Data

MIT CSAIL researchers have developed a new air safety system, called Air-Guardian, that is designed to serve as a “proactive co-pilot, enhancing safety during critical moments of flight,” reports Jace Dela Cruz for Tech Times. 

Curtis Northcutt SM '17, PhD '21, Jonas Mueller PhD '18, and Anish Athalye SB '17, SM '17, PhD '23 have co-founded Cleanlab, a startup aimed at fixing data problems in AI models, reports Alex Konrad for Forbes. “The reality is that every single solution that’s data-driven — and the world has never been more data-driven — is going to be affected by the quality of the data,” says Northcutt.

Writing for Nature, graduate student Jelle van der Hilst offers advice on determining whether the data resulting from an experiment is meaningful and useful. “Although in research it is crucial that you don’t fully trust your data until it has been triple-proven and peer-reviewed,” writes van der Hilst, “we do have to gain some operational confidence in our methods and results. Otherwise, crippled by self-doubt, we’d never bring any new research into the world.”

A new study by MIT researchers demonstrates how “machine-learning systems designed to spot someone breaking a policy rule—a dress code, for example—will be harsher or more lenient depending on minuscule-seeming differences in how humans annotated data that were used to train the system,” reports Ananya for Scientific American. “This is an important warning for a field where datasets are often used without close examination of labeling practices, and [it] underscores the need for caution in automated decision systems—particularly in contexts where compliance with societal rules is essential,” says Prof. Marzyeh Ghassemi.

Forbes reporter Rob Toews spotlights Prof. Daniela Rus, director of CSAIL, and research affiliate Ramin Hasani and their work with liquid neural networks. “The ‘liquid’ in the name refers to the fact that the model’s weights are probabilistic rather than constant, allowing them to vary fluidly depending on the inputs the model is exposed to,” writes Toews.

Principal Research Scientist Kalyan Veeramachaneni speaks with Fast Company reporter Sam Becker about his work in developing the Synthetic Data Vault, which is helpful for creating synthetic data sets, reports Sam Becker for Fast Company. “Fake data is randomly generated,” says Veeramachaneni. “While synthetic data is trying to create data from a machine learning model that looks very realistic.”

Studies by researchers at MIT have found “that shifting to electric vehicles delivers a 30% to 50% reduction in greenhouse gas emissions over combustion vehicles,” reports Tom Krisher for Associated Press. According to Prof. Jessika Trancik, “electric vehicles are cleaner over their lifetimes, even after taking into account the pollution caused by the mining of metals for batteries,” writes Krisher.

Prof. Daniela Rus, director of CSAIL, writes for Forbes about Prof. Dina Katabi’s work using insights from wireless systems to help glean information about patient health. “Incorporating continuous time data collection in healthcare using ambient WiFi detectable by machine learning promises an era where early and accurate diagnosis becomes the norm rather than the exception,” writes Rus.

Prof. Daniela Rus, director of CSAIL, emphasizes the central role universities play in fostering innovation and the importance of ensuring universities have the computing resources necessary to help tackle major global challenges. Rus writes, “academia needs a large-scale research cloud that allows researchers to efficiently share resources” to address hot-button issues like generative AI. “It would provide an integrated platform for large-scale data management, encourage collaborative studies across research organizations, and offer access to cutting-edge technologies, while ensuring cost efficiency,” Rus explains.

In an article for Forbes, research affiliate John Werner spotlights Prof. Dina Katabi and her work showcasing how AI can boost the capabilities of clinical data. “We are going to collect data, clinical data from patients continuously in their homes, track the symptoms, the evolution of those symptoms, and process this data with machine learning so that we can get insights before problems occur,” says Katabi.

Writing for Times Higher Ed, Prof. Carlo Ratti makes the case that in the wake of the U.S. Supreme Court’s ruling on affirmative action, big data and analytics could “help admissions officers quantitatively capture the kinds of disadvantages applicants face and the kinds of diversity they may represent.”

Neil Thompson, director of the FutureTech research project at MIT CSAIL and a principal investigator MIT’s Initiative on the Digital Economy, speaks with Politico reporter Mohar Chatterjee about generative AI, the pace of computer progress and the need for the U.S. to invest more in developing the future of computing. “We need to make sure we have good secure factories that can produce cutting-edge semiconductors,” says Thompson. “The CHIPS Act covers that. And people are starting to invest in some of these post-CMOS technologies — but it just needs to be much more. These are incredibly important technologies.”

Prof. Marzyeh Ghassemi speaks with NPR host Kate Wells about a decision by the National Eating Disorders Associations to replace their helpline with a chatbot. “I think it's very alienating to have an interactive system present you with irrelevant or what can feel like tangential information,” says Ghassemi.

Researchers from MIT have found that Twitter “bot detection tools can rely on funky, flawed data sets that replicate mistakes made within one another, rather than trying to accurately identify bots,” reports Chris Stokel-Walker for Fast Company. “We realized that this was a systemic issue in the data sets that are commonly used for bot detection,” says postdoc Zachary Schutzman.

Prof. Marzyeh Ghassemi speaks with NPR host Emily Kwong and correspondent Geoff Brumfiel about how artificial intelligence could impact medicine. “When you take state-of-the-art machine-learning methods and systems and then evaluate them on different patient groups, they do not perform equally,” says Ghassemi.