Alumni Profile: Peter Bock ’62 & Artificial Intelligence
The world of the future is developing today through Peter Bock ’62 of Washington, D.C., the fall 2012 Knop speaker at Ripon College. Bock is a former NASA scientist and a professor emeritus of engineering at George Washington University in Washington, D.C., where he was on the computer science faculty for more than 40 years. Before that, at Ripon, he and his wife, Donna Oberholtzer ’63, sang in a jazz group with Al Jarreau ’62 and Thomas “Duffy” Ashley-Farrand ’62.
Bock is an international expert on artificial intelligence (AI) and cognitive science. He pioneered the development of Collective Learning Systems theory, an adaptive statistical-learning paradigm for artificial intelligence. He and his research team developed the well-known AI engine known as ALISA (Adaptive Learning Image and Signal Analysis), used by prominent firms and institutions around the world. Since 1990, he has been working to extend the cognitive capabilities of ALISA and develop practical applications for it.
Bock’s expertise crosses multiple disciplines, from computer science to cognitive science, anthropology, developmental psychology and the humanities. These varied areas are necessary to understand how the human brain operates, he says.
“It’s a big challenge,” he says. “The DNA you inherited from your folks contains enough information to build your body, but comes nowhere close to the amount of information you need to fill your brain and live life. You begin with an empty computer for a brain. The rest is learned.”
ALISA first responded to images and signals, and in 2007 the ALISA team decided to train ALISA to create something. “This is how ALISA the artist was born,” Bock says.
“Humans record symbolically,” Bock says. “We write, paint and record our experiences in many different ways and pass them on, so our offspring don’t have to learn everything from scratch. That progress is critically important.
“We trained ALISA to imitate the great art masters. There is a great deal of thought and a huge amount of information that goes into her creations. She pauses, rejects and changes her mind, trying to make decisions about what to do. To get it right, it takes about six hours to complete a work, comparable to human effort.”
Bock’s primary research objective has been the same for several years: to design and create an artificially intelligent being named MADA (Adam spelled backwards), whose cognitive and emotional capabilities are on a par with humans. Bock and his team are on schedule for the “birth” of the personal intelligent being in 2024, and he says he anticipates this event with both excitement and trepidation.
“The greatest fear is that we haven’t taken to heart what Mary Shelly wrote about in her novel ‘Frankenstein,’ published in 1815,” he says. “The hero is a being that Dr. Frankenstein built but gives it no name, simply calling it the Monster. All the Monster wants is a friend, a mate, but Frankenstein refuses to do this. The Monster is alone, angry and emotionally bereft. And tragedy ensues.
“If we build an artificially intelligent being – MADA – what will stop us from building another and another? The human race has a long record of treating groups of strange or different intelligent beings very badly. In this country, we had 300 years of slavery; in Egypt, about 700 years; in England, 150 years. And it’s still going on today; we seem to have learned little about this as a species. This is a very real moral hazard for us humans. So we’re in real danger of building a creature that could be mass-produced and enslaved. We must not let that happen. If any beings are smart enough to make useful slaves, then they will be smart enough to understand their tragic predicament.”
Some question that if this danger exists, then why develop artificially intelligent beings at all? Bock says it’s an inevitable pathway, given the motivations that drive humans. If he doesn’t pursue it, somebody else will, and he feels his team’s appreciation and acceptance of the moral hazard in this pursuit are greater than industrial or military standards and practices.
MADA will not be born an adult but an infant with an irresistible thirst to learn, just like human children. “An infant doesn’t speak, but she sees, hears and processes everything around her, absorbing information at the rate of life,” he says. “Life will move around MADA at our pace. And although she eventually will be able to read and do other things faster than we do, she will have to interact with the rest of the world at its own speed. She may be a bit impatient with our pace of living at times. But then, most human children are, too.”
He says that ALISA, the current AI engine, is MADA’s grandmother. The daughter of ALISA, CONVERSA, will be able to converse with humans. It took humans 600 million years to evolve to this point. MADA’s evolution will be complete after 60 years of human guidance.
“We have been predicting MADA’s birth date of 2024 for the last 40 years, and we are right on schedule,” Bock says. “But there is a technological caveat looming in the near future: to make enough fast memory available for MADA’s brain, we need a technology change. We need a new technology that provides much more memory and generates far less heat. (Our brains give off only about 35 watts.) If we don’t have it by about 2018, the memory capacity of our computers will plateau at about the intelligence capacity of a dog. Not very interesting for artificial intelligence research.”
If MADA is brought into being, Bock says, “After a few years she will be able to talk and have intelligent conversations. She will learn, behave and be cute, and then she will grow up into an adult. But it is important to understand that MADA will not be human. She will be an artificially intelligent being, a brand new species. However, we’re going to bring her up with plenty of love and caring, so hopefully she will be respected at the very least and a beloved friend at the very best.”
For more information about Peter Bock’s work, visit his website at http://www.seas.gwu.edu/~pbock/