Pattern recognition

Having worked in so many areas of technology, do you see any recurring themes? In terms of the technology, my area of interest is pattern recognition. Character recognition is a classical pattern recognition problem. Speech recognition is one also. The music technology uses pattern recognition and advanced signal processing to recognize the salient patterns that make a piano sound like a piano. Now we're applying these speech and language technologies to Kurzweil Educational Systems for learning-disabled individuals.

Who is He? Ray Kurzweil's brainchildren read to the blind, convert speech to text and create music without using musical instruments. He's now at Lernout & Hauspie Speech Products NV, to which he recently sold his educational software company. He studied computer science (under Marvin Minski [sic]) and creative writing (under Lillian Hellman) at MIT.

The other major theme is trying to apply these technologies in ways that enhance human communication. One major application of that is applying computer technology to the needs of the disabled. We're not yet creating cybernetic geniuses; the intelligent computers we're creating are very narrowly focused. But they can perform functions we normally associate with human intelligence -- sharply defined ones, like reading a book out loud or recognizing human speech. So we're able to apply these technologies to overcome the narrowly defined handicap or disability of a person.

It's clear some of this is facilitated by Moore's Law. Moore's Law is a great facilitator. All the technologies I've been involved in are very hungry for MIPS and megabytes. It's only been in the last six to 12 months that PCs have been powerful enough to do continuous speech recognition. Just two years ago, you had ... to ... speak ... like ... this.

In your book, you project how pattern recognition and machine intelligence will grow and be used. What's next? One example: I'd say that by 2004, a common feature of your cell phone will be the ability to translate from one language to another. You routinely will be able to speak to people anywhere in the world and understand each other's languages.

What other major advances do you see in the next two to five years? In 2009, the $1,000 PC will perform about a trillion calculations per second. We'll have very high-resolution displays that exceed that of paper. In general, there won't be paper-based publications anymore. We won't have cables between our components; they'll communicate in a wireless fashion. The majority of text will be created using continuous speech recognition. We'll be talking things over with our computers. Most routine business transactions will take place between a human and a virtual personality. Most learning will take place with intelligent courseware; human teachers will be more like mentors or guides.

As the book's title indicates, you see computers as getting smarter than people. By 2019, a $1,000 computer will match the processing power of the human brain: 20 million billion calculations per second. By 2029, $1,000 of computation will match 1,000 human brains -- and that's the hardware intelligence. The software will take us longer than 2019, but by 2029, we'll be able to match the flexibility and intelligence of the human brain, in part by actually reverse-engineering the brain. Once the computer achieves a level of intelligence comparable to human intelligence, it'll necessarily soar past it [because] computers can readily share their knowledge [and skills that they've learned].

I think the world of human and machine intelligence will begin to grow together. We'll be placing intelligent neural implants in our brains to enhance our sensory abilities and our perceptions, memory, reasoning faculties. ... We're already doing that to a limited extent.

We'll also be able to plug in to the World Wide Web directly through our brains, without any external equipment. And virtual reality will be much more compelling than the very crude version we've experienced today.

Being a prophet is an interesting but challenging profession. It's been said that we tend to overestimate what can be done in the short term and underestimate what can be done in the long term. The predictions I'm making are based on technology you can touch and feel today ... what we can't anticipate are the discontinuities. There will be some [discontinuities] in the 21st century, and that can only accelerate things.

Dern is an independent author, speaker and consultant in Newton Centre, Mass. His E-mail address is ddern@world.std.com; his Web site is www.dern.com.