News

The Hopkins-led team demonstrated that an AI model trained solely on synthetic tumor data works as well as models trained on real tumors.

A Johns Hopkins-led team found that chatbots reinforce our biases, providing insight into how AI could widen the public divide on controversial issues.

Johns Hopkins researchers compare GPT models to “sloppy paralegals.”

Johns Hopkins researchers find that large language models can handle multiple interpretations of the same sentence—but only if they’re told to.

Smartphone videos are often the first to capture news events. Johns Hopkins researchers are developing a tool to make that footage more searchable and contextualized.

Johns Hopkins researchers investigate how machine learning classifiers can be made more resistant to adversarial attacks on their input.

To combat the machine learning phenomenon known as “shortcut learning,” researchers from Johns Hopkins and the FDA have developed a data-screening method to identify potential hazardous shortcuts these models may take down the line.

Hopkins researchers have leveraged the synergy between medical professionals and artificial intelligence algorithms to create the largest annotated multi-organ dataset to date.

Researchers release a new algorithm that promises to help restructure the human reference genome into a more powerful—and inclusive—graph-based representation.

Data gathered using software developed by Johns Hopkins University computer scientists will have “huge implications” for understanding human health and evolution.

Johns Hopkins and Columbia University computer scientists teamed up to combat the inaccurate correlations that artificial intelligence and machine learning models learn from text data.

Johns Hopkins computer scientists demonstrate that yes-or-no questions can aid in the usability of semantic parsing models.