A new artificial intelligence-based search engine may be an important step towards search results that are the result of a machine that “thinks” through answers rather than just indexing information.
Semantic Scholar is an AI-based search engine created by the non-profit Allen Institute for Artificial Intelligence, which is based in San Diego. The big thinkers there plan to trump their main competitor Google Scholar by upping their total research article database from 3 million at launch to over 10 million early next year.
“This is a game changer,” says Andrew Huberman, a neurobiologist at Stanford University. “It leads you through what is otherwise a pretty dense jungle of information.”
Sematic Scholar, which currently caters mainly to neurology and computer sciences, uses machine learning, semantic analysis and machine vision, a process that has been used primarily in robot guidance and in industrial inspections, to identify connections between relevant research papers. Google Scholar, meanwhile, works like a general search engine and simply searches for key words.
But Semantic Scholar has a long way to go. It has a meager database compared to Google Scholar, which indexes more than 200-million articles and can even access information that has been locked behind pay walls. Most researchers will probably end up using both, at least in the short term.
“The one I still use the most is Google Scholar,” says Jose Manuel Gómez-Pérez, who works on semantic searching for the software company Expert System in Madrid. “But there is a lot of potential here.”
Some where is this all going? Those who have been binge-watching the new HBO hit “WestWorld”, which is set in a futurisitc amusement park populated by synthetic androids, may get a little creeped out by the end game here. The goal for the Allen Institute for Artificial Intelligence is to create an AI system that will be able to read and understand scientific text and create its own hypothesis. This obviously has huge and wide-ranging implications and we trust that the answer to the question “What is the best Thai restaurant in Cleveland?” will not be “Kill all humans”.
Sematic Scholar is not the only search engine of its kind. Microsoft quietly released its own AI scholar research tool, called Microsoft Academic, in May of this year. Microsoft Academic is the replacement for an older product called Microsoft Academic Search, which was shelved in 2012.
At least to date, this field is chracterized by its cooperation as much as competition. Semantic Scholar, Microsoft Academic, and a few smaller AI search engines shares search algorithms so they as to hasten the development of current AI capabilities. Microsoft Academic uses Bing database so it has access to over 160 million publications and is also toying with the idea of making the engine user customizable.
“The Microsoft Academic phoenix is undeniably growing wings,” says Anne-Wil Harzing, who studies science metrics at Middlesex University in the UK.
But one expert says the machine learning aspect of AI search engines could prove problematic. Jeff Clune, a computer scientist at the University of Wyoming in Laramie told Science recently that he liked Semantic Scholar and found it to be a fun and useful service, but said he was at the same time worried that it was a “black box” of information.
“Will people understand where the numbers are coming from?” he asked.