Could artificial intelligence (AI) take over the world? The question captured media attention recently when Bill Gates, Stephen Hawking and Elon Musk joined the debate.
Gates said he is “concerned about super intelligence,” Hawking warned that “the development of full artificial intelligence could spell the end of the human race,” and Musk described AI as “our biggest existential threat.”
Tom Dietterich, president of the Association for the Advancement of Artificial Intelligence and distinguished professor of computer science at Oregon State University, has been at the forefront of the media discussion, giving an academic perspective on the issue.
“We will not let AI run the world … It’s a technology that should be used to enhance our humanity,” Dietterich said in an interview for NPR’s “On Point.”
Dietterich points out that computers cannot become fully autonomous on their own; they need to be programmed. And the real danger would be in depending on AI too much when it could be vulnerable to cyberattacks, contain software bugs or have user interface issues.
“Before we put computers in control of high-stakes decisions, our software systems must be carefully validated to ensure that these problems do not arise,” he said in an interview for Digital Trends.
Rather than a threat, Dietterich points out that AI is a powerful tool that can advance science and impact how we live. And it’s already happening.
When Facebook automatically recognizes a person in your photo, or when your credit card company detects a fraudulent use of your card — machine learning is at work. Machine learning is a type of AI that provides a way for computers to learn with experience.
“Over the last 30 years, AI has moved beyond carefully engineered laboratory settings. Now that so much data is being collected from smart phones and the web, we can use machine learning techniques to teach AI systems how to robustly understand speech, recognize objects in images, translate languages and even play Jeopardy!” Dietterich said.
Oregon State expertise in wild data
The faculty members in the AI and machine learning group at Oregon State share their expertise with colleagues across the globe in a broad range of areas including biology, computer vision, ecology, language processing, social media and medical diagnosis. Their collaborators span 13 departments on campus, 42 universities around the world, and 16 companies and research institutions.
Google, Yahoo and Facebook
Companies like Google, Yahoo and Facebook are very interested in machine learning techniques to help improve their software. In fact, Professor Liang Huang of Oregon State just received the Yahoo Faculty Research and Engagement Program Award for his research on resolving ambiguity in natural language. For example, the query, “How can I book Paris Hilton?” is ambiguous as to whether the question is about the person or the hotel. Huang also received two Google Faculty Research Awards and his research is being used in Google Translate. Professor Scott Sanner has worked with Facebook data to design a system to improve recommendations for book purchases and movie rentals.
Creating a system that can learn new knowledge from written texts is the goal of research by professors Prasad Tadepalli, Xiaoli Fern and Dietterich. Applications include assisting a doctor in making a diagnosis or helping a police detective piece together evidence of a crime.
Machine learning has even entered the world of sports. Faculty members Alan Fern and Sinisa Todorovic have combined the fields of machine learning and computer vision to teach computers how to watch football games. It’s more useful than it might seem at first. In fact, they are collaborating with a company in the area of sports video storage and organization to create an automated system for annotating and organizing video. But the technology could also be useful in hospitals or nursing homes to monitor patients and see who needs care.
The advancements in science are rapidly growing. For example, bioinformatics is an emerging field that uses computer science, including machine learning, to analyze and interpret biological data. The applications are broad, such as Professor Stephen Ramsey’s research on factors that contribute to heart disease, and Professor David Hendrix’s research in sequencing the hops genome. Careers in bioinformatics are exploding, according to an article in Science, and Ramsey and Hendrix are part of educating the next generation in this area of specialization as affiliated faculty for the Center of Genome Research and Biocomputing.
Rebecca Hutchinson, who joined the faculty this year, will also be teaching in the area of bioinformatics as part of the center. Her research focuses on ecology, such as a recent project for which she developed a new method to help scientists understand habitats of rare or endangered species.
“What I find really inspiring is that I can use computer science to have an impact in ecology — to help protect species that are threatened by climate change,” Hutchinson said.
Several other researchers in the AI and machine learning group contribute to studies in ecology. Weng-Keen Wong works with Cornell Lab of Ornithology to model the skill level of citizen-scientist birdwatchers who contribute data to eBird, a research program that collects observations of bird species around the world. Dietterich is using machine learning to analyze the eBird data and model the migration of different bird species. Raviv Raich and Xiaoli Fern have been working with Oregon State ornithologists to classify bird songs from recordings taken in the wild, with the goal of tracking which species are present at a site.
Raich’s research also impacts medicine. For example, he helped cancer researchers improve the analysis of blood samples to detect different types of blood cancer.
These are just a few examples of how AI researchers at Oregon State are helping to advance science and technology to improve our world, but the applications are endless.
What does the future hold?
The area of AI is still growing. In fact, Google recently announced that they are reorienting their entire company around machine learning technologies.
At Oregon State, future research will address the issue of AI systems involved in more high-risk decisions. Tom Dietterich and Alan Fern are developing AI systems that know their own limits — that ask for help from humans when a situation is unfamiliar and their learned knowledge is insufficient to make a decision. As AI systems are involved in more high-risk decisions, it will be important that we can trust them to act only when they are competent.
“Machine learning has become the most important and most practical way of creating AI systems. At Oregon State, we hope to lead the way toward a future in which AI technology can enhance our lives both by automating boring things and by helping people address the most difficult and complex problems that we face as a society,” Dietterich said.