How the Inventors of Dragon Speech Recognition Technology Lost Everything
First time accepted submitter cjsm writes "James and Janet Baker were the inventors of Dragon Systems' speech recognition software, and after years of work, they created a multimillion dollar company. At the height of the tech boom, with investment offers rolling in, they turned to Goldman Sachs for financial advice. For a five million dollar fee, Goldman hooked them up with Lernout & Hauspie, the Belgium speech recognition company. After consultations with Goldman Sachs, the Bakers traded their company for $580 million in Lernout & Hauspie stock. But it turned out Lernout & Hauspie was involved in cooking their books and went bankrupt. Dragon was sold in a bankruptcy auction to Scansoft, and the Bakers lost everything. Goldman and Sachs itself had decided against investing in Lernout & Hauspie two years previous to this because they were lying about their Asian sales. The Bakers are suing for one billion dollars."
I thought Goldman Sachs were the good guys?
billions! It's so much money even the singular takes an "s"!
Mostly random stuff.
Did Goldman Sachs employees engage in some kind of fraud?
I would give my two left lugnuts to see white-collar crime like this vigorously prosecuted.
Goldman Sachs and the $580 Million Black Hole
Lesson 1: Just because you paid someone 5 million dollars doesn't mean they have your best interests at heart.
Lesson 2: If a deal involves putting all your eggs in one basket, you should assume that the basket is faulty, and that everyone knows it but you.
More power to them. I don't know much in the area of finance and the like, but stories like this continue to give me the impression large financial institutions like to play fast and loose with other people's (read: little guy's) money. Too big to fail? More like too big to be allowed continued operation.
My blood hurts...
IN BIZARRO WORLD
Imagine that, Goldman Sachs involved in swindling people.
Will wonders never cease.
inevitably Goldman will weasel its way out of it.
I read this story when it was published, the the guy that owned the company and lost his technology on top of all that money had a phd and was making huge advancements in the technology much faster than predicted, he's probably the reason Siri can exist. The saddest part is that he still had many improvements he was working on and now he just can't, the tech has been sold, he can't touch it anymore without getting sued. In the end humanity loses.
I don't like GS (have had multiple experiences). But I'm still skeptical. Something seems off here. Given lack of good counsel, why did the Bakers go out on a limb, change course, and agree to an all-stock sale at the last minute? Why not walk away, or hire better advisors to get you what you need, and challenge GS fees in parallel? $580 mil was at stake. $5 mil is 1% of the pie. Seems like the only reason to sell is to get liquid. Why go to all that trouble just to swap shares of one stock for even higher risk and reward?
Who owns Scansoft, who apart from GS are the other big winners from this transaction? They got the world's best speech rec software for a fraction of its true value - I wonder who they were advised by?
Korma: Good
"After consultations with Goldman Sachs,..."
That's your problem right there.
And using the dictation option. But I can't create a post because every time I say g---m-- s----, all I see are a bunch of expletives in the comment box
intellectual property law is philosophically incoherent. it is your moral duty to ignore it or sabotage it
http://www.nytimes.com/2012/07/15/business/goldman-sachs-and-a-sale-gone-horribly-awry.html?pagewanted=1&_r=2&hpw
I lean left on local standards (those of European social democracy) so I'd probably be something like "extreme left" on American standards (if we consider Democrats a left-wing party). I have no love for either Goldman Sachs or the whole sector they operate in... that said, you can hardly call what they did "stealing".
Are they unethical? Sure. Have they broken some laws by deceiving regulators? Probably. Misleading advertising? Might be. Fraud? Depends on the contracts they've used... but stealing? No. They've simply not cared about the fate of their clients - or the society - except where they had the economic incentive to do so. That kind of stuff happens when you have free markets.
For any given amount of freedom in the markets, you get some good and some bad sides. You thus choose a level where the good sides outweigh the bad ones... and acknowledge that the decision also leads to some undesired results. What doesn't work is choosing one level, at first ignoring undesired results and then, when they become too apparent, call them stealing, etc. without making an argument for choosing another level of freedom in general.
... and everyone in the industry knew that they were full of shit with their finances. The two founders - Lernout and Hauspie - were accountants, not technologists. The company had significant investment from the Belgian government, and L&H were politically well-connected enough to keep milking that funding source as they rode the tech frenzy of the 90's. In every instance where we competed against them, they always bid at ridiculously low prices that couldn't possibly be economically sensible. They had some decent technology, but their business practices were always suspect. It was very likely the taxpayer financing which kept their bubble from bursting before it did.
To be fair, Dragon was primarily involved in desktop ASR, whereas L&H (and my company - Voice Control Systems), were focusing on telephony applications. So, the Bakers may not have been fully acquainted with L&H's reputation. But, really, they should have been extremely suspicious of this deal, especially when the offer was changed to 100% stock. TFA didn't say what the value of some of the competing offers were, but I'm guessing they were substantially less than $580M. If it sounds too good to be true ....
None of this is to excuse Goldman, which apparently was hired to do something that they did not do. And clearly, the Bakers were done an injustice. But, something tells me that they were willfully blind to the possibility that this offer was unjustified. And for that, they should blame themselves.
If humans are mostly water, and beer is mostly water, then humans must be mostly beer.
....we have the comfort of knowing Goldman Sachs is still less evil than Electronic Arts.
You're dead wrong, EA only really screws its employees, goldman sachs will screw anyone who gets close enough, which unfortunately happens to include everyone with a stake in the world economy. Every unemployed person on earth currently owes a partial debt of gratitude for their state of employment to goldman sachs.
-=Geoskd
I wish I had a good sig, but all the good ones are copyrighted
you can read countless articles about Goldman trades within the past 10 years, since they became a public company especially, and find numerous examples of where Goldman 'analysts' or 'consultants' were wishy washy with the 'trading' folks , either their own or others.
just becasue finance schools teach 'chinese walls' and they talk about it in presentations doesnt mean it actually happens.
im a little confused as to what you would actually consider 'stealing'.
my favorite definition is from Jennifer Aniston's character in Office Space.
you take something thats not yours. and then it becomes yours.
doesnt mean that its factually incorrect. Goldman Sachs has a huge influence in the white house and in congress, and anyone who studies the history of the early 21st century will have to study Goldman Sachs.
If you have a good business that generates revenues, it is better to rely on it, rather than hope that 419 scams and "get rich quick" schemes will work for you.
The differences between GS and that Nigerian scammer are the better writing talent and the better government protection that one of them is using.
Siri doesn't do its voice recognition on the cellphone. It farms that out to machines in a datacenter somewhere.
Even if the 4 junior guys assigned to Dragon didn't know about the previous Goldman analysis of L&H (and they probably didn't), someone up top certainly did. Their work should have been looked over by a supervisor and/or others up top at Goldman. And those people should have known.
Also, the Wall Street Journal reporters calling the L&H "customers" proved that even cursory due diligence here would have proved that L&H were fraud.
And it's way too convenient that Goldman was representing a competitor (Nuance) at the same time.
Goldman was negligent and possibly also double-dealing.
Goldman's only hope here is to get out on a standing issue. If Goldman actually has to go to trial on this, they're screwed. A jury will crucify them. And rightfully so.
My wild guess? This settles before trial for a couple hundred million.
http://www.wired.com/wired/archive/11.02/code_pr.html
As that older article points out, the Bakers also spent some time early on at IBM Research doing speech stuff (and from when I was working at the IBM Speech group myself much later, it did not seem completely clear what way most of the knowledge was flowing). My undergrad adviser at Princeton, George A. Miller, who did a lot in the psychology of natural language and knew the Bakers (I think from when he was at Rockefeller with them), told me about this loss more than a decade ago, as a cautionary tale. More than the money, what really hurt most for the couple was not being able to work on their project anymore. For anyone who really cares about what they are working on, this is a good argument for working in the free and open source software realm rather than trying to finance proprietary software somehow, even when you think you are the "owner" of the software. Imagine if the Bakers had released Dragon as FOSS back then and built a consultancy around it -- at least they would not be alienated from their 20+ year labor of love (or "third child" as they called the software). In general, you also can't expect the same people who put their love into creating great things for the world to be fully prepared to deal with business sharks (even business sharks like GS being supposedly hired to "help" them). I'm glad my wife and I released our own labors of love (like our Garden Simulator and PlantStudio software) as FOSS instead of taking on investors and making it proprietary, since at least we can always still work with the source code. Of course, the flip side of that is often not having the time to do that because of a need to do other things for money. We ideally need a "basic income" and similar social changes to solve that problem and to minimize a software industry based around "artificial scarcity".
http://www.basicincome.org/bien/
http://www.artificialscarcity.com/
A 21st century issue: the irony of technologies of abundance in the hands of those still thinking in terms of scarcity.
Was trading their company for stock in another company. That was monumentally stupid.
I've spun up and sold a handful of companies now and not once did I accept stock in the purchasing company. When I exit, I EXIT, as in, that's it. The only reason I start a company is to sell it for a ton more cash than it took me to start and build it. I can then use a small percentage of that cash to start another one.
When you swap stock, you're trading something you know everything about (your stock) for something you don't know ANYTHING about (their stock). You don't know how they're cooking the books, or lying on their SEC forms, or laundering assets away, or playing other ENRON games. Certainly they could take the same tack about you, but then, you're the one with something they want, so they should be the ones to take the risk.
I have no problem holding onto a company that is making me money, if someone doesn't want to buy. If they have doubts about the way I do things, that's fine. I'll just keep making money until some other buyer comes along.
The ONLY time someone is in a hurry to sell is when they're losing money, and the company's days are numbered (like, in this example, the purchasing company who wanted to trade their crappy losing stock for some good, profitable stock). I can wait.
You are right. Last time I was offered millions of dollars, I asked them to pay me in camels. Way safer!
Why can't
There is a common denominator that has been overlooked here... Gaston Bastiens. For those of you who go back far enough, you may remember a couple of companies known as Quarterdeck and Datastorm Technologies. Gaston Bastiens took the helm of Quarterdeck and went on a buying spree that included slurping up privately held Datastorm Technologies. Like his tenure at L&H, this period was known for cooking the books and leaving the original folks hosed. It was a glorious day when those of us who used to work at Datastorm saw Gaston being led away in chains... http://www.flickr.com/photos/markdionne/294806554/ We weren't the least bit surprised that he had done it again... we were just pleased that he got prosecuted this time.
There are very specific rules regarding investing. If you are acting as an economic adviser to another entity, there is all sorts of legal language attached to that transaction that is supposed to insure that you are acting in the best interests of your client. It's similar to acting on behalf of someone else as their legal representative (in other words, lawyer)
If someone hires you as an adviser for their investments, and you tell them to invest in something that you know is dodgy, then you are in hot water. At the very least, there is the concept of due diligence, where you are supposed to do a through analysis of the investment before recommending it. It appears as though Goldman had done their due diligence before and decided against investing in L&H, which means they are in trouble if they then recommend someone else make a similar investment.
My Other Computer Is A Data General Nova III.
I actually wrote a fair amount of UI code interfacing with Microsoft Agent as part of a research project, AutoTutor. While the L&H TTS engines were indeed the default for Agent, that's just because they were the default ones installed on Windows (2000 and XP) at the time. Agent allows you to load the TTS engine of your choice, so long as it supports the Speech API. Because the Speech API includes callbacks for phonemes spoken, Agent can synchronize lip movements of the character to what's being spoken by the speech engine regardless of its creator.
Ultimately, the poor quality of the L&H voices led us to SpeechWorks and AT&T's NaturalVoice products. Sadly, both the TTS and voice recognition fields went through major consolidations in the early 2000s, and now SpeechWorks is dead (acquired by Nuance). NaturalVoice is still available, more or less, from Wizzard Software.
The Freelance Wizard
It's very unfortunate that most of us geeks are not that well versed in financial thingy, and that we geeks are too trusting of others - methinks this maybe the result of we geeks, in our own geeky-universe, are generally trustworthy
Unfortunately, the world out there is filled with critters that will cheat everybody, even their own mothers !
The two inventors of Dragon Speech Recognizing System put their trusts in Goldman Sachs, which, by now, most of us geeks should know not to trust
It is even more unfortunately that the two inventors now have to rely on lawyers to sue Goldman Sachs - and lawyers are critters that are not that far removed from the kind of critters that occupies the Capital Hill / White House / Wall Street
Methinks the only way we geeks can survive the world out there is to turn ourselves into the baddest kind of critters - even more badder than the critters of Wall Street, critters in the Capital Hill and/or the White House
Muchas Gracias, Señor Edward Snowden !
It seems there is some karma in effect here.
http://www.planetcrap.com/topics/8/
The only problem with shooting them is that their personal money is probably safely tucked into offshore bank accounts. Threatening to torture their soon-to-be-orphan children to death should get them to relinquish those account numbers however.
The legal argument here will almost certainly be that, by accepting $5 million from the Bakers for consulting fees, Goldman Sachs had a fiduciary duty to act in their best interest regarding the transaction.
You can argue that they should have known better than to "put all their eggs in one basket". But the fact that the client should have known better is not a legal defense to a breach of a fiduciary duty. That's why you're paying the professional in the first place – because they're supposed to know better than you! They have a legal obligation to use that knowledge in your interests, not double-dip on the side.
The fact that Goldman considered investing in L&H and then specifically declined to do so is strong evidence that this recommendation to the Bakers was in violation of their legal duties.
The paid 5 millions dollars? And their advice was to sell their company to a sinking ship (either a ship Goldman knew was sinking, or with a tiny amount of work could have figured it out)
And their defense is "We completed the transaction." That is like saying "Yeah we sold your stuff to the con artist. We knew he was a con artist, and he paid in fake cash, here you go. We did our part"
I think sueing for 1 billion isn't enough. And good luck to the Bakers, from the story it seems they deserve to win.
It is painfully evident to anyone who has used Dragon software that they are not technologists.
I was referring to Lernout and Hauspie - the founders of that company - as not being technologists.
Dragon technology is the state-of-the-art in large vocabulary speaker-independent ASR. Jim Baker's basic mathematical framework is still the basis for all commercial ASR systems. This included the L&H ASR technology at the time of the deal. Each company's ASR engine has its own bells and whistles to distinguish it, but the main differentiator these days is the access to domain-specific data for training the statistical models - in combination with the many heuristic tricks used in building those models.
It is hard to know the source of your bad anecdotal experience with Dragon Naturally Speaking. The technology has its limitations, and just "doesn't work" in probably 20% of the situations (broadly defined) in which it is deployed. The field needs another 2 or 3 fundamental research breakthroughs before any random use case could be guaranteed to be effective. But, many people have found Dragon dictate quite useful, particularly in restricted language domains, such as legal, medical, and business emails.
Oh, and the ASR engine behind Siri _is_ a version of Dragon, so your post is somewhat non-sensical.
If humans are mostly water, and beer is mostly water, then humans must be mostly beer.
Go Jim and Janet, Go!
Don't forget legal fees and damages!
Former Dragon Employee, wishing you well.