NYU researcher wins $200K to bring AI tool to support under-resourced newsrooms

Mona Sloane, faculty at NYU Tandon and Senior Research Scientist at the NYU Center for Responsible AI (R/AI), and Hilke Schellmann, professor of journalism at NYU’s Graduate School of Arts and Science, has been awarded a $200,000 grant from the Patrick J. McGovern Foundation to bring an innovative AI tool to under-resourced newsrooms to significantly scale up their investigative capacity and democratize access to FOIA records.

The project will integrate the NYU-developed Gumshoe prototype — a Natural Language Processing Tool that identifies relevant and irrelevant sections in large text corpora — to help journalists effectively comb through thousands of Freedom of Information Request (FOIA) releases and other document sets. NYU will collaborate with MuckRock, an open-source journalism platform used by tens of thousands of journalists across 4,000 newsrooms to help request, analyze, and publish public documents. The effort will unlock decades of valuable information, data, and history contained in federal government records released under the Freedom of Information Act.

“Effective use of the Freedom of Information Act is a key public tool to encourage transparent, accountable government activity,” said Vilas Dhar, president of the Patrick J. McGovern Foundation. “This innovative application of AI by New York University and MuckRock makes important records accessible to under-resourced newsrooms that are shining a light on our country’s most important social justice stories.”

Gumshoe is an innovative NLP-based text analyzing tool that was developed in a collaboration between Schellmann and Sloane and a team of graduate students at the NYU Center for Data Science, under the supervision of Julia Stoyanovich, professor of computer science at NYU Tandon and director of R/AI. Gumshoe sorts and ranks content in FOIA requests for journalists. Because it is based on NLP technology, Gumshoe understands the meaning of words in context and “learns” about the individual investigation of the journalist over time - like a shoe that fits better with every walk.

“We built this tool because as a freelance journalist I was frustrated with combing through large Freedom of Information requests. I knew other journalists had the same problem and I am so delighted the Patrick J. McGovern Foundation is making it possible for us to bring this AI tool to small newsrooms across America that really need this,” said Schellmann. “Journalists in underserved newsrooms need this AI tool and the training MuckRock will provide to hold the powerful accountable and help strengthen democracy.” 

Through integration with MuckRock’s open-source DocumentCloud platform used by tens of thousands of journalists around the world, Gumshoe will help journalists on a wide variety of critical stories.

“Combining AI innovation, social research, and investigative journalism is extremely important for advancing the public interest in technology”, said Sloane. “We must work together to advance innovation in the right direction and level the playing field of journalistic practice across the U.S. and beyond.”

“We’ve seen a widening disparity between the newsrooms that get access to the advanced tools and support needed to tackle the complex analysis and those that struggle to fund basic resources and training,” said Michael Morisy, MuckRock’s co-founder and chief executive. “I’m grateful for the McGovern Foundation’s support of this collaboration with Hilke and Mona, which will help us start to address that gap while supporting local accountability reporting.”

Over the past year, MuckRock has been expanding the way that it collaborates with local organizations, particularly those serving communities of color and innovative new outlets helping fill the voids left through news deserts where legacy media no longer exist. This includes extensive editorial partnerships helping dig into the impact of the COVID-19 pandemic as well as a revamped FOIA training and coaching program. These expanded collaborations, which have resulted in tens of thousands of newly released documents, provide the perfect opportunity to rapidly scale the impact of Gumshoe by integrating with MuckRock’s platform while further expanding editorial and investigative support to newsrooms around the country. 

Along with access to the tool, the collaboration will offer select newsrooms additional training, editorial guidance, and financial support in the form of microgrants and training stipends to ensure that key stories in communities around the country can be told while journalists from a wide variety of backgrounds have an opportunity to put advanced technology to work in their reporting. These close collaborations will be leveraged to iteratively improve both Gumshoe and MuckRock’s approach to building intuitive and impactful transparency tools.

ThePatrick J. McGovern Foundation is global 21st-century philanthropy bridging the frontiers of artificial intelligence, data science, and social impact to create a thriving, equitable, and sustainable future for all. The Foundation’s work focuses on bringing together academia, practitioners, and civil society to pursue the potential of AI and data science to address some of the world’s most urgent challenges.

Zooming into the black hole at the centre of our galaxy

This zoom video sequence starts with a broad view of the Milky Way.We then dive into the dusty central region to take a much closer look.There, a swarm of stars orbit around an invisible object:a supermassive black hole, 4.3 million times that of the Sun.As we get closer to it, we see these stars, as observed by the NACO instrument on ESO’s Very Large Telescope (the last observation being from...
...

Read more

Watch Stars Move Around our Galaxy’s Central Black Hole (ESOcast 248 Light)

o learn more about the Milky Way's supermassive black hole, Sgr A , scientists zoomed in towards our galaxy's centre with the help of ESO's Very Large Telescope Interferometer to watch how stars move around Sgr A .This video summarises what they found.
...

Read more

Salk Stites lab's computational methods could contribute to improving treatment options for cancer patients

Patients with colorectal cancer were among the first to receive targeted therapies. These drugs aim to block the cancer-causing proteins that trigger out-of-control cell growth while sparing healthy tissues. But some patients are not eligible for these treatments because they have cancer-promoting mutations that are believed to cause resistance to these drugs. From left: Edward Stites and Thomas McFall

Now, Salk Assistant Professor and physician-scientist Edward Stites have used supercomputer modeling and cell studies to discover that more patients may be helped by a common class of targeted therapies than previously thought. The findings were published December 14, 2021, in Cell Reports.

“Colorectal cancer patients who have tried all of the standard treatment options but still seen their cancer progress need new options. Our study suggests that one already available targeted therapy could benefit up to 12,000 additional colon cancer patients every year,” says Stites, the paper’s senior author. “Our findings are pre-clinical, and we hope this research will motivate clinicians to develop clinical trials that further examine our results.”

Stites was interested in examining drugs that target a protein called EGFR (epidermal growth factor receptor). EGFR is known to drive a subset of many different cancer types, including lung cancer and colorectal cancer.

In 2004, the US Food and Drug Administration (FDA) approved cetuximab, the first drug to block EGFR activity in colorectal cancer. Since then, other drugs that target EGFR also have received approval. But from the early development of these drugs, doctors believed that patients with a mutation in any one of the families of proteins known as RAS would not respond to EGFR drugs. Therefore, whenever molecular testing of a patient’s tumor revealed a RAS mutation, the patient was not offered these targeted therapies.

Earlier research by the Stites lab suggested that not all RAS mutations act in the same manner, and was able to explain one well-known, but poorly understood, the exception to the rule. In the new study, the team combined computational and experimental approaches to using this new explanation to find more RAS mutations that should not cause resistance to the EGFR drugs.

The researchers used cells from cancers that were identical except for specific RAS mutations. This allowed them to compare how each specific mutation influenced the response to EGFR-inhibiting drugs. They found that some RAS mutations did not prevent the drugs from working. These experiments also allowed them to validate their computational studies, which helps establish how new computational methods could contribute to improving treatment options for cancer patients.

The investigators also examined how well different RAS mutants bound to another protein, called NF1. Stites’ previous mathematical models hinted that NF1 could play a key role in the cells’ response to targeted drugs. In their new studies, they revealed that the RAS mutants that do not bind NF1 well retain sensitivity to EGFR drugs, while the RAS mutants that bind NF1 well are resistant to EGFR drugs. This relationship to EGFR drugs was not originally apparent, but the computational modeling was able to uncover it from within the available and varied data.

Ultimately, the investigators identified 10 distinct RAS mutations that do not preclude the use of EGFR inhibitors. Many of the drugs that would work for these mutations are already approved by the FDA for other uses, which means that doctors could start prescribing them for their patients “off label” even before clinical trials are conducted.

Stites, who holds the Hearst Foundation Developmental Chair, stresses that this study also helps to validate the mathematical and computational methods developed by his team. “Models can solve scientific problems that traditional methods cannot,” he says. “We hope that future clinical trials will help identify the magnitude of benefit as well as whether all the RAS mutations we identified are equally sensitive to the EGFR-inhibiting drugs and how other mutations in addition to RAS may influence the strength of the response.”

The first author of the paper is Thomas McFall, a former postdoctoral fellow at Salk who is now at the Medical College of Wisconsin.

VLTI uses machine learning to help find stars moving around the Milky Way’s supermassive black hole

The European Southern Observatory’s Very Large Telescope Interferometer (ESO’s VLTI) has obtained the deepest and sharpest images to date of the region around the supermassive black hole at the center of our galaxy. The new images zoom in 20 times more than what was possible before the VLTI and have helped astronomers find a never-before-seen star close to the black hole. By tracking the orbits of stars at the center of our Milky Way, the team has made the most precise measurement yet of the black hole’s mass. These annotated images, obtained with the GRAVITY instrument on ESO’s Very Large Telescope Interferometer (VLTI) between March and July 2021, show stars orbiting very close to Sgr A*, the supermassive black hole at the heart of the Milky Way. One of these stars, named S29, was observed as it was making its closest approach to the black hole at 13 billion kilometres, just 90 times the distance between the Sun and Earth. Another star, named S300, was detected for the first time in the new VLTI observations. To obtain the new images, the astronomers used a machine-learning technique, called Information Field Theory. They made a model of how the real sources may look, simulated how GRAVITY would see them, and compared this simulation with GRAVITY observations. This allowed them to find and track stars around Sagittarius A* with unparalleled depth and accuracy. Credit: ESO/GRAVITY collaboration

“We want to learn more about the black hole at the center of the Milky Way, Sagittarius A*: How massive is it exactly? Does it rotate? Do stars around it behave exactly as we expect from Einstein’s general theory of relativity? The best way to answer these questions is to follow stars on orbits close to the supermassive black hole. And here we demonstrate that we can do that to a higher precision than ever before,” explains Reinhard Genzel, a director at the Max Planck Institute for Extraterrestrial Physics (MPE) in Garching, Germany who was awarded a Nobel Prize in 2020 for Sagittarius A* research. Genzel and his team’s latest results, which expand on their three-decade-long study of stars orbiting the Milky Way's supermassive black hole, are published today in two papers in Astronomy & Astrophysics.

On a quest to find even more stars close to the black hole, the team, known as the GRAVITY collaboration, developed a new analysis technique that has allowed them to obtain the deepest and sharpest images yet of our Galactic Centre. “The VLTI gives us this incredible spatial resolution and with the new images, we reach deeper than ever before. We are stunned by their amount of detail, and by the action and number of stars they reveal around the black hole,” explains Julia Stadler, a researcher at the Max Planck Institute for Astrophysics in Garching who led the team’s imaging efforts during her time at MPE. Remarkably, they found a star, called S300, which had not been seen previously, showing how powerful this method is when it comes to spotting very faint objects close to Sagittarius A*.

With their latest observations, conducted between March and July 2021, the team focused on making precise measurements of stars as they approached the black hole. This includes the record-holder star S29, which made its nearest approach to the black hole in late May 2021. It passed it at a distance of just 13 billion kilometers, about 90 times the Sun-Earth distance, at the stunning speed of 8740 kilometers per second. No other star has ever been observed to pass that close to, or travel that fast around, the black hole. 

{media id=266,layout=solo}

The team’s measurements and images were made possible thanks to GRAVITY, a unique instrument that the collaboration developed for ESO’s VLTI, located in Chile. GRAVITY combines the light of all four 8.2-meter telescopes of ESO’s Very Large Telescope (VLT) using a technique called interferometry. This technique is complex, “but in the end you arrive at images 20 times sharper than those from the individual VLT telescopes alone, revealing the secrets of the Galactic Centre,” says Frank Eisenhauer from MPE, principal investigator of GRAVITY.

Following stars on close orbits around Sagittarius A* allows us to precisely probe the gravitational field around the closest massive black hole to Earth, to test General Relativity, and to determine the properties of the black hole,” explains Genzel. The new observations, combined with the team’s previous data, confirm that the stars follow paths exactly as predicted by General Relativity for objects moving around a black hole of mass 4.30 million times that of the Sun. This is the most precise estimate of the mass of the Milky Way’s central black hole to date. The researchers also managed to fine-tune the distance to Sagittarius A*, finding it to be 27 000 light-years away.

To obtain the new images, the astronomers used a machine-learning technique, called Information Field Theory. They made a model of how the real sources may look, simulated how GRAVITY would see them, and compared this simulation with GRAVITY observations. This allowed them to find and track stars around Sagittarius A* with unparalleled depth and accuracy. In addition to the GRAVITY observations, the team also used data from NACO and SINFONI, two former VLT instruments, as well as measurements from the Keck Observatory and NOIRLab’s Gemini Observatory in the US.

GRAVITY will be updated later this decade to GRAVITY+, which will also be installed on ESO’s VLTI and will push the sensitivity further to reveal fainter stars even closer to the black hole. The team aims to eventually find stars so close that their orbits would feel the gravitational effects caused by the black hole’s rotation. ESO’s upcoming Extremely Large Telescope (ELT), under construction in the Chilean Atacama Desert, will further allow the team to measure the velocity of these stars with very high precision. “With GRAVITY+’s and the ELT’s powers combined, we will be able to find out how fast the black hole spins,” says Eisenhauer. “Nobody has been able to do that so far.