The Other Side of the Coin: The AI Bubble Pop

After my last post launched, I was pleasantly surprised to read about some problems on the horizon for OpenAI.

That, alongside Andrew’s comment on the post (thanks for reading, Andrew!), compel me to explain that though I personally believe AI has some value, and though I believe we will eventually be ousted by attrition barring definitive action on our part (like me raising about $3 million to pour on a school and some media action! Dreamers can dream, can’t they?) — I still acknowledge that the colossal amount of money being spent on AI is likely to result in a big bubble burst and potentially cause another AI Winter.

And though I don’t intend to make a habit of using AI generated stuff on my website, I once again let AI do some googling for me just to see what it might say.

Is there an AI bubble in terms of investing and can it lead to an AI winter?

I’ll search for current information about AI investment trends and concerns about a potential bubble or AI winter. Now let me search for information specifically about AI winter risks: Based on the search results, here’s what the current situation looks like:

Yes, there are significant bubble concerns

AI investment as a share of the economy is nearly a third greater than internet-related investments during the dotcom bubble, and many experts are warning about bubble-like conditions. Even OpenAI CEO Sam Altman has acknowledged that investors as a whole are “overexcited about AI,” though he maintains AI remains fundamentally important.

Key warning signs include:

Over 1,300 AI startups now have valuations exceeding $100 million, with 498 “unicorns” valued at $1 billion or more
Tech companies are projected to spend about $400 billion this year on AI infrastructure
Circular revenue deals between major players like OpenAI, Nvidia, and Microsoft, where these companies are simultaneously investors in and customers of each other
OpenAI is losing billions annually while making massive investment commitments, with projected revenues of only $13 billion against deals worth around $1 trillion

Could it lead to an AI winter?

The picture is more nuanced here. The IMF’s chief economist predicts that while a bust is possible, it would be less likely to be a systemic event that would crater the US or global economy, since AI investment is financed by cash-rich tech companies rather than debt.

However, an MIT report revealed that 95% of enterprise generative AI programs are failing or returning no measurable value, which suggests the practical applications aren’t yet living up to the hype.

The key difference from past AI winters: Unlike previous AI winters, companies like Google, Amazon, and Microsoft that survived the dotcom crash are now the primary AI investors, and they’re already profitable with strong fundamentals. This suggests any correction might be less catastrophic than historical AI winters.

While some companies will likely collapse under unsustainable economics, AI itself won’t disappear—it’s evolving toward more practical, specialized systems rather than mega-models. So we might see a market correction and consolidation rather than a complete “winter” where AI research grinds to a halt.

Back to Christopher Day, the author of most posts on Stenonymous.com

In short, it’s not the technology that’s going to kill us. More likely than not it’s the shortage exaggerated and exacerbated by the Speech-to-Text Institute and the large court reporting companies. For example, when U.S. Legal used its money to buy and kill Stenotrain and proceeded to bet everything on digital, or when Veritext built Blue Ledge into its pipeline instead of, say, using its money and power to create a stenographic program and build that into its pipeline. More or less, they saw an opportunity to bump out the “overpaid” stenographers, and they took it. And that’s what I spent the last few years documenting.

With some words on the internet and $10,000 from readers like you, we dismantled the fraudulent STTI, publishing again and again that they were using outdated, unadjusted figures to make their claims. We ended up informing thousands of potential jobseekers that digital was not the best way to get into court reporting. To this day I get emails and comments about various fraudsters in our industry. I’m one guy. There are 17,000 to 30,000 of us. That’s a whole lot of potential and brainpower.

So, I still hold out some hope we’ll do something special and reverse the course. My favorite Latin phrase is, after all, a posse ad esse, “from possibility to actuality.“

P.S.

Like Ms. Rachel says, bubble, bubble, bubble, bubble, pop!

Ah! Dad jokes! Guess how many times I’ve heard that one!

Every movement needs its propagandists, right?

AI Will Revolutionize Court Reporting, You’ll See!*

*This is a creative writing exercise meant to make the reader think. While it will talk about real world things, it does so in a way that is meant to be funny-ish, dark humor. One of my readers mentioned these joke posts can be unpleasant because the reveal comes at the end. We’re going to try doing it at the beginning now.**

AI has been used as a buzzword to milk investors for years now. It’s only natural that the tiny court reporting industry follow suit. After all, AI has never killed anyone or gotten them raped in jail. What could go wrong when we’re dealing with the accuracy of court records and the strength of meritorious appeals? After all, all trials result in a perfect verdict all the time. The system never goofs.

Look at how AI improved medical transcription. It made sure your doctor’s eyes are on that transcription during your appointment. What? You want them to examine YOU? Quality care is not part of the Leonard Green model of looting hospitals for poor people. Similarly, Leonard Green asset, Veritext, continues its push to lower quality wherever profitable.

From 95% in 2016 to 25% in 2020, technology is exponential, so there’s nothing you can do, just give us your jobs. Actually, no, just come work for us for less. Digital reporting is the future. You can just take less money to clean up our ASR. And that’s progress because then your money is in our wallet.

Look, even if you think there may be some quality issues with the AI stuff, it’s inevitable. These folks have more money than us. There’s definitely no world where court reporters make so much noise about digital reporting that the cost of advertising for digitals outweighs the cost of not being a scheming dickhead. No, that could never happen.

Give up and do nothing. That is the way of the “virtuous person.”

Si vis pacem, para bellum.

**So that whole private equity thing where the companies buy, hold, and sell other companies, it’s a cool way to make money. But as it turns out, my blog has been increasing investor knowledge too. Big box might not be able to sucker another firm into buying their company. This might become one of so many leads that turns into a big nothing burger, but stay tuned.

Addendum:

Following this post, a valued reader sent me this, showing the Florida Bar seems to be aware of AI pitfalls, and this, indicating that the FTC is looking at the potential power of the “Big AI” cartel.

Another valued reader pointed out the Taylor Swift AI controversy.

Stenonymous Tries ChatGPT

I got to sit down with ChatGPT, a program by Open AI that takes what you give it and gives you back what you’re asking for. The way it comes across to me is a “smart search,” essentially taking what you’re asking for and providing a simple, “humanized” output explaining something or answering you.

What you’re about to see are a series of things I asked it. I had some serious questions, some vanity questions, and even a conspiracy theory joke question in there. I even ask it what can be done to combat worker shortage. Please note that the co-founder of Open AI, Sam Altman, is quoted by VICE as having tweeted that it would be a mistake to be relying on it for anything important right now. Enjoy.

The first message I got when I went to try ChatGPT. The creators are upfront that it can produce misleading information.

The ChatGPT system gives an honest look at its capabilities and limitations.

Of course, my first questions for ChatGPT were the vanity questions.

Stenonymous defeats AI for working people everywhere (joke).

Christopher Day asks ChatGPT a joke conspiracy question.

More vanity questions by Christopher Day to ChatGPT. But ChatGPT mentions AAERT and shuts that down fast.

ChatGPT has as much trouble with this question as Christopher Day, part 1.

ChatGPT has as much trouble with this question as Christopher Day, part 2.

ChatGPT defends AI by refusing to answer (joke).

To be fair, while most question answers impressed me, this one stuck out as something where a traditional search is actually more helpful.

Spotlight on Trade Associations by the FTC

That said, even ChatGPT knows what the gold standard is.

You can also give it information. And while I work on another article to highlight CoverCrow, I figured I’d let ChatGPT in on it.

Notably, this doesn’t always work.

Addendum:

Why Active Readback’s No Steno Man is Wrong

Some have seen this video. I got around to it. I have honest reservations about giving someone like him more press and attention, but then, my audience outnumbers him by a lot, so if you all have the data, it’s a force multiplier and family he doesn’t have. You can tell Readback is terrified of us because they don’t have the guts to leave the comments on and get called out on their lies. Let’s take advantage of their fear.

He likens court reporting to medical transcription. I made a short TikTok on that. I’ve spoken to Mitch Li from Take Medicine Back. Emergency room physicians are being pushed out for nurse practitioners in the same way big money is trying to push us out for digital. Guess what? The doctors largely don’t like that their scribes were pushed out, and the quality of medical transcription has been suffering because of its lean to automation. As a matter of fact, as a young reporter, I was getting requests to get involved with medical transcription (MT). That was only ten years ago. Nowadays the Association for Health Documentation Integrity says there’s a transcriber shortage.

Wow, sounds like the exact excuse they’ll use in court reporting.

It’s so bad that they STILL want to attract court reporters to do medical transcription. So how good was automation for MT anyway?

“Automation is so wonderful, please God send the court reporters to save us.”

Well, isn’t it interesting that jerks like no-steno man (NSM) created problems in an industry that they didn’t bother to stick around and solve? “Oh, people are dying from the reckless automation of something important? Exit stage left. Time to try court reporting!” Guess what? We’re not medical transcriptionists, and we’re not letting you destroy our industry without a fight, you jackalope.

Ultimately, I decided to eviscerate your puffery with cold, hard, facts.

His entire line about automating medical transcription and making it cheaper is fluff. What good is cheap, useless, garbage? And make no mistake that automatic speech recognition, natural language processing, artificial intelligence, or whatever fancy label we want to put on it, is just that. The objective science that exists today says that it’s 25 to 80% accurate from all the major players. When was the last time you had a 20% untran and called yourself “neartime?” This also kills his argument about the technology being revolutionary. He’s comparing our 99% real-world accuracy rating to AI’s 80-at-best average accuracy and calling it revolutionary. This is more like if Google maps led you the wrong way down a one-way street about 20% of the time. It’s not acceptable and we shouldn’t be forced to pretend that it is. If they’re not using full automation, they’re using human transcribers, and that means there are zero efficiency gains from a manpower perspective. This is a hide-the-ball trick of saying technology is better than it actually is to fool investors and consumers. It only fools people who have not seen the trick before.

Next strawman argument by the liar: Court reporting costs have gone up. In actuality, we’re working for less than we were 30 years ago adjusted for inflation. Let’s call this out for what it is, a ploy to get court reporters scared of demanding the rates and pay that they deserve. Less money in our wallets means less money for us to spend on our associations to fight for us. The push to get court reporters to accept less has been largely successful in the last decade, and it has been driven by low-intelligence businesspeople that look at the labor expense as something to be cut no matter who it hurts. There are over a million lawyers in the United States and about 30,000 of us. We’re a rare commodity and need to start acting like it — keeping pricing reasonable, but not abusively low.

Notably, NSM refers to the democratization of technology and talks a good game about how realtime is too expensive for the little guys to afford. Anir Dutta of Stenograph also referred to the democratization of technology in the Speech-to-Text Institute podcast. What does this tell us? This is a coordinated buzzword in whatever business circle they’re all playing in. They’re using democracy as transfer propaganda. Who doesn’t like the sound of democratization in a free society like the United States? This ignores that in actuality adopting his active reporting model would likely hurt democracy in the form of disproportionately hurting the quality of black and minority speakers’ records. We have put immense effort into ensuring everyone has an equal record. Are we willing, as a field, to allow technological snake oil to kill the equality we stand for every day in every proceeding?

The puffery in the advertising is on full display:

Active Readback gets several things wrong in their advertising.

This looks intimidating to a stenographic court reporter that doesn’t grill it a bit. First, questioning our accuracy. How dare they? I just gave the science. They’re not guaranteed accuracy. Nobody can guarantee accuracy. What happens if a word is wrong? Does everyone get the service for free? That would be a guarantee. Tellingly, they make no such promise. Audio available? Stenographers have been using audio for years. It’s called asking nicely or getting a subpoena. Lawyers don’t want to re-listen to depositions anyway, that’s why they hire us. Exhibit handling, stenographers literally led the way and trained clients on that after COVID. The rest of it, hey, we can give all that away for free too, but we like our businesses to be profitable instead of losing $13 million a year like VIQ Solutions. We need profitable businesses so that we can continue to provide the same great service we have for over half a century. NSM’s investors must have their mouths agape. He’s not charging what the market can bear, and that’s a recipe for low returns and disaster in business. I’m pretty sure I learned that in business 101. What’s this guy’s excuse?

The low, flat rate that he talks about in the presentation isn’t really that low, which tells me that this process isn’t automated. Just to break it down, there are stenographers working for less than $4 per page in New York City right now. Assuming 60 pages an hour, that’s $480 for two hours. Lawyers can get the tried and tested stenography for a little bit more than the brand new maybe-this-works-maybe-it-doesn’t Active BS. This isn’t a sell, it’s an embarrassment.

“Active Readback, charging 72% of what a stenographer charges for half the quality, guaranteed.” (Parody)

Final point I’ll address is his mention about the shortage and how the stenotype is “hard on its operators.” We’ve been cremating our shortage despite some of the biggest names in the business, Veritext and US Legal Support, actively sabotaging us. Additionally, our technology is a lot easier on the hands than the Mechanical Turk game that Active and others are probably playing. Mechanical Turk lets services crowdsource transcribers. When people buy into active reporting, they’re likely buying into inefficiency and hurting workers.

Stenographers, I cannot stress this enough: Hold your ground. Our industry is worth $3 billion and we control most of that. The people that are trying to convince you to give up and run away are not doing so out of the kindness of their heart. These are liars, nothing more. Now that I’ve peeled back the curtain and exposed some of the flaws, I hope you will follow the Protect Your Record Project motto of “connect, educate, advocate.” I hope you will follow the STRONG motto of “we are strongest together.” I hope that if you found this blog post helpful, you will take the time to donate below.

I also hope that Active Readback will come on here and comment. We do not cower behind censorship like them. Perhaps that is all the world needs to see to know whose version of events is truest.

The more money I make from my media, the harder I can fight.

Donate Once:

Donate Monthly:

Donate Annually:

Choose an amount

¤15.00

¤30.00

¤50.00

¤5.00

¤10.00

¤20.00

¤60.00

¤120.00

¤240.00

Or enter a custom amount

Funding for the stenographic free press.

Donate Donate monthly Donate yearly

Stenograph’s Disrespect of Stenographers Continues with Texas

Stenograph has been in hot water because of its degradation of quality and service. This led to a boycott of the company by stenographers across the country, a boycott which continues to this day. As stated in my Oh My update, Stenograph’s push into automatic speech recognition is not being done properly. It’s being sold as a productivity boost, but available science says AI/ASR is a productivity killer. Anir Dutta, Stenograph’s embattled president, doesn’t care. He ignored a personal letter from me alerting him to these issues.

As if these issues were not enough, Stenograph promised to meet with Texas Court Reporters Association members and address their concerns. The company then retracted its agreement and set up its own meeting, likely to confuse consumers and attempt to manipulate us. TCRA addressed Stenograph’s behavior as follows:

Stenograph claims its plan is to meet with TCRA members.

Stenograph apparently pulls out because one member they don’t like might be in attendance.

Then, perhaps under the delusion that stenographers are stupid, Stenograph decided to hold its own meeting:

Stenograph attempts to create its own meeting in place of the Texas Court Reporters Association town hall.

This is a bait and switch. This behavior is disgusting and in my opinion we shouldn’t condone it as a field. It’s very clear what’s happening. Stenograph does not have an answer for why it is requiring stenographers to get releases for data it wants to steal from us or the liability it wants to be put on us, as per its licensing agreement:

“You understand you are responsible for obtaining consents and authorizations for data we may or may not be using.” – Stenograph (parody)

“We may be using your data to build our digital reporting products, but you’re not entitled to anything from it, which we’ve just unilaterally decided.” – Stenograph (parody)

Since Stenograph doesn’t have an answer, it doesn’t want to be in a position where that’s revealed. Again, I know factually that there are great people that work for the company and great software trainers for the software. That does not excuse what they’re doing. They’re barreling into automatic speech recognition in a haphazard, might-makes-right, and manipulative way that should give us all pause. We are the profession of blatant honesty. You say it, we write it. Can we not agree that this is not a direction we want a company, one that is practically our namesake, to take?

I have a message for Anir Dutta and Stenograph: We may not be computer programmers or $10 million companies, but human intelligence is not linear, it’s on a bell curve. We are not “stupid scribes” for you to play word games with. Words are all we know. We listen to people for a living, and we know when we’re hearing lies. If you have deluded yourselves into believing that you are so far ahead of all of us on the curve that you can lie to us with impunity, then I offer you the same stenographic proverb I offered Naegeli. TKPWHRUBG.

Stenograph’s Phoenix Won’t Rise From the Ashes

There are two ways to handle the November 4 STTI podcast with Anir Dutta, president of Stenograph. I can go point-by-point and try to poke at every little gripe I have, or I can go “big picture,” give people a rough outline, and let people decide for themselves. But first, let me just point out how obvious it is that STTI is a digital court reporting marketing tool. It’s November 2021, they’ve been around for two years, and they have one podcast. Now let’s compare that to a real field. Anna Mar’s Steno Talk first launched in March and is already on Season 2. Shaunise Day’s Confessions of a Stenographer also has done about 30x the content in the time it’s taken STTI to do one. It seems very strange that the “declining, shifting” industry has so much more content. Maybe there’s a lot more to talk about in an actual industry with actual news.

Now let’s do some big picture work. Artificial intelligence, AI, in its current form, is easy to understand. In brief, programmers use a recipe or instructions called an algorithm to tell a computer what to do. The computer is then fed lots of data. In automatic speech recognition, this data might be people speaking paired with accurate or semi-accurate transcriptions. Simply put, the algorithm tells the computer to go through the data and make future decisions based on patterns. Luckily for us stenographers, real life does not adhere to perfect patterns. Investors and companies that trust computers to make them money off of AI have a big chance of failure. Gartner predicted that 85% of AI business solutions would fail by 2022. We now have a real-life example of this. Zillow was using algorithms to predict the housing market. The value (market capitalization) of Zillow’s shares just plummeted $35 billion. Automatic speech recognition, ASR, which is AI for “hearing” and “transcribing” speech, has much larger companies than Zillow working on it. IBM is one of those companies, and for comparison, their market capitalization is, as of writing, about $103 billion. In a 2020 study of IBM, Apple, Google, Microsoft, and Amazon, companies with a combined market capitalization (value) of $8.873 trillion, the ASR was 25% to 80% accurate depending on who was speaking. My argument? If these behemoths haven’t figured this out, nobody else is close.

What is Stenograph’s answer to that? They mention in the podcast that they created the engine that their automatic speech recognition is running off of. This is meant to assure the listener that though there is clear science and data that suggests there’s no chance in hell Stenograph’s automatic speech recognition is better than anyone else’s, you should just try it, because there’s no risk to you — you only pay for it if you use it. They know that if they can get you to try it, some of you will experience post-purchase rationalization and keep using it even if it’s not very good. We, as consumers, need to be honest with ourselves. ASR is open source. Anyone can play around with it for free, customize it, and sell their version of it. Whatever Stenograph’s amazing programmers have cooked up is much more likely to be a tweaking or reworking of what is already available than a bona fide original work.

The podcast supports my assertion. It’s heavily laden with corporate propaganda techniques. Some common propaganda examples used generally in tech sales:

Fear appeals: Keep up with the technology or get left behind! Buy! Buy! Buy! OR ELSE.
Bandwagon: If you’re not paying for support, you’re not supporting the profession. EVERYONE must have this.
Name-calling: You don’t like TECHNOLOGY? Luddite! YOUR PROFESSION WILL GO THE WAY OF THE HORSE AND BUGGY, HAHAHA.
Card stacking: Our product is new, and wonderful, and we’ve put a lot of time and effort into it. But we are going to forget to mention that it would hurt minority speakers and allow large private equity companies to offshore your jobs with impunity.
Glittering generalities: Think buzzwords. Increase in productivity. Custom-built engine. If you don’t know exactly what something means, the salesperson does not want you to question it.
Transference: This takes someone’s good feelings about something and tries to transfer it to the company or product. Anir did this during the podcast when he said in the future technology will be “democratized.” We live in a republic that loves the concept of democracy. This is so powerful that when I heard Anir say it, I felt good. Good feelings make it harder to remember the bad things people do to us. It’s not really that different from any abusive relationship, it’s just a business relationship.
Ad nauseam: Repeat the same message over and over. If no one stands in your way, it becomes truth. For example, did you know technology is getting better every day? I guess someone forgot to tell that to whatever technology powers Stenograph customer service. Or perhaps technology is not getting better every day and that is a story we are sold ad nauseam.

I do not believe these things to be inherently evil or wrong. A sale is a sale. But when salespeople are used as instruments of ignorance, the wielder has gone too far.

On the topic of tools, some have pointed to our field’s adoption of audio sync and how that was widely hated and is now ubiquitous. Let me go on record and say audio sync hurt our field. Agencies started telling my generation of reporters “don’t interrupt, just let the audio catch it.” We trained an entire generation to sit there like potted plants while testimony was lost. No wonder so many from my generation left the field. They never developed the crucial skill of communicating our need to get every word. Dealing with the very simple skill of asking for a repeat became a harrowing and dreadful experience. More than that, audio sync kills productivity. In a dense layout, my transcription time is somewhere in the ballpark of 20 pages an hour. I used to use audio sync, and on a bad day, my transcription time was probably half that. I doubled my productivity by completely rejecting the “new” technology. I don’t disparage people that use audio sync, it’s a tool in our arsenal. Almost every reporter I know uses it to some degree. But beyond our post-purchase rationalization of “it is a wonderful tool for us,” I have not seen empirical evidence or reliable data that suggests it improved productivity or profit. It made us feel better because we could let some stuff go, and now it’s being weaponized to say “see, that worked out okay! This will too!”

On or about November 2, 2021, I wrote to Anir Dutta via snail mail. A copy will be downloadable below. I was very honest about my intentions. Stenograph had a chance to stop the boycott and didn’t even try.

Christopher Day’s letter to Stenograph Download

Maybe its trainers should sue the company for the lost income experienced during the boycott. There’s federal law against false advertising in 15 USC Section 1125. Seems to me that by continuing to press the ASR to consumers against available data and evidence, Stenograph has set itself and its independent trainers up for a massive loss. Stenograph is also potentially cutting into the earnings of its customers by pushing its ASR as a productivity booster when it may very well turn out to be a productivity killer. So if the company continues down this path and finds itself facing lawsuits, you read it here first.

Just to drive home my point about tech sales, I created a computer program that produces thousands of transcript pages a minute. The program code and a sample transcript are available for download.

Fake Transcript Creator Code Download

GeneratedTranscript407512 Download

Then I announced to the world that my brand new program could do transcripts faster than every stenographer in the country. None of you can disprove that. It’s true.

Tell your clients $1 per transcript. This is the situation we are all living together. Caveat emptor.