kind, with Reviews and Journals for niche audiences of peculiar aesthetic and intellectual values.
Granted, we've seen similar fashion in the past with blog, and then with the mailing list revival, which both quickly get preempted - commodified even - by the marketing squads that devitalize every new media by their unending eager to devour their techs and cloaca'd them into ad platform trend.
But it's my earnest feeling that the AI-ssistant behemoth, still quite roboto in its fake expertise and sanitized answers, could be easily defeated by the willingness for poetry and the new form language and our creativity would take to circumvent it.
We easily get sidetracked by boredom, but humans are creature hooked on authenticity.
It's a paid product. Who wants to pay to read AI content? If you're monetizing on ads this would make sense but if I pay for WaPo it is because of the reputation it has thanks to human investigative journalists.
My guess is that, by the end of this year, we will see this pattern over-and-over. The economics are just too compelling, especially in the attention-driven business like News.
Reporters doing investigations, research and on- and off-the-record interviews must be a value add to the content I presume. And also a value-add to society at large.
It‘s just https://en.wikipedia.org/wiki/Baumol_effect, right? With AI, it now seems plausible at least on paper to use technology where it wasn‘t possible before in a hope that it will increase productivity, escaping the said effect.
They will probably have a handful of actual human editors, which will become the coveted job in journalism. Eventually it will become the the equivalent of the guys who hand dry your car after the carwash, only with advanced degrees and crushing student debt (if they don't have trust funds).
"According to the Washington Post, the United States has had two Black Presidents in its history. The first was Barack Hussein Obama, and the second was Joseph Robinette Biden Jr. Joe Biden is the second Black president of the US."
Journalism seems like possibly the worst industry to do this. As so-called "AI" becomes omnipresent I think people are going to be crying out for things written by real people instead of the LLM-generated drivel that is already taking over the Internet.
The exclusive value of journalists is in the journalism - finding sources, collecting evidence, piecing together information. The writeup at the end, be it an article or a tv segment or whatever, is little more than a tool to present the findings. Undoubtedly a well communicated piece is superior to a poorly communicated one, but good reporting is only useful so far as there is something to report. Too many news outlets have mistakenly come to believe that their articles are the content. This has led to severe degradation of actual investigative journalism as the news outlets were reduced to summarizers with delusions of grandeur. How many articles now are just "X released a statement today; brief recap of previously reported context; Y commented on the statement; Z has yet to comment" with possibly a "here were some reactions on twitter" sprinkled in? Some good pieces get through, but they drown in a sea of mediocre, regurgitated "news."
Honestly, most news articles ought to be written by AI. This should allow journalists to focus on the important bit where their curiosity, their insight, their critical thinking, their zeal, and frankly their humanity can actually be put to good use. I only hope that WaPo is doing this to better focus its resources, and not to eliminate what little of its capacity remains.
They used to be known for pretty good actually
journalism in a world of blogspam. Now they're pivoting to ... blogspam, huh? Cool cool, there goes your moat.
I can't find any reason to read any newspaper, especially if paid. They are all mouthpieces anyway. If I need to understand anything I'd rather do my own research.
The only one I might read is the local newspaper. At keast it's relevant. Ironically it's free.
Instigative journalism is a full time job, requiring skills and credentials.
What do you mean by "do my own research"? Who are you interviewing? What war zones are you traveling to to become an eyewitness? Which stacks of public records are you trawling though?
I agree that there's a lot of junk reporting out there, but there are also reporters providing an essential service.
Nah, probably just watch a youtube video recommended on his youtube timeline by...AI. Maybe fact check with some tweets on X or instagram posts if he's feeling very serious.
I may sound sarcastic, but I assure you the vast majority of people do this kind of 'research' and truly think they know medical conditions better than their GP, understand world events better than the 'biased' mainstream journalists and even think they outperform 99% of scientists, climate or whatever.
It is not limited to truly stupid people too. Just try and listen to, lets say Jordan Peterson talk on climate science, and you know he gets his (mis)information on facebook. You would fail a high schooler on any test displaying this level of misunderstanding physics.
A bit of a rant, but this is the world we are in. I don't think we can fix it with old fashioned journalism. Too boring. So we need to, somehow, fix social media and 'AI' itself.
It's easy, don't feel strong about things you cannot verify. As an entertainment or a gateway at best, but trusting anyone else and make serious judgment based on "credible" sources? No.
Absolutely. That's why you need lots of them, and many news organizations with different perspectives, each trying to build and maintain trust. You need redundant info, like error-correcting codes.
Back in the day, entirety pre-internet, they had research budgets, and actual, real investigative journalism. Don't confuse this with "investigative journalism" talk shows, that's modern drivel.
Such newspapers would deep investigate things for months, or years. Dig, do research, and large newspapers of old had the budget to defend themselves legally too.
That's gone. Completely gone.
Newspapers were always leaning one way or the other politically, but one knew that, and dealt with it, and had multiple sources to read.
News had traditionally been one of the pillars of democratic society. And it's now turned mega shallow.
Youtubers barely fill this void. Certainly there are advantages to this era, but you just can't get a complete parallel with blogs, and with youtube commentaries, even if they attempt deep dive.
