This AI will be 10 years ahead of science – if we allow it


The survey committee, which received input from several small panels, considered a lot of information to develop research methods. Even if the Academies don’t release the final recommendation of the NASA committee in a few more weeks, scientists are eager to know which of their questions will work on it, and what will be left.

“The Decadal Survey has really helped NASA judge how they can lead the future in human space discovery, so it’s important that they know very well,” he said. Brant Robertson, a professor of astronomy and astrophysics at UC Santa Cruz.

A group of researchers want to use artificial intelligence to make this process faster. Their proposal is not for a specific mission or line of inquiry; rather, they say, their AI can help scientists make difficult decisions about which other proposals to prioritize.

The idea is that by training an AI to see areas of research growing or declining rapidly, the tool can make it easier for survey committees and panels to decide what needs to be done. on the list.

“What we want is to have a system that can do a lot of the work that the Decadal Survey does, and let the scientists working on the Decadal Survey do what they can best,” he said. Harley Thronson, a retired senior scientist at NASA’s Goddard Space Flight Center and lead author at the suggestion.

Even if the members of each committee are selected for their expertise in their respective field, it is impossible for each member to understand the nuance of each science theme. The number of astrophysics publications is increasing 5% each year, according to the authors. That’s a lot for anyone to process.

That’s where Thronson’s AI comes into play.

Construction lasted more than a year, but in the end, Thronson’s team was able to train it in more than 400,000 studies published in the decade leading up to the Astro2010 survey. They also taught AI to sift through thousands of abstracts to identify areas that are at the bottom and have a lot of impact from two- and three-word subject matter such as “planetary system” or “extrasolar planet. ”

Consistent with the white paper of the researchers, AI has successfully “revived” six popular research themes of the past 10 years, including the rise of meteorological exoplanet research and the observation of galaxies.

“A challenging aspect of artificial intelligence is that sometimes they can predict, or detect, or analyze things that completely shock people,” Thronson said. “And we see it all the time.”

Thronson and his colleagues envisioned the steering committee to use their AI to help review and summarize the many texts that needed to be scrutinized by the panel, leaving the human experts to make the final call.

Their research is not the first to attempt to use AI to analyze and shape the scientific literature. Other AIs are already in use to help scientists review as well work with their colleagues.

But can it be trusted in a task as important and influential as the Decadal Survey?



Source link

admin

Leave a Reply

Your email address will not be published. Required fields are marked *