Listen to Kurzweil Interview, Part 4
Ray Kurzweil (continued): I mean I have a dozen honorary doctorates. That's where the doctor sometimes comes. I don't insist on that.
Tony Candela: And this first business that you started. What was it?
Ray Kurzweil: When I was a sophomore in college, I created a system that matched high school students with colleges. It was a pretty comprehensive system and we ran tens of thousands of students and sold it to Harcourt, Brace and World, now Harcourt, Brace, Jovanovich, a publisher.
I put myself through college and helped support my parents. My father was ill at that point, with heart disease.
Then in '73, three years after graduating, I started Kurzweil Computer Products with the idea of... Well, we'd developed Omnifont optical character recognition and then it was kind of a solution in search of a problem: what would this technology be good for?
The technology was the ability to recognize printed letters in any type font. Because there was character recognition at that point but it used something called template matching, which couldn't even do size normalization. It basically matched pixel to printed image with just a picture of what the character was supposed to look like.
So the system was just to electronically match points, pixel for pixel. So if it matched the picture of an A then you'd say that it was an A. It wasn't able to intelligently infer that it had the properties of an A. It wasn't able to even break touching letters. Letters had to be fixed space. They couldn't be proportionally spaced. Almost everything printed is proportionally spaced. And they could only read one type style. And then they came out with two type styles: Courier A and OCR B type fonts. But it was by template matching.
You might wonder what that's good for, since nothing exists in the printed world with those characteristics. But literally they would retype material, using a Selectric typewriter and then scan the type sheets. You might wonder what the benefit of that is, cause you still have to re-key it. Why don't you just key it into the computer to begin with?
Well back in those days, a computer terminal was very expensive and a Selectric typewriter only cost $400 or $500. So you could type it on a cheap Selectric and then scan it, as opposed to using an expensive computer terminal.
But anyway, we invented the first optical character recognition that could recognize any type font and also deal with printing errors, like touching letters, broken letters, third generation photocopies and so on.
And we had a number of applications in mind, including the blind reading problem. It was a plane trip I took where I sat next to a blind gentleman who was articulating how blindness was just an inconvenience. He represented his company. He flew all over the world. Not a problem to get to meetings, to attend meetings.
But then he said actually there was one problem, one handicap associated with visual impairment that would be good if there was a solution. Which was the ability to access ordinary printed material because recorded books and braille were only available for a small fraction of books, a couple of percent. Business memos, and so on, were not available.
He could get readers but you have to schedule them and there's a privacy issue. So that would be great. If he could actually read printed material. That was the only handicap that he could think of.
And that was inspiring. That kind of put us over the edge to make that the application that we pursued. We needed then to develop two other technologies, which were the flatbed scanner, cause that didn't exist. So we invented the CCD flatbed scanner and we also invented the first text-to-speech synthesizer. We put those three technologies together to create the Kurzweil Reading Machine.
Tony Candela: Now the flatbed scanner technology really didn't exist? There was nothing like that?
Ray Kurzweil: There were sheet fed scanners that were not based on CCDs. There were no CCD flatbed scanners. We developed the first one. And all these scanners you see around, they are all CCD flatbed scanners.
Tony Candela: And the actual speech synthesis...
Ray Kurzweil: We developed our own full text-to-speech synthesis. There had been pieces of it. There was a Votrax chip that could articulate phonemes but you couldn't feed text into it. So we developed a system that could take the text, recognize the text from the OCR, and speak it. It was the first full text-to-speech synthesizer.
Tony Candela: I don't think a lot of people realize that, that that element of it needed to be developed as well.
Ray Kurzweil: There were really three technologies there.
Interestingly, the optical character recognition, the Kurzweil OCR, has remained in existence and continues as the number one OCR package in the world, for thirty years in a row. It's called Xerox Textbridge now, from Scansoft. Scansoft used to be a Kurzweil product but it's been in continuous development for thirty years.
We introduced it actually on January 13, 1976. I remember that because Walter Cronkite used it for his signature sign off. He used to end every program with, "That's the way it was," and he'd give the date. So this was the first time that he ever had, that he didn't deliver the sign off himself. He had the reading machine sign off, "And that's the way it was, January 13, 1976."
That press conference generated a lot of attention. Jim Gashel of the National Federation of the Blind demonstrated the machine and it was on all three nightly evening network news programs. I then took it on the Today show. And that's actually how I met Stevie Wonder because he caught that show. He heard me on that show and then called us up and said he wanted to stop by and get one and we actually just had one production model we could give him so he came over and we taught him how to use the reading machine and he went off in a taxi with his reading machine.
Tony Candela: Did you sell it to him?
Ray Kurzweil: No, we gave it to him.
And that was the beginning of what has been a long-term friendship. So we spent time together when he came to Boston or when he was in Los Angeles. A number of years later, in 1982, we had this conversation about these two worlds of music: the acoustic world and the electronic world and wouldn't it be great if we could create a bridge between them and actually have electronic synthesizers sound realistic with the sounds of acoustic instruments.
That's how the first Kurzweil Reading Machine came into being.
Tony Candela: And you continue to develop synthesizer technologies through the years as well?
Ray Kurzweil: Well, Kurzweil Music was founded in 1982 with Stevie Wonder as a musical advisor. In 1984, we came out with the Kurzweil 250. It was recognized, has been recognized as the first synthesizer that could realistically capture the piano and other orchestral instruments and we developed then a whole line of products, just like the reading machine. They got more powerful and less expensive.
I sold that company to Young Chang in 1990 and it continues as a major line of musical instruments. It's a possibility that I may be involved in buying it back one of these days.
Tony Candela: When was the decision made to go away from the hardware/software reading machine (which started out as a very expensive machine and started dropping in price but still stayed very expensive) to a software solution?
Ray Kurzweil: The rest of the history of Kurzweil Computer Products is: 1978 we created a more capable version of optical character recognition and put out the Kurzweil Data Entry machine. That became a front end to databases. There was a whole database field being created. Lexus and Nexus, still the leading databases for law and news, were created using the Kurzweil Data Entry machine.
We sold the company to Xerox in 1980. They saw it as a bridge back from the world of paper. They had all these machines that took electronic information and put it out on paper. We had the scanning and the optical character recognition to bring it back into the world of electronic information.
They also liked the blind reading machine project as it fit in with their image of a socially conscious company.
So, we sold it in 1980. I remained actually a very active consultant for that business through 1995, fifteen years.
Around that time, they had really over the years lost interest in the reading machine area. They did continue quite aggressively with the optical character recognition. It was called Xerox Textbridge. That continued to be the leading optical character recognition (software). They took that company public. It's called Scansoft now. It continues today to be a very successful company.
So in 1996, with Mike Sokol, who had been working for Kurzweil Computer Products and the Xerox subsidiary for many years, as head of marketing and sales, we started Kurzweil Educational Systems in 1996.
At that time we felt that the consumer versions of personal computers and scanners were powerful enough that we should do an all-software solution.
So in 1996 we developed the Kurzweil 1000, which was introduced at the end of the year and we put together a distribution channel with all of the dealers that had worked for earlier versions of the Kurzweil Reading Machine company. That company continues today as a leader in reading machines for both the blind with the Kurzweil 1000 and kids with learning disabilities and reading disabilities who use the Kurzweil 3000.
Tony Candela: And there was a period in there where, of which I'm sure you would like not to remember where Lernout and Hauspie had a piece of this action.
Ray Kurzweil: Yeah, we sold the company in '98. I remember it was actually a pretty good deal for our shareholders. It turned out it did not affect the company at all. They had sort of bigger fish to fry, bigger problems. They were acquiring companies like crazy and they almost seemed to... forget that they'd acquired this company.
So Mike Sokol continued to run that company. Nobody left. He and the management team just ran it, the same way that they had done it before.
So then in 2000, well in 2000, Lernout and Hauspie hit a wall with its own financial problems and so we had the opportunity to buy it back, which we did. And through all of those transitions, nobody left. It was the same team, same products, so there was complete continuity. It has grown over the years. It's fifty or sixty people now.
And so it continues today as a strong company.
Tony Candela: We're in Wellesley Hills right now. They are not in the same town.
Ray Kurzweil: They're in Bedford, ten minutes from here.
Tony Candela: It's not that far.
Was there anything that had to happen for you to be able to go to a software solution?
Ray Kurzweil: Well, an enabling factor was to have the necessary consumer hardware that could do what you needed.
Tony Candela: It was the hardware.
Ray Kurzweil: But scanners had gotten to be very powerful and very inexpensive and personal computers were also capable of doing what we needed. We didn't need any specialized hardware any more and we thought that the user would benefit by being able to use consumer electronics because that's where you get fantastic price performance so they can take immediate advantage of the ongoing improvements and not be so stuck with niche applications that don't change.
Tony Candela: Did you know that this was what was going to happen next as soon as the hardware became more powerful? Were you pretty much ready and waiting for it?
Ray Kurzweil: We were waiting for it for awhile, some time. I don't know that we talked about that early in the history of the Kurzweil Reading Machine project. But in the years leading up to 1996, the idea was around. There was kind of an overlap period where there were software solutions and hardware solutions both in the market. But the price performance benefits of the software solution quickly outperformed the hardware solutions and so there aren't any hardware reading machines anymore.
You'll find some in Europe, just because of the reimbursement...the governments there. Like in Germany, will reimburse the cost, if it's a specialized device. And so if it's consumer hardware and software, they don't treat it the same way for reimbursement so just to kind of get around that government regulation they'll put them together and say [that it's] not a computer or scanner [but] a specialized reading machine for the blind.
Tony Candela: The equivalent of the durable medical device in American health insurance, sounds like.
Ray Kurzweil: Exactly.
Tony Candela: Well Ray, I know time has gone very short here and I know where you are now and I have a feeling that we're all waiting to see what really will end up happening in the future and that you'll be proven right.
For now I'd love to thank you for a great interview and having your voice on tape is a wonderful thing and we're all going to enjoy listening to it.
Ray Kurzweil: I enjoyed the dialogue very much and I will say that although these various technology trends really are very persistent, and I believe we can confidently predict these capabilities for the future, the future will be even more remarkable than what we can imagine now. Because we'll have millions of people who are applying their creativity in these future years, coming up with all kinds of exciting inventions and applications using these very powerful computers and communications systems.
I think it will be a very remarkable future.
It is true that both our creative and destructive sides are amplified by the exponentially growing power of technology but I do think the creative side has an edge, providing we maintain an open, democratic system that respects the free exchange of ideas and our sort of perhaps hard to define ideas around liberty and freedom are important because I do think they give the edge to the creative side of our technology.
Tony Candela: If you saw that society was not handling this well, would you react in some way? Would you change what you were doing? Or do you think you would continue forward with the hope that we would straighten ourselves out?
Ray Kurzweil: Well it's a complicated question. There are certain things we shouldn't do.
I think we should try to avoid making the tools available to do great harm easily available. I would not support putting the design for an easy, homemade atom bomb on the Internet. Which, by the way, has been on the Internet and there's been great activity to get it off the Internet.
I think there are some limitations and I think there needs to be some regulation, although it's a complicated question. I'm very concerned that we slow things down too much with medical regulations.
You might note that there's no regulation whatsoever in software. You don't have to get a license from the government in order to be a software practitioner and software products don't have to be licensed by the government and it's arguably the most productive industry in human history.
And when problems occur, like a software virus, there's a solution that emerges in hours and so our sort of technological immune system that deals with all of these software pathogens keeps evolving as well.
So the legal system works quite well and there's no regulation. I'm not advocating that we do away with all regulations in the medical field but I am concerned, for example, that a bio-terrorist does not have to put his inventions through the FDA. Whereas the scientists we are counting on to protect us do and it slows down every step of the process. We need to really streamline that process.
And we don't balance risks appropriately. One kid dies in gene therapy trials and they stop the research for six months. Well, how many hundreds of thousands of people will ultimately die because of that delay?
These unattributable risks, we know that if we delay a therapy that could stop heart disease for a year it's going to kill hundreds of thousands of people. But nobody cares about that because people have been dying of heart disease for eons. Whereas, if somebody gets hurt, five people get hurt by something, that's attributable. Then the CNN cameras are there and Congressional hearings and so on.
So we balance the unattributable to statistical risks a hundred times greater than the identifiable problems. That's actually not a good strategy. It leads to a lot of suffering and a lot of delay. I've had experiences where there's friends who have cancer and desperately need some drug and based on a lot of investigation, I believe there's an experimental drug that has a good chance of helping. You can't get it.
It's a whole other issue but there are situations where there's certain destructive applications that should be inhibited.
Tony Candela: Well, I'd like to finish up on a light note with a question. You have cats in your office but they're model cats.
Ray Kurzweil: Actually at home I have 300 cat figurines. That's one thing I collect. You asked what sort of surprising hobby I have. That, I guess is a surprising hobby, that I collect cats. Some are one inch, some are a foot high.
The cats here are fat cats and that's sort of a symbol for my current company called FATKAT, which stands for Financial Accelerating Transactions from Kurzweil Adaptive Technologies, F-A-T-K-A-T.
And we're applying my field, which is pattern recognition, to finding patterns in the stock market that we can utilize in automated trading.
We have a system that makes 80-90% a year, completely automated, that we've been experimenting with, with real cash and we plan on launching a hedge fund based on that technology.
Tony Candela: Excellent.
Ray Kurzweil: Another little pattern recognition project I'm working on.
Tony Candela: This one I want to know more about, but we'll do that off tape.
Thank you so much, Ray. I appreciate it.
Ray Kurzweil: Yeah. Enjoyed it very much.
