Could a sophisticated algorithm be the future of science? One innovative economist thinks so.
Phil Parker, who holds a doctorate in business economics from the Wharton School, has built an algorithm that auto-writes books. Now he's taking that model and applying it to loftier goals than simply penning periodicals: namely, medicine and forensics. Working with professors and researchers at NYU, Parker is trying to decode complex genetic structures and find cures for diseases. And he's doing it with the help of man's real best friend: technology.
Parker's recipe is a complex computer program that mimics formulaic writing. "What I have done is I've collected information," he says, saving researchers valuable time they would have spent in the lab. Parker is confident that all that work of collecting, surveying and estimating data can be automated, "meaning computer programs can accelerate discovery in the sciences faster than scientists can, if the computer is trained to behave like one."
Parker says the average researcher spends 40 to 50% of their time reading other people's work. That, coupled with relatively short careers, leads to burnout, and stymies discovery. He says his innovation helps free up time to focus on discovering new data.
"I'm excited," Parker says. "[It could help] some diseases in the world of agriculture and plant diseases, especially ... If I can find the formulated methodology behind the science, ultimately, a computer might write that."
Parker explains the value of the algorithm as a tool to map different species, which can then be cross-referenced and used as a matrix of data to map and compare different flora and fauna.
"In essence, you have a computer program guessing what scientists will discover and then writing a report of findings of what you see in nature," he says. "The application of these kinds of things can open windows to subjects that people didn't know were there ... Right now, there are 400,000 plant species and sub-species out there in the world. Research and agriculture has only been able to cover about 1% of those species. But some of [them] could be very, very useful."
How useful remains to be seen. Parker says he plans on making his findings public knowledge, and is already publishing to totoagriculture.org, a site that shares world-agriculture information. Still, he's tentative, though results have been positive so far.
"We're posting the facts that might be an engine for a formula," he says. "I want to be very careful with the academics. We have already done it for chromosome counts; predict the chromosome counts within plant species. The results were pretty encouraging."
Those results include naming new species. "It's [the algorithm is] predicting the language that people would use."
Parker's been at this for years. His formula, originally used for printing, is able to churn out entire books in minutes. It's similar to the work being done by Narrative Science and StatSheet, except those companies are known for short form auto-writing for newspapers. Parker's work is much longer, focusing on obscure non-fiction and even poetry.
It's not creative writing, though, and Parker isn't interested in introspection, exploring emotion or storytelling. He's interested in exploiting reproducible patterns — that's how his algorithm can find, collect and "write" so quickly. And how he can apply that model to other disciplines, like science.
Parker's method seems to be a success; indeed, his ICON Group International, Inc., has auto-written so many books that Parker has lost count. But this isn't the holy grail of literature, he insists. Instead, he says, his work is a play on mechanizing processes to create a simple formula. And he thinks that "finding new knowledge structures within data" stretches far beyond print.
What remains to be seen is whether his new data model will revolutionize the world of deductive reasoning and, beyond that, medicine, forensics, or other scientific disciplines.
Image courtesy of iStockphoto, dra_schwartz
No hay comentarios:
Publicar un comentario