A new study by Google is aiming to teach powerful large language models (LLMs) how to reason better with graph information. In computer science, the term 'graph' refers to the connections between entities - with nodes being the objects and edges being the links that signify their relationships. This type of information, which is inherent…
Researchers at MIT, Harvard, and the University of Washington have shunned traditional reinforcement learning approaches, using crowdsourced feedback to teach artificial intelligence (AI) new skills instead. Traditional methods to teach AI tasks often required a reward function, which was updated and managed by a human expert. This limited scalability and was often time-consuming, particularly if…
During the kickoff event of MIT’s Generative AI Week, the “Generative AI: Shaping the Future” symposium, Rodney Brooks, co-founder of iRobot, cautioned attendees about the dangers of overestimating the capabilities of generative AI technology. Brooks, also a professor emeritus at MIT and former director of the Computer Science and Artificial Intelligence Laboratory (CSAIL), warned that…
Time-series analysis is indispensable within numerous fields such as healthcare, finance, and environmental monitoring. However, the diversity of time series data, marked by differing lengths, dimensions, and task requirements, brings about significant challenges. In the past, dealing with these datasets necessitated the creation of specific models for each individual analysis need, which was effective but…
In the modern digital age, individuals often interact with technology through software interfaces. Even with advancements towards user-friendly designs, many still struggle with the complexity of repetitive tasks. This creates an obstacle to efficiency and inclusivity within the digital workspace, underlining the necessity for innovative solutions to streamline these interactions, thereby making technology more intuitive…
Five MIT researchers—Michael Birnbaum, Regina Barzilay, Brandon DeKosky, Seychelle Vos, and Ömer Yilmaz—are part of winning teams for Cancer Grand Challenges 2024. Each team, made up of international, interdisciplinary cancer researchers, will receive $25 million over five years.
Associate Professor of Biological Engineering Michael Birnbaum is heading Team MATCHMAKERS, comprised of co-investigators Regina Barzilay (Engineering Distinguished…
MIT researchers have developed an algorithm called FeatUp that enables computer vision algorithms to capture both high-level details and fine-grained minutiae of a scene simultaneously. Modern computer vision algorithms, like human beings, can only recall the broad details of a scene while the more nuanced specifics are often lost. To understand an image, they break…
Recent developments in Artificial Intelligence (AI), particularly in Generative AI, have proven the capacities of Large Language Models (LLMs) to generate human-like text in response to prompts. These models are proficient in tasks such as answering questions, summarizing long paragraphs, and more. However, even provided with reference materials, they can generate errors which could have…
The Sparse Mixture of Experts (SMoEs) has become popular as a method of scaling models, particularly in memory-restricted environments. They are crucial to the Switch Transformer and Universal Transformers, providing efficient training and inference. However, some limitations exist with current implementations of SMoEs, such as a lack of GPU parallelism and complications related to tensor…
The development of effective large language models (LLMs) remains a complex problem in the realm of artificial intelligence due to the challenge of balancing size and computational efficiency. Minimizing these issue, a strategy called Additive Quantization for Language Models (AQLM) has been introduced by researchers from institutions such as HSE University, Yandex Research, Skoltech, IST…
Stanford University researchers are pushing the boundaries of artificial intelligence (AI) with the introduction of "pyvene," an innovative, open-source Python library designed to advance intervention-based research on machine learning models. As AI technology evolves, so does the need to refine and understand these advancement's underlying processes. Pyvene is an answer to this demand, propelling forward…
Machine learning (ML) workflows have become increasingly complex and extensive, prompting a need for innovative optimization approaches. These workflows, vital for many organizations, require vast resources and time, driving up operational costs as they adjust to various data infrastructures. Handling these workflows involved dealing with a multitude of different workflow engines, each with their own…