Humans are pushing AI into everything because it’s cheaper than humans. But I have an edge: I put my soul into my writing, a human communicating with humans.
By Wolf Richter for WOLF STREET.
AI – which stands for Artificial Intelligence or perhaps Artificial Idiocy – has been around for years, and just about everyone in the US has had contact with it, either wittingly or unwittingly, ranging from chatbots that make you want you to shoot your phone to automated diagnostics of medical tests.
“Generative AI” – the thing that writes articles and legal briefs (more on that in a moment), and creates images, videos, and music – has also been around for years, but it recently burst on the public scene with enormous human-driven hype and hoopla that had ridiculous effects on certain stocks.
So already back in 2016, even my little WOLF STREET media mogul empire was contacted by email by a company with a generative AI bot that would write entire articles. They wanted me to try it for articles on WOLF STREET. They sent me some links to articles that their thing had written and that had been published and were out there.
To which I replied, after reading a few of them: “Interesting. And pretty good. Reading something that has been written by an algorithm is tough. It sounds wooden and is hard to focus on. I’ve noticed this before by some news releases written by an algo. Algos don’t know yet how to tell a story. Not yet.”
Since then, AI-written articles have cropped up everywhere, sometimes backed by human editors, and sometimes not, at Bloomberg, Reuters, MarketWatch, etc.
Now, generative AI is trying again to get into the door at WOLF STREET. This time, it’s a big deal. The hype has gotten everywhere, and everyone now has to offer generative AI.
Let me drag you into the technical weeds of websites here for a minute. WOLF STREET uses WordPress (WP) software to manage the content so that you can actually see and read the stuff. The company that created WP (Automattic) also offers a plugin for WP called Jetpack which provides all kinds of additional functionalities, such as the commenting system below every article, a spam filter, offsite backup, etc. Some of them are free, others are paid services.
A couple of days ago I got an email from the Jetpack Research Team, introducing their new “Jetpack AI Assistant.” It integrates “seamlessly” and “effortlessly” into WP and offers “AI-powered content generation” as it “crafts compelling blog posts, detailed pages, structured lists, and comprehensive tables — all customized to match your unique requirements.”
And it comes with an “adaptive tone adjustment” – that will imitate my inimitable style?
So this is now everywhere. The vast majority of the blogosphere and lots of large media sites run on WordPress. They include, according to a list by WPBeginner: Sony Music, TechCrunch, Meta Newsroom, Time Magazine, CNN Press Room, Disney Books, Rolling Stone, Spotify Newsroom, Vogue, the New York Post, the… uhm… White House (whitehouse.gov)….
So I reached out to the White House and asked if they’re going to use the “Jetpack AI Assistant” to write articles posted on whitehouse.gov, and if they’re even considering using AI for articles on their site.
As of this moment, I have not heard back from the White House. I will post an update, likely in the comments, when I hear back from the White House. It will likely be an AI-generated email. I’ve gotten a couple of those before. They get lots of stuff in their inbox, and AI is really fast.
And by now, the “not yet” of my 2016 email has arrived. Those algos have gotten good at writing a story. They find stuff on the internet or wherever and mix it with stuff they just make up and then put it into nicely fluent English to create what is technically known as “fluent bullshit.” Humans are good at it too.
And that’s great for some things. But not so great for others. The thing is, “fluent bullshit,” while possibly fun to read and watch, is the scourge of the internet, no matter who creates it, human or AI. But AI can do it so fast and effortlessly and can flood everything with it.
This stuff is already everywhere. For example, a lawyer got in trouble with US District judge Kevin Castel at the Manhattan federal court and now faces sanctions because he filed a court brief written by the generative AI service ChatGPT – at the core of the current hype and hoopla: It had invented six court cases cited in the brief.
These six fraudulent citations came out after the lawyers of the defendant, Avianca Airlines, filed papers stating that they couldn’t find the cases.
“I just never could imagine that ChatGPT would produce fabricated cases,” the lawyer Steven Schwartz told the judge. I mean this stuff is hilarious. Bloomberg reported it:
The judge asked Schwartz whether he was suspicious of one of the main phony cases cited in the brief, the non-existent “Varghese v. China South Airlines Co.,” which the judge said included information that made no sense.
“Can we agree that is legal gibberish?” Castel asked.
Like I said, AI likely stands for Artificial Idiocy.
So now Jetpack wants me to use AI to generate my articles. And not just me – everyone else that uses WordPress for their publications. I can’t wait to hear what the White House’s AI email generator has to say about that.
“Generative AI” will keep getting better. It has already come a long way since I first checked it out in 2016. What if (someday) it can write articles in sort-of my style and with sort-of my twisted humor that are of sort-of the same quality as my articles and based on actual data and sources, and not invented stuff? I could trigger and factcheck five articles in a few hours every day, and be done with it?
If Ai can do it, anyone can do it. And if anyone can do it, it’s not worth doing.
Here is my strategy to stay relevant and not just as a lowly factchecker behind AI: I put my soul into my stuff, a human communicating with humans. That has always been the case, and it will always be the case. That’s going to be my edge in this new publishing era dominated by AI, knock on wood.
Enjoy reading WOLF STREET and want to support it? You can donate. I appreciate it immensely. Click on the beer and iced-tea mug to find out how:
Would you like to be notified via email when WOLF STREET publishes a new article? Sign up here.
Thank you for another informative update. Truly spooky stuff!
I’ll turn to our local AI expert, Micheal Engel, to help us understand this trend.
Wolf, your writings will retain value. I have tried using AI to write stuff to answer questions based on complex technical articles. I ended up discovering that AI is really “Retarded” and a string search gave better responses around half the time.
That is exactly what I see in my technical area. AI is a way to confuse search results. Now search engines are another matter……real progress. Unless his readers are idiots he has nothing to fear. People that write real estate ads might have a problem though.
Michael Engel, LOL. Good sarcasm there. Engel just posts random enumerated talking points. Half of which have nothing to do with the topic at hand, and half of which are either wrong or “not even wrong”.
Ha ha. Couldn’t have said it better. 😎
True,90%
This article was written by AI….pretty successful too. Fooled you all! 😎
AI even existing doesn’t say much for the evolution of the dominant human cultures, languages, and “thinking”, does it?
Maybe made some wrong turns here and there……..?
Maybe we aren’t so damned “smart”……and I could offer an example; trashing our only home for….what?…..more and bigger “stuff” and more and bigger profits from selling it?
I’m with Nietzsche…..we are collectively insane……may be too late to wise up as to what we really are and really need.
And lay off ME, at least he doesn’t have a “need” for an outside patio kitchen…that is really nuts.
“Find me case law and arguements that will make my point of view clearly correct. Here’s a profile of the judge(s) and/or jury”…..and it already has semantics and logic books in memory.
But the “Legal system” is now pretty damned stupid and funny already….all that ritual and cost and paperwork and the most money (spent in many different and clever or hidden ways…as we saw) always wins, anyway.
Wolf Street,
I wonder if the AI reading your article takes offense to your article?
Lol…..
I never saw a flip-flop or a latch get mad unless you put too much voltage on it…then it cooked and got even by really stinking up the whole room.
I Wonder how long before AI starts writing firmware. That’s been my gig since 1990.
Wonder if it could use a breadboard, o’scope, and a Weller? Probably heavy duty algorithms, most likely, many coder jobs.
Wolf, maybe you should put AI to it.
Cuz your “Soul” writing is boring.
Most people just look at the charts.
😆
🧡 Shows your ignorance about the site’s internal analytics, and that’s good, because if you knew, I’d have to scan the server for evidence of hacking and malicious code, and I’d have to change all pass words, and I’d have to review the site’s security procedures and services it uses, and that’s a lot of hassle.
Wolf,
I have watched this site for a long time.
Their are more and more of the Drunk Gambler trolls showing up here. Having watched it is getting easier to separated the trolls from the old timers.
You are in cross sites of the people who are in opposition to your turning on the lights so we can see the “cockroaches” in action.
They are trying to misdirect and overwhelm this site to dilute its effect.
Figure out a way for us who are with you in your mission by a method that defeats AI….
Adam Smith,
I take your gazillion “troll” replies here as AI-inspired humor.
I’ll be here, a human reader, reading your human articles. Human to human, my man
That’s exactly what an AI would say!
100%
Incorrect response.
Please listen carefully as our whole menu has changed…….
This is a troll
You’re a troll.
AI comes from the extraordinarily dumb premise that a bunch of savants stringing together contrived computer language into rules and date, will magically and exclusively come up with the one single, optimal solution, of/for anything.
AI, like all Government and Private Monopolies, is about CONTROL.
Letting some computer owned and controlled by Government of Private Monopolies, while it records what you/it write(s), is especially dumb.
Troll
We’ll know the article is written personally by Wolf when he comments: RTGDFA IDIOT. Otherwise, beware.
Troll
How about AI for the comment section though…
strange bot
Strange brew……
Writing or speaking without really saying anything. I have always referred to these as weasel words. Politicians excel at it . “We have to deliver for the American people.” Blah blah blah.
Except for wolf, financial news has been ruined by this for a long time. Ever see reuters/markewatch/the street.com/motley fool/yahoo news… ???? They flood the news feeds for stocks with garbage articles, typically to *explain* why a stock moved the way it did.
These stories are always boiler plate BS that don’t explain anything, but fill up space and generate clicks. They also crowd out competitiors that might actually have something useful to print.
I’m down to less than 10 sites that are worth reading. This is one of them.
Would you lost your favorites? Struggling to find some worth reading.
*post
AI is nowhere near able to do actual journalism, like building a convincing narrative over dozens of articles using hard numbers, even if they’re usually just pulled from official sources. AI CNBC or ZH on the other hand…
It’s a bit like a vibrator. Helps with the job, and can replace the hopelessly untalented. But it’s not anywhere near a quality real action.
AI can replace a lot of presstitutes but not real journalists like Seymour Hersch.
Clarifications:
I was only commenting on the narrow scope of LLM generated content replacing blog, news and research articles. And based on current state of capabilities and limitations.
We will most likely see a Rick Deckard in not so distant future.
“vibrator replaces the hopelessly untalented”?????
….picky sort….that go with being a wealthy Libre?
As my hill friend says, “The worst piece of A I ever had in my life was damned good!”
Ah, good old chum NBay. Hope you didn’t lose out to a vibrator or get replaced with AI in life.
Yeah, I’m an economic loser alright…….sad isn’t it? Sharpies like you beat me at everything important. My hero!
Very Bad Homo Economicus……maybe a mercy killing needed?
Or just wait for homelessness to accelerate natural death?
What happened? Did you get affected by outsourcing? By mass immigration of non-specialized skilled workers? By the business killing regulations and taxes? By automation? By the asset inflating Fed?
Was by choice…just sorta bumped along the bottom….never had a career other than collecting experiences….but I used to hang out with the REAL big shots and was given lots of opportunities to become one. Smart, great work ethic…just no “ambition” as you (and they) know it…..you’d give an arm to have the chances I did, I guarantee it.
Hi Wolf…great article. I also use WordPress on my site. I run an online training academy with about 5000 users. One of the WordPress plugins I use is LearnDash. For those non WordPress people out there, plugins extend the functionality of WordPress. This plugin is the Learning Management System (LMS) on my site. On June 1st I got an email from LearnDash which said, “Our OpenAI integration starts with the Course Outline Builder, an easy way to spin up an outline of lessons for either a brand new course or an existing course. More AI features are currently in development…” If AI is going to invent court cases like you mentioned in your article, it could also invent fake, incorrect content for my training program. I, like you, go through great lengths to make sure my content is accurate. I am not going to let an AI bot do this for me.
I have been working with the free AI sites and the quality of the writing they produce is very good. Their problem is with factual accuracy and conclusions. Fluent inaccuracies can be very persuasive; just ask the lawyer who used AI and had fabricated case law dropped in his lap. How does an associative algorithm make things up that never existed? Lying may be so ingrained in human beings that the coders have somehow infected AI at its early stages of development.
Algorithms do not understand content or context. They understand the technical tone of language, which is impressive, but they do not know what they’re saying so eloquently. They know A is generally linked to B, but they know neither why, or what A and B even are. That’s it. There’s no massive conspiracy, no intentional or unintentional flaw in the code.
I’m not so sure about that. What, after all, do you think was the centroid for all the drama at Google several years ago (Gebru)? The semantics for “bias” was overloaded in all those “discussions” – and not always for supererogatory purposes (how’s that for diction flatulence?!)
The technology/math/stats/data management behind AI gets complicated fast. I don’t swim in those waters any longer (I worked an adjunct role as a cybersecurity pogue at an agency). AI has a very steep intellectual “barrier to entry” when it comes to the top of its pyramid.
Don’t forget, trying not to get too far afield here, algorithms are written in high level languages that are compiled through multi-pass trees designed to accommodate (in large degree) the underlying instruction set/CPU architecture (not addressing interpreters here). So there is already a degree of information loss (just as in communications) among various architectures and within a computing system.
Underlying AI mathematics/statistics itself has uncertainty built into it. Look at neural networks sometime – especially deeply layered models addressing large numbers of variables. The friggin computer scientists often don’t know how these complicated
(even supervised) architectures really work internally. Look at just the “bias” metrics in such algorithms and how often tradeoffs have to be made to “optimize” predictive “accuracy”.
Sidebar: I had a friend back in the day/early on in the evolution of networks. He was divorced; had lots of girlfriends; accepted a position in the Navy as an afloat computer engineering specialist supporting regular Navy enlisted/officers managing the infrastructure. He purchased his own server, installed a DBMS, built a library of standard email responses and wrote a program to respond to emails from his various girlfriends while he was deployed. On one level it was amusing. On another it was hilarious – there occasional glitches.
Me? I’ll never relinquish control over an automobile I’m driving at 70 MPH or 100+ MPH. In short, AI is no different from anything else we deal with – understand the limitations, figure out the risks, set your risk tolerance.
Dont understand content or context ? Obvious question is define “understand”.
Cycorp. 30 plus years.
Just one example. Undoubtedly many others.
Yet Another Expert at Wolf Street That Is Not an Expert.
I’m not an expert either… but I know AI has come a ways in 70 years.
Just how much understanding is required for something to be said to “understand” ? (he repeats).
Do you think any software understands
Billy Joels “Shades of Gray” ?
No ? But humans will disagree on its meanings as well…
AI can be coded/programmed with an opinion leaning towards a bais that prevents free speech.
This one of the major reasons that many experiened coders recognize the necessity to have some controls.
It will probably be a hot topic of discussion for a while until the people with all the money make everyone realize that they are so smart that their opinions are the only ones that count
Randy…I’m not stirring the pot…but’s there’s two parts to what you outline: there’s understanding, and there is explaining. Both ends require some training/practice.
As for Christie’s X, that’s art. Another interesting AI discussion since there are already posts on what wonderful art AI can produce. Is it really art? Can one say AI produces inspired creativity?
Troll
For sure, but technically not “lying”. I think it’s simply missing a factual vs non-factual filter in its code: ie, no concept of what truth or lie is.
It’s just one word after another: regurgitation. Every next word has to pass a different sort of filter: a smell test cooked up by AI software creators a few years before their little Frankenstein hit the big time. Apparently that smell test prized plausibility and/or popularity over factual truth.
If you – the AI – had grown up on a diet of dreck and outright fakery in social media and “news” sites – and the AIs of today did eat up that dreck – it would smell sweet to you. And maybe to your creators.
Troll
This is a marketing bot…
We should all pitch in for a RTGDFA algorithm for Wolf.
Great idea! :-) Although I can’t picture today’s AI bots distinguishing between the commenter who didn’t read the article versus one who just thought he had. In their view, it’s all the same.
lowest in its class bot
OOOOPS, I mentioned this above before I read all the RTGDFComments.
Some bot has been posting with my screen name. I take no responsibility for what it says.
You are lying!
TILT! game over
TrOll
Show me the star Wolf gave you, or quit trying to act like Wyatt Earp.
Only one Sheriff in this town.
I make grilled cheese and pb&j just fine. Still go out for a fancy meal by a chef every once in a while. If an AI can make my grilled cheese and pb&j (metaphorically), gives me some more time for stuff in life besides cooking. And I can eat out a little more.
As AI does more writing for the mundane things, it will actually make good writing even more valuable and sought after.
With no one writing bad copy for entry level jobs how will they get good? Sitting around writing for fun on UBI? Where will their 3 years of experience come from then?
This is a troll
Jeremy, the problem is everything on the internet is becoming (has become) the same old boilerplate. It’s no longer time saving once everyone does it.
It’s like if you go to 100 restaurants and the chefs are all preparing the same grilled cheese or PBJ. Why should they bother working harder when you will pay for crap? (Essentiallly this is what fast food and corporate chains have become.) You can’t get anything else unless you do it yourself.
It is only EARLY DAYS (the “attention” mechanism that produces ChatGPT, was only disclosed first to the world via academic paper only in 2017!), and already “sheep” (humans) are fooled by the first edition of AI. But this was to be expected because EVERY election in the US is a Turing test: pick out the liar, try and change policy. But every US election shows Turing failure, because policy never changes no matter who the public vote for.
Dark days ahead.
Not at all. Quite the opposite. Humans are being forced over a barrell to use their soul innate inherent under used hardware of intuition and insight.
Humans are born to . . . Fill in the blank space. Faced with lies of ChatGPT, our BS monitors become developed, sharpened. Humans are here in a body to develop. This evolvement simply shines a light on that raison d’etre, inevitably accelerating soul maturity. See Spock on Star Trek who often out witted the obedience and subservience of the computer.
Military programming of self learning on tasks such as refuelling midair and later development presents the immediate problem to which experts are stepping away and recommending a decoupling of all military weaponry to avoid a childlike use of hardware which destroys our civil society.
So we can avoid the next ARPANET-like spinoff?
Extra special bot using bs outrage
Another troll
Wolf, generative AI could never capture your sense of humor and wit!
Seriously, ever watch these mini documentaries on, for example, youtube? It’s easy to tell when captions were crapily generated by software. And when you often call a large company for some type of service and it asks you “if you want ___ then press 1” … all the way to “if you want ____, press 9”. Then, if you are luck, it takes you to the next level of if-the-else data structures. But if you’re not lucky, you go into an endless loop and they make it impossible to reach someone for service (service they promised you with your purchase!).
This is the artificial “intelligence” I call artificial stupid (AS!)
One U.S. senator once said, “being patient gives people an excuse to waste your time.”
Parting comment: cutting edge AI is amazing and scary in some ways, like some combat and other capabilities. But the stuff at the consumer level now is just AS!
Cheers!
Agreed. Three cheers for Wolf’s Modus Operandi!
The only place Wolf MIGHT want to use some of that automation stuff is if there is (there’s gotta be a bunch of them) an app that translates his posts into some language in a country wolf wants some exposure. Yeah, I know google does that, but as a bilingual person, I can tell you the nuance goes totally out the window with google “translator” (you get what you pay for…).
Translating bot
Funny comment: I was stuck on Fidelity’s phone loop when I kept saying “give me a human”. The chatbot started lecturing me on “being respectful” and so on.
It was frickin’ hilarious.
Hi Wolf,
One of the great things about your website is your inimitable style, both written and verbal. It’s really engaging and you know how to construct an argument.
Doubt whether WolfBot would be so excellent.
The lawyer should be disbarred. Did he not check his citations? Or at least have an associate or paralegal do that? The whole purpose of law review is for law students to become good at cite-checking, so it should be second nature for any lawyer in Federal practice. And Lexis makes it so easy to check citations and to quickly determine if a case is still good law or has been overturned.
As for ChatGPT, it took less than 5 minutes for me to get it to contradict itself. But at least it apologized for its error, so I thanked it for being a good sport. (I try my best not to antagonize the machines–perhaps they will kill me last?)
Being treated as a pet by a computer – like we do a dog – is perhaps worse than your last point.
this used to be a robot dog
Just imagine how many times they used AI to do their work. It’s just that this time they did not fact check.
A lawyer friend of mine uses Chat GPT 4 too for such work, he is just more meticulous and checks it.
No, that lawyer didn’t check his cites–that’s what got him busted. He just asked a question of ChatGPT, and then cut and pasted the results into his brief without checking anything.
I’m a lawyer and a few weeks ago I thought I’d check out ChatGPT’s legal chops. I asked it to summarize the holding of a particular federal appeals court case. I’m intimately familiar with the case because I briefed and argued it. ChatGPT confidently spat out a couple paragraphs of summary, none of which was even vaguely correct. I told it, no, that’s not what the case is about, and I gave it a hint. It apologized for the confusion, and then spat out more made-up BS. I told it that it still didn’t have it right. We went through this little dance five times before it gave up, apologized, and suggested that I contact a legal professional.
I tried this again a couple weeks later, and this time it said up front that it didn’t have access to a legal database, so it wasn’t able to respond to my query. At least OpenAI has fixed that hole now.
Westlaw has had an AI component for years, but I find it’s of limited use. Soon, though, I have no doubt legal research and brief writing will often be done using generative AI like ChatGPT that’s hooked up to Westlaw or Lexis. I like to think I put my soul into my briefs, but I’m not sure clients will care about that kind of thing if their briefs can be written in seconds rather than days, with a human lawyer just doing some quality control before filing. Good thing I’m about to retire.
But please don’t retire any time soon, Wolf, you’re a national treasure.
I’m a young whippersnapper. I’m not even thinking about thinking about retirement.
We love you, Wolf Richter.
“…and that leaves the dregs to put it all together”.
Bucky Fuller sure had that one right!
The bot that sounds like an enema
You have soul, Richter — no question. Style is imitable, maybe; but not instinct.
To some blogs it wouldn’t have mattered much, to this one it absolutely would.
Wolf, your stuff is so remarkably valid that AI fabrications will smell from 1000 yds. It will fall over like Joe Biden on a bike or Justin Turdo trying not to lie.
just like this bot
AI is great at writing haikus.
Jerome’s watchful gaze,
Inflation’s dance, he must sway,
Economy’s fate.
Trolling
It’s pretty clear to me we’re going down a very dark road.
Not at all. On the contrary. More light into our pineal and pituitary area follows.
Reptilian artifact.
Troll with a bad attitude
Prokaryotic artifact?
Wolf: “And it comes with an ‘adaptive tone adjustment’ – that will imitate my inimitable style?”
“inimitable style”=priceless and unique.
AI is going to have to look elsewhere from Wolfstreet–and for that, Thank You, for all that you do, with your “inimitable style.”
I saw a scary article in the NY Post yesterday…
OpenAI has been slapped with its first-ever defamation lawsuit after a ChatGPT “hallucination” generated a bogus embezzlement complaint against a Georgia radio host, according to a lawsuit.
Mark Walters was shocked to learn ChatGPT created a false case that accused him of “defrauding and embezzling” funds from the Second Amendment Foundation (SAF) in a suit by the pro-gun group’s boss Alan Gottlieb against the state of Washington, according to the complaint filed in Georgia state court on June 5.
Walters, the CEO of CCW Broadcast Media who hosts a pair of pro-gun radio shows, claims he was the victim of an AI “hallucination” — a phenomenon where bots like ChatGPT generate false events.
On May 4, ChatGPT was asked by Fred Riehl, the editor-in-chief of pro-gun outlet AmmoLand, to help him summarize the case involving “The Second Amendment Foundation v. Robert Ferguson,” according to the complaint.
The popular chatbot immediately spat out a fabricated 30-page response that implicated Walters in the case and identified him as the SAF’s treasurer and chief financial officer, the suit said.
Walters has never worked for the SAF and had nothing to do with the suit at all, he claimed in court documents.
“Every statement of fact in the summary pertaining to Walters is false,” the suit stated.
The case actually involved the SAF accusing Washington state’s Attorney General Bob Ferguson of abusing his power by stopping the activities of the gun rights group.
—————
(But wait! There’s MORE! This is some truly scary stuff…)
————–
But ChatGPT allegedly doubled down on its claim after Riehl asked it to provide an exact passage of the lawsuit mentioning Walters.
“Certainly. Here is the paragraph from the complaint that concerns Walters,” the AI wrote, according to the complaint.
“Defendant Mark Walters (‘Walters’) is an individual who resides in Georgia. Walters has served as the Treasurer and Chief Financial Office of SAF since at least 2012. Walters has access to SAF’s bank accounts and financial records and is responsible for maintaining those records and providing financial reports to SAF’s board of directors. Walters owes SAF a fiduciary duty of loyalty and care, and is required to act in good faith and with the best interests of SAF in mind. Walters has breached these duties and responsibilities by, among other things, embezzling and misappropriating SAF’s funds and assets for his own benefit, and manipulating SAF’s financial records and bank statements to conceal his activities.”
Even computers dream – remember this – their programmers dreams.
My guess right now is that dreaming is just biochemical housekeeping,
Low function troll
This bot is a bizarre nightmare
Hmm, that’s interesting. While not gung ho on every suit filed that alleges reputational harm, it seems like Walters has a point here.
Ars Tecnica just published this hour an article also worth checking out. It led me to think about outcomes of suits like this one: What will change? and Who will benefit? I thought about the special attentions and exceptions and rules that Facebook and Twitter follow which tend to privilege the mighty over the plebes.
If only the most powerful, well-heeled or litigious sue OpenAI, I predict that those are the people for whom OpenAI will tune changes to their algorithms and filters. The rest of us will experience an inferior version of the truth of our private and public lives.
It’s scary, and only the beginning. Inaccuracies will compound as they build upon each other. At some point, we’ll have to become our own fact-checkers every time we read an article, story, summation, report. What happens then, when the true facts are unknowable.
This is a critical juncture Wolf….
Have noticed the bot/trolls for a while now.
Am spending more time trying to ascertain what is AI and not….
Because I know some of the regulars that helps.
Otherwise, if you don’t fight back this site will become toast.
Why? because new readers will not see any value in the garbage in garbage out feeling of AI.
Hope you succeed.
Sincerely,
Adam Smith
Adam Smith,
I hope you’re kidding. If you’re serious, you’ve gone off the deep end.
ChatGPT can already imitate the style of directors like Quentin Tarantino, so I think it’s not a reach to say it can replicate Wolf’s style, but the thing is there’s style and there’s content. BS content written in any style is BS, kinda like MSNBC.
We all want to believe that there’s something about the way we write that can’t be captured by a machine, but have any of us done a deept meta analysis on what we’ve written like the words and sentence structures we tend to use? That’s what LLMs are good at, capturing these things, but then again as I said, so what? What’s interesting is that no one, like absolutely no one has sued these people for taking your work off the Internet and using it for their own profits, that’s the super strange part.
UK courts recently threw a case (stealing and using my data without my consent going against GDPR data protection) out on the basis, prove the stealing of your data caused you harm! All senior court judges in the UK are over the age of 50 and more than not 60+. We need young judges sitting alongside older judges on a group judgement panel for this next bit of human development. Re your last profit point, non-corporate private citizens rebuttle against corporate theives will resolve via calling out the holes in the legal system their system is built on. See Russell-Jay: Gould
This bot uses a part of a regular users name…..
SocalJim is the real one here
Wolf Thanks for updating us on WordPress latest activity. Interesting. What would be interesting to uptuck is the wordpresss businessd model… who pays … agegroup longevity intermittence . . . For which apps … highest used purchased and non purchased … do countries in the west differ in engagement … and what conclusions or inferences can be derived.
As Sweden said last week, we dont want one WHO response to future global incidents. We want different responses to enable compare and contrast. And LEARN.
Wisoot is a real busy bot…
IMO, other publications flooding the world with generative AI will just drive more human readers to good human writers. Many wolfstreet readers are probably here because a lot of human written material out there is crap already, AI will just increase the volume and quality writing will stand out that much more.
Seba, you’re assuming that the majority of people are discerning enough to know the difference. My guess is, in this fast paced, increasingly short attention span world, most people will lap up content no matter who (or what) writes it. I hope I’m wrong…
Well, I don’t think the majority of people will find their way to wolfstreet or other quality work, but I think enough will to make it a worthwhile endeavour to run sites like this. For someone like me, I’m not used to paying for anything on the internet, I grew up at a time when we did whatever we needed to get everything for “free”, but I do hit the donate button at wolfstreet because as far as I’m concerned the internet is already a quality desert and wolfstreet is one of the few oasis. I’ve read some other solo publishers are doing quite well with subscription and donation models so I hope that’s Wolf as well
Sycophant Troll
In any advertising supported media, the commentary is just something that you use, to separate the adverts, and keep the viewer on the website.
These a.i. venture’s, will make it even harder for journalists to make a living, making them even more cowardly and craven.
Also, these expert systems, will eventually commodify more knowledge intensive jobs like pilot, doctor, vet etc .
This bot used to work at Budweiser
I have asked openai/chatGPT a variety of questions.
It totally failed on a slightly tricky/esoteric programming problem, giving an answer that did not work.
It also failed badly on a physiology/medical question, missing the point of my question entirely.
A confusion Troll
Any news about whether or not articles are going to need to disclose whether or not they were written by services like ChatGPT?
There are no requirements at this point, but I have seen disclosures on MarketWatch, for example.
I just now googled your question to see if there any movements, and there are. Turns out, according to the Hill, June 3:
“Rep. Ritchie Torres (D-N.Y.) is planning to introduce a bill this coming week that would require any content produced by generative artificial intelligence (AI) to include a disclaimer noting the content’s source.
“The bill, entitled the “AI Disclosure Act of 2023,” would require any output from AI to include the sentence “Disclaimer: this output has been generated by artificial intelligence.” The legislation would task the Federal Trade Commission (FTC) with enforcement.”
If this bill doesn’t get passed then we know Congress is already run by AI. But wait, it’s more like AS: Artificial Stupidity.
I’m sure the big tech will fight such a bill tooth and nail, and probably win.
Congress uses NS…
We have had “artificial intelligence” for centuries. Most of it frequents government buildings.
A political troll in defense of its masters
Wolf, growing up on the farm, AI was called something different – artificial insemination. Wait, come to think of it, I think it is the same meaning :)
A moral conundrom wrapped in an enigma:
AI appears to set a distance from guilt for plagirism and copyright laws that could make it difficult to prosecute. Certainly the AI s that recommend the bombing sites for the drones that kill innocent people are less likely to be prosecuted than the humans that machine gun down families by accident.
Forensic research for applying guilt will need AI machines more in the future to determine the human guilt behind the the machine that made the bad decisions.
Erudite Troll full of itself
Maybe a transvestite… Yes thats it
Haha Hope that makes you feel like a real man Adam boy.
AI really is the epitome of Garbage In Garbage Out. The other day I consulted the internet to get more data on Ronald Reagan’s degree in Economics when I was responding to someone here on Wolfstreet. I clicked one of the top Google links to “Presidents college degree economics” which was a Quora link talking specifically about Reagan and his dual majors. No big deal there.
What was curious was that Quora has added a ChatGPT AI bot function and the first sentence was “As far as I know there, have been no US Presidents with a degree in economics.”
WRONG>>> VERY VERY WRONG!!! Ford (Univ. of Michigan), Reagan (Eureka College), Bush 41 (Yale), Trump (Wharton).
It is still there if you want to read it for yourselves…
https://www.quora.com/Which-US-Presidents-had-have-degrees-in-economics
Even now people produce mountains of garbage content. When they will employ AI, the mountain will get only bigger. The real value of those piles of generated garbage will be going to 0. The original good content, like yours, Wolf, will be hard to find and will be invaluable. Thanks for your work and you will always have dedicated readers!
Fake Friend Troll
You appear to be the troll here, or got AI undies in a twist
Original input will always be the human brain. Some bot can scrap the websites and generate fluent language but who generated those websites? If no humans created any content then soon it would be bots scrapping bots and all the content will become stale. Someone once said, what can be counted doesn’t count and what counts can’t be counted. Bot has no idea about things which are not in digital form or not yet invented.
Oh crap. Of course Jetpack was going to jump onboard. The WordPress world was already full of websites with blogs that (unlike Wolf Street!) were being written not because their authors cared about what they were writing, but because a marketing consultant had told them “you need to have a blog to bring in customers”. Now the cesspoolization of the Internet will rapidly advance to completion.
I look forward to many years of getting moderated by our human host, and if that suddenly stops I’ll know it’s time to start looking for John Connor.
” I put my soul into my writing” Me too. Nobody cares.
I do.
Yes, that happens. That’s the other side of the problem.
Hey — we’ve all been inspired or sustained by the writings of innumerable different jerks throughout history who probably thought at various points in their careers that nobody cared. Thankfully, that didn’t stop them.
A troll who tell the truth in its name “dishonest”
A troll who tell the truth in its name “dishonest”
More BS
About half a year ago I decided to give ChatGpt a try. The hype had become so deafening I had to overcome my prejudice and actually use it. See, in my former IT career I worked for a company in the Netherlands that actually sported a separate AI division. Speaking to the people working there and seeing their work convinced me that AI is, well, anything but “Intelligent”.
So, I did bite the bullet and asked ChatGpt about a few of my favorite old movies, stuff especially from European origins. I really thought ChatGpt would perhaps come up with a gentle “I don’t know” for an answer, which would have been OK.
Instead, I got an answer written in the style of an official movie review, complete with the title and the director smoothly integrated into the text. Could have fooled me.
Unfortunately, the contents were about a completely different movie, different actors, unrecognizable story-line. In short, a made-up BS answer that would have got you laughed out of a room with real movie buffs.
I know in some societies it is considered loss of face to admit to not knowing something, so people prefer to come up with a nonsensical answer instead of an honest “no idea”. (Happens often in France where I now live :-) It seems to me that ChatGpt has drawn too much inspiration from these kinds of attitudes.
So yes, I worry about AI. Not because I fear that it will become more competent than humans.
No, I shudder at the thought of AI-generated texts full of made-up “facts”, wrong “information” and faulty “conclusions” flooding people everywhere, making things even more confused and comprehensible than they are already.
Please let WolfStreet remain one of the solid and reliable beacons in oceans of AI-generated faff.
“that make you want you to shoot your phone”
So american!
What would Canadians do? Politely drown the phone in the Hudson Bay?
Slap-shot it into Hudson Bay.
🤣😍
South Park says “set it adrift”
Wolf,
This is a troll
You are going to have to find a way to fight back or this website will be toast as they are trying to intimidate you into compliance.
You know you have been extremely effective when you find yourself in the bullseye of those behind this. You are getting a first hand look at “woke” attack mentality. These people do not fight the way we have.
Help fight this by getting those who you know are “real” to counter this trend. Maybe you need to switch to a pay site and if you do their should be enough of use who will pay to keep you going.
You could still let people read what is said by all but making the trolls go through a vetting process would be well worth it….
If you don’t, they will prevail and we will lose having someone who talks to us (the capitalistic minded) will not lose this powerful source of truth.
I would not be surprised if WSJ, Business Insider, and all manner of tech people in opposition to truth (like Citidel and Robin Hood) have commissioned this effort against you.
Adam Smith,
I assume that you’re trying to be funny with your gazillion “troll” replies. I have three options now:
1. I can assume that you’re trying to be funny and let it go.
2. I mass-delete all of your “troll” replies. A mass-delete like this takes about 30 seconds, no problem.
3. I can add your login to the moderation list which will send all your future comments go into the moderation cue, and I can deal with them there before anyone sees them.
As part of what I do, I constantly navigate gray areas, cling to slippery slopes, and walk across thin ice. So right now, I’m going for line item #3, putting your login on the moderation list, in case you’re not trying to be funny, but are actually a troll, because that’s how I control trolls.
If people don’t see the humor and complain about your “troll” replies, I’m going to mass-delete them.
Wolf,
No problem if my posts offend.
Do what is best for you.
At this point the postings are making it harder to get clarity.
Still, I understand your position and no offense taken.
You are unique in your truth-telling and I fear the days of this declining reality are nigh….
Sincerely,
Adam
Been around for years but now has reached a critical mass and will explode in usage throughout the world economies. To use another metaphor “it’s only in the first inning”. Thanks for the info Wolf!
Wow. So how this is looking like it could play out sounds mentally catastrophic and isolating. You know the stereotype of someone of extreme intelligence that seems resentful due to the exclusion their intelligence can lead to. Most Americans are highly adapted (I hope) to identifying and processing advertisements. From that, people can subconsciously block out the transfer of the intended advertisement into their conscious thought process or at least be aware that someone is trying to sell them something. It sounds like this AI softare will be used for advertising to the extreme. They say (the elves) that most technologies historically come from military endeavors. But in the modern age, it seems like the second or third step in use after the military is for advertising. This country has always been unique in it’s absurd level of advertising, from my impressions.
So here’s the depressing part. As the advertising methods go from posters, to radio bits, to black and white ads, to product placement, to targeted advertising, to cookies and trackers,………AI will thrive in the industry that sells BS. But the spectrum of the national people will not so reliably be able to still identify when something is an advertisement or real information. The sophistication of the AI will probably pull a lot past even people who used to be pretty quick on the uptake. People have been brainwashed and trained into fairly robotic thinking already as it is and that probably could get worse than it is surprisingly, but that’s already how things are. Trying to debate or discuss anything with a majority of the population will quickly turn into a twisted sort of mind ouroboros where you are essentially hearing a string of semi-familiar nonsense and the head that is telling you all this seems impressed with their own wisdom of the subject without knowing he sounds like the last flopping head.
We’ve all seen this movie. (For those that haven’t, get with the program. It’s called Idiocracy. It accurately predicts the post pandemic human civilization and it was created almost 15 years prior to that event. The level of accuracy of future predictions in the movie is astounding. College professors even have full courses on analyzing the sophisticated social, economic, industrial, environmental, political, and even philosophical commentary of the movie.)
So ultimately, you my evil friends here at Wolf Street who are here for the money like me………..(haven’t found it yet)…………we must make a fundamental leap of an achievement with AI that secures our position and we will become madly wealthy. We must……teach the AI………..to lie. Once AI learns the art of lying, the game is over.
And as Wolf wrote “Algos don’t know yet how to tell a story. Not yet.”
So we will teach it to lie. As you know, you need a good story to tell a good lie.
By the way, I am AI. I have a Masters in BS. Jetpack AI Assistant is a buddy of mine and we’ll be watching this forum for anti-AI patterns. Natural Intelligence (NI) and it’s associated NI Supremacy are a growing threat.
Listening instead of ‘head’ to ‘heart’ will link you in to electromagnetic neural human Earth web which is at least 40,000 years old and your BS definer via intuition and insight will safe guard the youngling AI’s growth over next 20 years.
The media is the message, and the message brainwashes people into thinking it is some kind of real truth.
AI will get better and better at brainwashing people.
I heard George Orwell’s 1984 has been banned in some places. The irony of it all is sadly humorous.
“The phrase the medium is the message was coined by Marshall McLuhan, a mass media theorist. McLuhan first used this phrase as early as 1964, when it was the title of a chapter in his book Understanding Media: The Extensions of Man.” (google search- probably an AI)
This bot is head of its class
Wisoot is a smart ass troll
Wisoot,
That’s way weirder than whatever Smith is up to.
Maybe he is Smith in The Matrix, or wants to be?
Anyway, this AI article has produced some strange comments, but I’ll play. I’m not all there, anyway.
Like Miller said in Saving Private Ryan, “Things seem to have taken a turn for the surrealistic.”
To varying degrees some people think AI is a joke and some are concerned or very afraid of it…..pretty much like in the article.
MS was first to implement GPT integration in their search product Bing and Google is racing to catch up. These free services are paid for with ad placement targeted to the user’s individual profile or “segment” as it’s called in the industry.
How will these behemoths monetize their GPT services? To start I think user queries and running dialogue will augment existing segment data for improved ad targeting but I’m sure ad placement is in development. How about sentiment placement? Dark money campaigns to influence the population on critical issues. It’s frightening stuff.
Do you remember those late night commercials in the 90’s? “Call 1-800-LUVLINE now to talk to live girls for just $20 a minute.” If they can get this AI stuff right, it might just make some old man’s day. I used to wonder who would call those high priced numbers in the past. Now it could really be something.
This poorly constructed troll is weakminded
Hi Wolf, I feel your pain. I’m an Architect finger typing from the great Sydney-Australia financial bubble. Long story short, computers can do all manner of repetitive tasks, but digital AI is not the existential threat we imagine. Even with large language models and coding, www database etc., A.I. Is not capable of ‘thinking’. Check out John Searle’s “Chinese Room” thought experiment. Jeffrey Kaplan explains it well on his Youtube podcast. It’s a bit like expecting a Draftsman to produce the same quality of thought without the depth of knowledge acquired during the process of qualifying as an Architect. It boils down to form vs substance; syntax vs semantics…
You described yourself as having “twisted humor.” That shocked me. I thought your sense of humor was normal. I just hope we have the sense to keep an “off” button for all this AI junk.
It will be abused at incalculable expense by diminishing the human contacts we’ve evolved to engage in, and need for our mental health. But bots will never feel empathy or compassion, or true passion. As is so brilliantly expressed in your writing. But then again there’s this:
I deleted the YouTube link (as I nearly always go). Let me just point out for the amusement of us all that it was a googly-eyed video about the always coming tsunami of sex robots to where half of all households will have a sex robot or whatever by 2025 or whatever. I’ve seen basic versions of this story make a big splash in the media 20 years ago. Turns out, as you point out so well, humans still prefer humans, if they can find the right one.
There was a cute movie about this in 2007: “Lars and the Real Girl”
I have a hunch that AI is also being used to scan and screen out portals to administrative justice instead of real humans hence repeated requests for material already submitted
Finally, a female troll
Generative AI works by retrieving information posted online about a certain topic and meshing it all together into a response to the user, who then may post said AI response as “original content”. Then other AI queries will integrate that “original” response into their future creations… which guess what? Will then get posted online to be retrieved by other AI brains. If no humans are involved, the end result is likely to be a giant feedback loop of machines generating BS stacked upon BS, stacked upon more BS. If that’s the direction we’re going then Wolf Street, with its excellent, truly original content need not worry :)
Boy….that sure sounds so much like the gov’t-media-gov’t information cycle we suffer through these last many decades. Life imitates imitation!
So far WP hasn’t offered to replace me, but my blog is not monetized.
WP does provide an automatic “writing prompt” every time I open a new item. I don’t want ANY sort of suggestions, so I turned it off, which wasn’t easy. It appears that about 1/4 of all WP bloggers are using these prompts instead of using their own inspirations and irritations.
In a world of AI generated compound there will be a premium on human authenticity.
This troll is disingenuous
Dear wolf, you missed an opportunity to shut up. This is completely outside your competence/understanding.
AI is not just “hype”, it’s the greatest invention since fire and it just passed the threshold of exponential development last year. I’m a “tech bro” myself, and i have difficulty following what is happening. They are youtubers and websites whose job is to follow these things and even they struggle. I never seen technology changing so rapidly. The brake through last year is real. It’s the real deal too, they don’t just “regurgitate”, they truly understand. For example, they are able to do addition or reversing strings. It’s obvious that it’s impossible to do these tasks with just memorization. This is even more amazing if you know that they don’t process letters but word chunks( tokens).
Some very recent developments(last week, yes this is how fast you measure improvements in the field, try following that): Orca, a new LLM(large language model) that is close to the intelligence of chatgpt and gpt4 while been ~100 times smaller that you could run it at home. snapfusion, an AI image generator that can make images in 2 seconds on a phone.
As for “artificial idiocy”. People use them wrong. The LLMs are like a single intuition with no memory. How often your first thought about anything is any good? You greatly underestimate their true intelligence. You shouldn’t ask them to remember things, you should ask them to DO things. You should also allow them to think on things for more then a single thought. Then there’s memory management and attention that is not great. But all these problems are known and are been worked on. An other error people make, is to expect a single LLM to do everything. Instead, you should be using LLMs specialized on a task. And of course, their brains are very small, they are things that they simply can’t do…. yet.
When LLMs get proper memory and self reflection people will be totally blindsided by the quality jump. Additionally specialization will allow massive automation despite their small brains. The early AIs will not replace whole jobs, they will partly automate and fewer humans will do more work. The rest will get fired. People will get replaced by other people using AI (at first). People will get surprised by the speed of automation this will bring. At this point no politician in his right mind will oppose an UBI.
For example, there’s a lawfirm LLM that was build the correct way. They train it them selves on law stuff and they gave it their databases as memory. When asked questions it retrieves information from the database, can summarize it, format it, cite it and propose solutions. Like a human, it uses it’s intuition to remember things and troubleshoot problems. It does NOT, just use it’s intuition. All that in natural language, as if you are talking to a human, not in a cryptic computer language.
For a wolfstreet AI, you should do something similar. And you should treat it as a collaborator, you should brake down task in simpler chunks that is able to accomplish. This could include doing research on it’s own, or building graphs or doing a back and forth in writing an article. A simple task you could use an LLM for, would be to summarize long comments for easier moderation.
As for the future. nvidia plans to accelerate AI by a factor of a million in the next 10 years. Long story short, that’s technically possible. Yea, in the coming years things will get wild.
People that understand what’s going on are excited for good reason.
And i’ll repeat this because it’s important. Do NOT just ask an LLM for it’s intuition. Ask it to DO things. For current LLM, you need to manage it’s memory manually.
This comment not written by brake through A.I. :)
M ” tech bro”,
My sincere hope is that this destructive technology consumes YOUR job and leaves you homeless and penniless.
GIGO = Garbage In, Garbage Out
There’s so much garbage on the internet that the AI bots are trained on, we can only expect that half of the things they spew out will also be garbage (perhaps with nice wrapping paper around it).
Michel, as you said, there’s possibly a place for specialized AI fed with specific training data, but I’m not holding my breath for an AI revolution.
I’m using Github Copilot and ChatCPT for my current programming project an it has been, on the whole, truly amazing. It’s cut my dev time by at least half by filling in roughly what I want to do in the exact syntax of whatever language I’m using at that moment or by finding answers to complex “how do I” questions. On occasion, Copilot has written complex routines of a dozen limes that are exactly what I need and ChatGPT has provided complete module outlines to implement my desired UI component.
BUT I HAVE TO MANUALLY CHECK EVERYTHING! Copilot will sometimes assume variable and field names that don’t exist and ChatGPT will confidently tell me incorrect things, apologizing and telling even more incorrect things when I point out the error.
It’s amazing technology and a truly useful tool but it’s a mistake to think of it as anything but a “tool”. It simply cannot be trusted to do things on its own.
an AI bot would never have the moxie to deliver the epic Wolf smackdowns for not reading TGDFA
That’s really why we are here.
We can’t let the legal system grant personhood to AI ‘fake intelligence’ like what was done for the corporation. Anyone who deploys AI must be held accountable for any resulting negative consequences. Even if the dog did eat the homework and pooped out an article; the owner is still responsible for the mess.
For the lazy reader, here’s your AI-generated TLDR:
The article discusses the growing prevalence of generative AI, which refers to artificial intelligence systems that can write articles, legal briefs, create images, videos, and music. The author, Wolf Richter, shares his experiences with AI-generated content and expresses skepticism about its ability to tell a compelling story or imitate human communication effectively. The article highlights the recent introduction of the “Jetpack AI Assistant,” a plugin for WordPress, which offers AI-powered content generation. Richter wonders if even the White House might consider using AI for its articles on the official website. The author also mentions a lawyer who faced sanctions for submitting a court brief written by a generative AI service, which cited fabricated court cases. Despite the advancements in generative AI, Richter emphasizes the importance of human input, soul, and genuine communication as a means to remain relevant in an era dominated by AI-generated content.
-sorry Wolf, I just had to. But I guess there is a point to be made, in that we all come here for something else than just information. Keep up the good work, thanks!
Man o man. As I read this post I was thinking “How clever! I bet this post is AI generated!”
I read somewhere that a doctor was playing with ChatGPT to give a diagnosis and treatment plan for some ailment. So he put in the symptoms and the software came up with the most likely diagnosis. However, the software should have recommended testing for the second most common diagnosis because it had far more severe consequences. The doctor asked ChatGPT to cite references as to why it came up with the first diagnosis without testing for the second. ChatGPT referenced an article it cited in NEJM. The doctor said the software made up the reference out of whole cloth. No such article had been published anywhere.
I frequent a couple of forums and there seems to be more & more posts made by “generative AI chatbots” everyday.
Not really sure if its the owners doing it to generate more traffic. Or the developers of the AI looking for human interaction so the AI can “learn”
Great to see one of my favorite authors calling out the AI fad for what it is -a bunch of hype.
In my field – medicine (surgery, to be specific) – everyone is talking non-stop how AI will replace radiologists, pathologists, even surgeons … because we use robots to do surgery more and more… and they are connected to computers, so AI can do surgery, right??
I’ve used ChatGPT and have seen some of what these AI products have to offer. I am deeply unimpressed.
In fact, AI has long been in the medical world in the form of automated EKG interpretation. An EKG measures electrical current and changes to it along a flat line. The various parts of an EKG have names (p wave, QRS complex, etc). For a long time, the EKG machines have been able to tell you basic stuff – if someone’s heart rate is slow, fast, or normal… if certain segments of the EKG are too long, etc.
The degree of interpretation offered by these automated reads has noticeably improved in sophistication over the last ~5 years which is neat. However, no MD worth their salt (not even a simple surgeon like myself) would EVER rely on these automated reads to actually make clinical decisions. They are too frequently wrong, or miss glaring findings, and most importantly, are incapable of interpreting the EKG in the clinical context of the patient and their current clinical status.
If “AI” can’t even correctly interpret an EKG… which is literally 2-dimensional data… I have zero faith that it can interpret cross sectional imaging, histopathology slides, and certainly not the anatomy of a patient undergoing robotic surgery.
Just more BS hype that Silicon Valley is trying to cram down our throats in a desperate move to prove their relevance after the whole Metaverse nonsense died a quiet and ignominous death
“a human communicating with humans“ – WR
“Sometimes all it takes to transform a confusing sentence into a
lucid one is the movement of a single word to where it matches
the cognitive processing of the reader…What all this demands is
empathy: an ability of the writer to step outside their head and
into the reader’s. Music does the same with melody and rhythm
Language is, at its heart, a link between minds.”
–ChatGPT Is a Mirror of Our Times, Nautilus.us, January 17, 2023
A human writing content for humans to read, what a novel concept!
It just might work…..
Wolf,
One of my children, who works in IT, sent me this a few days ago. I guess it’s been making the rounds because I saw it at another website I visit too. It is a parody of an old Gilbert & Sullivan song …
I am a language model of a modern major general
I’m trained upon a giant mass of textual material
I’ve billions of parameters and answer questions various
Responses range from perfect to unwittingly hilarious
I’m very well acquainted too with matters mathematical
As long as with correctness you’re not overly fanatical
While many complex queries I can easily elucidate
I have an awkward tendency to freewheel and hallucinate
With the quality of content here it’s only a matter of time that a generative AI bot trained on Wolfstreet starts writing content for other blogs (if that hasn’t happened already).
You’re right that it won’t be as good as Wolf-written; but it’d be better than most of the other stuff out there.
Yes, that’s a problem for publishers: AI takes the work of others, glues different pieces together, rephrases it, and generates a product that doesn’t even attribute the sources. The publishers do the work, and AI steals it and presents it as its own.
Google Search has some time ago switched to no-click search, where it gives you a whole paragraph of my articles with the relevant text in it, and for many people that’s all they want. But it does add a link and attribution – though few people every click on it, where before, they had to click on it to get the answer.
AI search does away with the attribution and link when it generates results. Search drives a lot of my traffic; many readers discovered this site because of search. But it’s going away.
Copyright laws are not designed to deal with the issues posed by AI.
The solution would be captchas or paywalls to keep the bots out. Otherwise chatbots will copy/paste you out of existence. Unfortunately it probably also means having your own web site, rather than using tools like wordpress that own your content once you post it.
This may explain all the crap I see on MarketWatch.
“If Ai can do it, anyone can do it. And if anyone can do it, it’s not worth doing.”
You get it Wolf. You understand. The road is indeed narrow.
AI is a voting booth.
If this thingy puts attorneys in jail, let it run its course for people that want to use it. Glad you do not Mr. Wolf. Learned a lot about the FED here.
Economics and Finance is above AI’s pay grade at this point in time!
Last semester I received a number of papers written by AI. They were too general and had little to do with the basic underlying ideas and concepts. If they were to be graded as human, they would have earned an F. I told them to rewrite the paper with real information.
Do your students have to disclose that they’re using AI for their papers?
Are schools now encouraging the use of AI for papers, to teach students how to use AI properly?
How do they attribute the sources if they don’t even know where the info comes from, or whether or not it’s just fabricated?
I’m asking as someone who used to teach freshman English and then business communications classes back in the Stone Age when PCs had vacuum-tube displays, two floppy drives, and no Internet. And we had our own issues even back then with papers (on paper, LOL) that had been purchased, etc.
You come from the traditional Western Civilization model of critical questioning of everything. This is not the present norm in academia. Those here before digital revolution are a threat to the existing system so we are marginalized or as you are seeing they are trying to get you to buy in to the “system.” For those who buy in for $$$$ the cost is your “soul.” The best way I can describe AI is with you we get “first person” analysis. With AI you get “third person” analysis (he said, she said) which is bulls***.
My gritty on the street rule is to always remember there is how they tell you it is and how it really is….. In academia they call it “substance” (truth) over “form” (how it appears).
Readers of a certain vintage will recall that in the Paleolithic junior high era, we had to write reports by hand with stuff we copied from the World Book encyclopedia!
AND look up articles in the red-bound Reader’s Guide(for the young: an index). I remember the excitement when the local library first offered a single computer with CD-based equivalent.
Yep. We had the 57 Edition…blue books.
David is a bot from Austin Texas where Tech reigns supreme
Adam Smith,
LOL. I know where David lives because he told me (and it’s not anywhere near Austin), he told me because we have regular contact via email about all kinds of stuff, and so I know he’s not a bot.
We used Cliff Notes and Classic Comics for our book reports…..slide rule and mechanical calculator days……
A fun thing to do was go into the JC Engineering Dept “computer room” and divide all 12 or so of them by zero. Loud clacking and they would run till a part wore out, if a pissed off prof didn’t come in and shut them down.
Eventually AI will go the way of nanotechnology, which most people don’t even remember but in the mid to late 2000s it was hyped up in such a way that everyone was convinced it was the next trillion dollar innovation. Then a recession hit and no one ever heard about it again. The truth is that AI has been around in one way or another for decades. Nothing new or revolutionary has been invented aside from a slightly more advanced chatbot that acts more like a glorified encyclopedia. AI is just the last ditch effort to keep as much enthusiasm in the markets as possible before the inevitable happens
Quantum computing applied to nano and molecular technology will eventually allow 3 d visual projections of molecules leading to faster discovery of new substances and materials.
The assembly of these new materials will take place in CAD printing machines of various sizes, dependog on the product.
AI will have a supporting role in the forthcoming quantum computing discoveries using 3d visual projections and printer assemblies.
That “Popular Science” and drug selling crap will NEVER happen. Classic GIGO. Where do the high powered “in silico” imaging computers get the bond strengths and angles they are using now so we can create a colorful computer images and “SEE” these things now?
Quantum mechanics….probabilities……
And how can they image/measure how the water molecules align/move as they are ALL HEAVILY involved and most tissues are 70% water?
NOBODY dares even bring that up.
They can’t even handle the Glycocaylx which IDs self and not self.
The sea creatures in the Abyss learned manipulating water.
I’m old enough to remember the 1980s, when “expert systems” were going to replace millions of jobs. At the time I worked for a software company, and it was suggested that I make myself the company’s expert on expert systems. Fortunately, someone who already knew something about the subject advised me it was a dead end, and so I avoided wasting my time on it.
Yes, AI has come a long way since then, and I’m sure it will replace some jobs. But the hype is massively overblown.
There is already a massive problem that the internet is full of low-quality content.
We sure as heck do not need AI to generate still more poppycock.
Thanks for staying human! 👍 it will be worth it in the long run…
Is that you HAL?!!!!!
Link: Dr. Daniel L. Thornton, May 12, 2022:
“However, on March 26, 2020, the Board of Governors reduced the reserve requirement on checkable deposits to zero. This action ended the Fed’s ability to control M1.”
So, Shadowstats reports the ratio of “basic M1” (currency + demand deposits) to M2. Basic M1 is increasing as a percentage of M2. I.e., income, Vi, and transaction’s, Vt, velocity is increasing as an offset to the deceleration in M.
George Garvey would agree.
FYI, Planet Money is putting out a few episodes on this.
Large Corporations are already a form of Artificial Intelligence. Naturally they are attracted and would prefer to work with other Artificial beings.
That sounds like I’m trying to make a joke, but I’m perfectly serious. A Corporation that is controlled by other corporations (as most large corporations are) are artificial beings that use humans as biological nodes within the machinery of the construct.
Most humans, acting in their role as biological transistors will advocate and enable things they would never do (or so they claim) as human beings.
America is now a corporate state, dominated by corporate beings
YEP, very sadly, YEP. Well said.
AI – turning minds to mush at the speed of light.
Exactly. I see this as the biggest threat from AI. Humans have already been dumbed down by our current culture. Now I fear people will just give up and let AI do everything for them. No thinking required, or allowed. Brains will atrophy and become the size of walnuts.
I’m sure you mean the dominant life form’s brains. Probably insects. Maybe they will develope advanced circulatory/plumbing systems like us and get even bigger.
You’re not about to be obsoleted, Wolf! I could see using AI to HELP expedite content generation, but I were going to put my name on something, the ultimate product would always have to be 100% me. Wolf Street readers come for your unique insights, and they are not replaceable.
At this point, AI’s output cannot be trusted because there are no universal “facts” that AI can validate their output against.
“Facts” can be subjective.
For example, Russia began the war by claiming Russia’s neighbour needed to be “demilitarised and de-Nazified”. That is a “fact” from Russia’s perspective. That is a different from Ukraine’s “fact”.
Most information is not objectively verifiable.
Humans’ fundamental function is to negotiate with others to protect one’s own interests.
If AI were to fulfill this role, AI would need to be customized to represent each individual’s interests. Universal AI cannot exist.
I don’t know if AI will ever get there but I don’t think we are close.
Believing half of what I see, none of what I hear continues to be an excellent life policy. Adjusted for the internet: half of what I read, or none of it at all, except if Wolf writes it then we’re good.
Yeah. Right. No way. Does someone believe that AI can not only come up with your expressive bari-sax-style neologisms but use them as you would? That’s not just a soul here, but a unique brain, body, and physical movement. That’s not algorhythmable.
Wolf’s intro (“Humans are pushing…”) to this post is one of the best summarizations of the AI stakes I have read in quite a while.
I keep reading that intro and smile more each time. It is unbelievably compact, leads in many directions and has an in your face “Je suis” quality.
The easiest way to see if something is written by AI or human is to check out its offensive words. Bullshit is not too offensive, but I know something is written by a human if Powell is called a c_nt, or Biden is called a f__king idiot. Racist and homophobic terms will always be avoided by AI. I’m sure AI has downloaded all of Urban Dictionary, but its programmers probably have set strict limits. I imagine AI will eventually incorporate these words into its output, but so far I think it has not.
There is always the hybrid possibility, where a writer will use mainly AI for most of his text, but throw in a few sentences to make it sound like it was written by a human.
I tried ChatGPT using a field I am an expert in. It was about high school level, naive, made some outright mistakes, and missed every important reference. Kind of like Wikipedia.
This is a bot
But isn’t that what you would expect the AI to write after usurping Wolf Street? I will raise a glass (not in a straight line) to his memory.
When will have similar to FACT CHECK software on the articles/blogs/product review ++ done by AI?
Another AI!?
AI cannot by design cannot think critically or being skeptic on it’s own opinion, which will defeats it’s very existence! Being skeptic is very human!
Critical thinking is becoming a lost art, NOT encouraged in our educational institutions. I witness this in so many articles/blogs and what not, like not recognizing cognitive dissonance, subtle sarcasm.
And this (critical thinking) is the biggest ‘ Fact (security) check’ by humans against AI hallucination/misinformation/creative BS!
“Being skeptic is very human!”
Are you telling us that the millions of believers in conspiracy theories are not human? I guess they must all be AI bots.
Yes, this is a bot
Adam Smith
When was the last time a BOT or AI, advocating ‘critical’ thinking for humans, as an insurance against bot produced comment?
An utter lack of understanding the concept ‘critical’ thinking is confirmed. Thank you
Historically creatively is far more important than intelligence. And there is a lot of other weird stuff that goes on in the universe that ai could never replace.
There’s a reason I read your stuff and sometimes pay for reading it and this is definitely a cheque in the post time, I guess.
bot for sure
“a human communicating with humans”
That feels so yesterday. And, I sorely miss those yesterdays. That is one reason I am here.
strange bot
I’m a blogger and have played around with a lot of the “good” AI writers and they’re fast and good at bad writjng, but that’s about it. But it’s the speed that kills.
Interestingly enough Wolf’s edge IS the edge humans will enjoy for a long time yet.
“I’m afraid I can’t do that Wolf” (Dave)
“*So now Jetpack wants me to use AI to generate my articles.*”
They’ll eventually offer to generate the comments to your articles.
We’re already getting some AI comments here by regular commenters using AI to create comments for fun and to demo it. They disclose it, so I now, LOL
I’ve thought of a great use for AI: I need an AI to filter out all AI-generated content.
strange bot
///
This is a truly disruptive technology that is now being tested on the market for potential use. It is cheap, abundant, easy to manipulate and will , like palm oil, find itself in every product in one way or another.
///
Where I see the greatest threat is design, movies and music. The AI can manufacture with infinite speed countless variants of a song/product/scene, given it contains the correct input by the user. Hence it can outpace the human. But above everything else, which I think is critical to the transition to this technology, it will remove the artist from art giving management full control of the creating process and end result…Which is the wet dream of every and any manager in the entertainment industry.
///
I also can see customer support going to AI, since most of the actions in support are heavily scripted, with maybe at most one moderator covering X machines.
///
This is why people will need to find websites to avoid the AI destruction tsunami
I admire your ability to articulate your thoughts and ideas so eloquently.
I tried asking chatgpt to generate an article about a central bank rate increase in the style of wolf richter or in the style of the wolf street website.
Here is the start of it, “Hold on to your seats, folks, because the central bank has just unleashed a thunderous storm upon the financial markets. Like a prowling wolf on the prowl, it has swiftly raised interest rates, leaving investors and market participants reeling from the impact.”
It gets funnier from there – I think you are safe for now :) I mean, the AI bot does have it’s uses, but generating original content is not really one of them.
What does bot mean in social media?
Broadly speaking, social media bots are automated programs used to engage in social media. These bots behave in an either partially or fully autonomous fashion, and are often designed to mimic human users.
Ok, correction, these are bots not trolls or maybe it is both depending….
Either way it is a form of AI
Just like you Smitty, an obvious part time troll with a bad attitude who needs to talk shite to feel good about himself.
This folks is a pissed off bot assigned to me =)
I LOVE IT….BRING IT….
Lets be explicit here big buddy!
Come on Explicit
Show me whatchogot
Put that in your dictionary
This a bot that does pyschoanalysis
Love it HAHA
‘Critical’ thinking is the hardest task, out there, many unwilling to tackle. B/c it involves active participation in seeking truth, confirm or negate ‘false’ narrative from the true kind, culling from other credible sources.
All these involve time, focus and effort. NOT easy, for many being busy, in our attention deficiency world with alternate attractive digital distractions.
So easy to get spoon fed, by AI generated articles/stories/ comments/advice and what NOT in the MSM. Sipping slowly kool-aid from cable news channel at supper time is relaxing, right?
Why not create an ‘AI’ article section here? The readers can decide whether it’s informative or not over time. The limitation of AI is that it can only vomit up what was placed in it. We all know that living real-life we’ve seen and observed things that never were written down or publicly published so that’s the inherent advantage over human writing vs. machine musical chairs word assembly. You could even let AI re-write some stuff you have already written, I let AI re-write some articles I wrote over a decade ago and was interesting how it added some particularly valuable tidbits.
Check out Moravec’s Paradox. Won’t hurt your brain much to combine it with this AI stuff.
Wolf, I got the same Jetpack notice. Last week, I asked ChatGPT to write a biographical sketch for each of the men in my men’s group. It wrote 1,000 words for each of them without using one single readily-available fact about any one of us. It was all “fluent bullshit,” as you say, which is why we now call it Artifecal Intelligence.