I’ve been following Verbit for a while. I’ve pointed out my skepticism regarding its series A funding claims of 99% accuracy in record time. That was well before the racial disparities in automatic speech recognition (ASR) study really hit the news and showed us that all the biggest players in the ASR market were having trouble consistently getting 80% accuracy. That was well before the captioning article showed us “record time” was 8 hours. My skepticism was revived when I read about Verbit floating the idea that it might go public. It seems very much to me like they’re unprofitable, like VIQ. It’s a game of getting investors to keep piling money on as far as I can tell. It’s the Uber strategy of throwing money around and trying to control a market whether or not the model is sustainable. We have a lot of reasons to doubt what Israel-based Verbit says.
Time to add one more to the pile. They claim they’re a New York-based company. They claimed that in May 2020.
The claim was repeated in a June 2021 article by Peter Cohan in Forbes.
And despite my attempts to alert them to the inaccuracy in July 2021, nobody could be bothered to correct the article. It’s still wrong as of September 14, 2021.
And just to make this really clear, it’s fact checkable using New York’s business search, which takes maybe 60 seconds. Verbit is a foreign corporation, meaning it is not based in New York.
This might seem like a minor thing, but it points to a larger problem. Media people are not bothering to fact check anything. They’ll go on and on about how the technology is great and new, and how this company is a unicorn valued at a billion dollars, but they’ll miss simple realities, like 85% of AI business solutions being predicted to fail. IBM Watson wasn’t the holy grail of ASR and IBM makes $70 billion in annual revenue. What are the chances that Verbit beat IBM with its $5 million in revenue? I’ll give everyone a hint. Vince McMahon’s theme song tells us exactly what chance they have.
And Verbit is not doing anything to correct false perceptions. They reposted their May 2020 article again on September 14, 2021.
Just for fun, let’s dive into the implications they list here, since it’s being published a second time as if it is still true.
1. The rise of non-compete litigation. I see no reason to believe that this is an accurate assessment. States like New York are banning non-competes under 75,000. Even our sitting President of the United States doesn’t seem to like non-competes very much. So it probably wasn’t true in May 2020 when the courts were closed and probably isn’t true now.
2. Courthouses are closed. True in May 2020. Not really holding true now.
3. Working from home culture. Stenographers adapted to this. There’s no edge to Verbit in that department.
4. High demand for lawyers. Can’t argue here. Our nation of laws needs lawyers, especially in rural areas.
5. Technology is key. They mention how lawyers that know how to send documents electronically and perform video conferences are more desirable. Is this surprising to anyone?
6. Fewer courtroom cases. Verbit has pointed to our stenographer shortage in the past as the casus belli requiring our replacement. If there are fewer courtroom cases, demand is lower than anticipated, and therefore stenographers can meet demand and the whole theme that we cannot has been a marketing farce.
7. Smaller law firms thrive. They’re writing this because smaller law firms have fewer resources to spend figuring out that the article is a sales pitch. Marketing is about how you make people feel. They want to make smaller law firms feel good and try Verbit.
8. New court reporting strategies. In May 2020, laws regarding oaths and the swearing-in of witnesses were changing to adapt to the pandemic environment. This has been a major debate in our field where some businesses ignore procedural rules while others zealously defend them. New York itself has fairly simple guidelines for depositions taken within the state, without the state, and in a foreign country. As page 32 of the Summer 2020 Vermont Bar Journal told us, this situation gets complicated. So it’s not a false statement they’re making, but this is an example of framing. “New” and “court reporting” are designed to make the reader feel like court reporting is changing. Our strategy is the same it’s been for a hundred years, stenotyping what you say while you say it. We just do it with better technology than we had in the 80s.
9. Courts are going digital. Yeah. Sure. But as I have pointed out to digital court reporters, stenographers have been digital for a while now.
10. The rise of the remote deposition. Automatic speech recognition thrives via the remote work because the audio quality tends to be much clearer, assuming everyone’s connection is good. It’s a closed scenario where everyone is speaking into a microphone. By contrast, the stenographic court reporter can survive anything. Check out 25 seconds of one of my early freelance jobs and let me know how well automatic captioning does there. I was a 20-something year old kid next to a steam radiator. If I had not been taking notes on my stenotype, there’d be no legal record of the proceedings. Automatic speech recognition fails in court reporting for the same reason court reporters get stressed out at lawyers. We have to get every word. Sometimes they stick us in spots where it’s really hard to do our jobs. In today’s world we are occasionally looked down on for asking to change our seat or relaying that a situation is unreportable. We will be very upset if the legal field suddenly decides “yes, we can create the ideal hearing scenario for the computer that we couldn’t bother to do for the human beings we work with every day.” But my money is on one simple truth, people are people and most of them will never jump through hoops to make a computer “happy” when they can work with a live stenographic reporter who will jump through hoops to make them happy. It’s the same reason customers dread calling any kind of service center nowadays. Getting bounced around by an automated system has got to be one of the most infuriating experiences in modern life. Applying that to the legal record is a masterful level of stupid.
But you know what’s screwed up? Here Verbit is calling digital court reporters highly trained, but not long ago, they were claiming that digital reporting required a workforce that is not highly trained. Again, this is a company with no conviction or facts backing it. It is a chameleon, ready to blend in with whatever way will make it money or sound good.
Let’s keep on reading some digital reporting myths.
AI never has a bad day? Well, in my October 2020 article, YouTube thought the caption for defeating the enemy and extinguishing his life was “to feed my enemy, I extinguish his wife.” In my June 2021 article YouTube AI thought “raise your right hand” was “rage right hand.” There’s two bad days right there. If Verbit’s got better ASR than YouTube, why haven’t they sold it to YouTube yet?
To understand why this is wrong, you have to know a little about the tech and concepts at play. Alexa and Siri are constantly able to learn your voice and tune to your voice. That’s like voice writing. In order to create a uniform ASR program that can get all English speakers all the time and automate that transcription, you need tons of data from all those speakers in all different types of environments. Since new people are being born every day and language is changing a little bit every day, this is basically hopeless. As written in Scientific American, ASR is not perfect and may never be. Just think criminal prosecutions. Does anyone really believe we are going to get defendants to sit there and help the court system train the computer to their voices? “Ah, yes, I think I will just assist the state in my prosecution.”
For anyone that hasn’t caught on, there is a pattern here. There is little substance, a lot of fluff, some great sales tactics, and no real court reporting knowledge. Perhaps most offensive is their reliance on quotes and ideas from the National Center for State Courts, which as far as I can tell just doesn’t like stenographers, since they continually call for digital recording despite some evidence that costs are similar and stenographers are more efficient. I hate to say that about NCSC since they seem to admire community court solutions as much as I do, but that’s where we’re at, they don’t like that my job exists.
I really feel for investors. They’re being recklessly encouraged to throw millions of dollars into something that, from any reasonable view of the facts, has a high chance of failing or stagnating. As I pointed out in my science article, they’re paying Kenyan transcribers maybe a fourth of what Americans are paid for the same work. Any alleged savings doesn’t go to the consumer, it goes to the company. Does the court reporting consumer want the creators of the legal record to be outside of his or her subpoena power? Does the captioning consumer want a company to push down prices so that captioners have a hard time affording continuing education? Is everybody really okay with what is apparently a zombie company coming in and sinking millions of dollars into Rev 2 under the false notion of “future technology?” Livne himself has admitted they’re “over-subscribed” when it comes to funding. It’s quite clear to me that they’re overfunded because they’re turning out to be an overblown transcription company and not the cutting edge of technology. After all, just compare their “over-subscribed” funding of maybe a couple hundred million dollars to the money pit of real AI research. When the media will admit that or when investors will catch on? That remains to be seen. But very much like US Legal, anything from Verbit needs to be viewed with extreme caution.
For investors looking for a stable return, consider getting involved with stenographic firms. Voice recognition and transcription has been identified as a market with billions of dollars in potential. Stenographers are the most efficient modality in that regard. Where technology companies will overpromise and underdeliver, the stenographic writer has worked out a system that has been going strong and evolving for over a hundred years. A Kentley Insights 2019 report showed a 10% profit as a percentage of revenue for court reporting businesses. As far as I am concerned, a far safer and more stable return is in stenography. If any investor wants to be directed to the more entrepreneurial minds of our profession, I am happy to direct. Please write me at ChristopherDay227@gmail.com.
I reached out to Jim McMillan from NCSC and I have to correct my above position on the organization. He explained that he believed quote Verbit used from him was from a 2013 post and that that was well before speech-to-text automatic speech recognition was close to usable. The position that NCSC takes tends to be on courtrooms that do not require the transcription of many matters. Obviously, I will always be an advocate for the stenographic reporter, but this is a far different take on it that I previously had and important for our field to see.