AI is getting bigger and smarter in Seattle. The Allen Institute for Artificial Intelligence, a sister company to Paul Allen’s Institute for Brain Science, plans to hire 25 people in the next year as it prepares to take its Aristo technology to the eighth grade, moving on from teaching it fourth-grade science.
Artificial intelligence has been all over the news lately, and this time not solely over fears that machines may eliminate people’s jobs. AlphaGo, an AI program developed by DeepMind, a Google company, beat one of the world’s foremost Go champions in March. The machine’s victory over Lee Sedol at the numerically broad game shocked many people who believed AI was years away from such a feat.
But it didn’t happen overnight. The program may be fairly new, but the underlying research has been developing for decades. “AlphaGo’s overnight success was kind of 50 years in the making,” Etzioni said. In Seattle, Etzioni believes the team at Allen’s AI2 is attempting a much bigger challenge by teaching a computer to be a mastermind in all disciplines of science.
AlphaGo can train by playing the board game 24 hours a day, seven days a week. AI2’s Aristo cannot. There are only a finite number of scientific test questions in existence, and the AI2’s Aristo technology must learn using them.
The machine also learns from researchers across the world. AI2 recently wrapped up a global competition in which teams ran their own programs through an eighth-grade science test to see which could get the top score. The Allen Institute awarded the $50,000 grand prize to Chaim Linhart, a senior researcher at TaKaDu, an environmental-services company in Israel.
Next year, the stakes will be much higher. AI2 plans to offer a $1 million prize if any team can beat a certain threshold, the exact percentage has not yet been determined, followed by smaller monetary prizes for runners-up. “It helps to raise awareness and focus on important problems,” Etzioni said. “To win, teams have to publish their code.”
Eventually, AI2 wants to have a smart enough system that it can act as an assistant in different fields of science. For example, it could scan hundreds of thousands of medical-journal articles in an instant to give doctors immediate answers to their questions.
Early versions of that project, called Semantic Scholar, have launched for computer science and it is expected to launch for neuroscience this year. Etzioni’s hope is that in about 20 years, the machine will be able to go even further, gaining the ability to form its own hypotheses from reading scholarly journals. He calls it a “scientists’ apprentice.”
“People who are running businesses, for the mundane but nonetheless insurmountable problems presented by big data, are turning to machine learning to make sense of this world of noise,” said Matt Bencke, CEO of Seattle tech startup Spare5. “…The incredible headline-gathering events are in some ways getting us distracted from the more mundane reality of how AI and machine learning are impacting our every day lives already.”