TAPAS can estimate the classification performance of a neural network for unseen input datasets in fractions of a second, without training.
Highlights from IBM Research AI in 2018 in three key areas - advancing, scaling and trusting AI - and predictions about what's to come in 2019.
Scientists at our lab in Kenya are working with Hello Tractor, a mobile platform that enables farmers to access tractor services on demand.
IBM Research launches new AI Experiments hub featuring prototypes of tools and resources that will unleash the power of AI.
An approach to efficiently train deep learning models on cloud platforms when the training dataset consists of a large number of small files.
IBM scientists show, for the first time, successful training of deep neural networks using 8-bit floating point numbers while fully maintaining accuracy.
Bridging the gap between interpretability and performance by transferring information from a high-performing model to a simpler, interpretable model.
The Algebraic Gradient-based Solver (AGS), a novel solver for approximate marginal MAP inference, shows how ideas from planning can be used for inference.
Delta-encoder enables AI to classify an image from a new category with only a limited number of examples from that category.
IBM Research AI shares new ideas and research results at the Thirty-Second Conference on Neural Information Processing Systems in Montreal.
The Pockels effect allows changing the optical properties of certain materials through an electrical stimulus.
Can silica gel desiccant packs, often found in leather shoe boxes and electronics, be the answer to our energy needs for buildings?
IBM Research has made progress in coupling AI with speech and written text analysis to help clinicians better evaluate mental health.
AI approaches for parsing and understanding text require clean input, which in turn implies a necessary amount of pre-processing of raw data.
IBM scientists demonstrate the advantages of the Graph2Seq model for semantic parsing and natural language generation at EMNLP 2018.
Word Mover’s Embedding is an unsupervised framework for learning universal text embeddings that can be used for downstream tasks.
After a few meetings with the Industrialization Ministry the IBM lab in Kenya agreed to a joint collaboration focused on improving the country’s Ease of Doing Business ranking.
A new deep learning framework that detects glaucoma from optical coherence tomographic (OCT) imaging with an accuracy of 94 percent.
A novel 'show and tell' technique for training an AI agent to operate optimally in a given environment while aligning its behavior with societal values.
The Ocean Tensor Package is a comprehensive, flexible tensor-computation package designed for general-purpose use, both within and outside of deep learning.
This feat opens the door to using the nucleus as way to sense and control magnetism at the atomic scale.
IBM researchers explain how advanced technologies can help financial institutions in the Sensemaker series by Innotribe at the Sibos 2018 conference.
Scientists prove quantum advantage for certain mathematical problems: shallow quantum circuits are more powerful than their classical counterparts.
IBM researchers recently developed a cyber reasoning paradigm named Threat Intelligence Computing to facilitate cyber threat discovery.
IBM researchers propose guidelines for novel analog memory devices to enable fast, energy-efficient and accurate AI hardware accelerators.
Coronary Artery Disease (CAD) is a condition in which plaque forms on the walls of coronary arteries, causing them to narrow.
The promise of the Internet of Things: This integration should ideally map to a decentralized hardware and software platform.
On the anniversary of the MIT-IBM Watson AI Lab, we look back at the shared history of MIT and IBM and look forward to the future of AI.
A novel approach for AI dictionary expansion employs a lightweight neural language model coupled with tight human-in-the-loop supervision.
Bringing the power of Watson to farmers.
An expert-in-the-loop to working closely with AI technologies can save experts up to 80 percent of their time on time-intensive tasks.
At MICCAI 2018, IBM researchers present their latest findings in applying deep learning to overcome some technical challenges in medical image analysis.
A new approach to video scene detection that formulates the task as a generic optimization problem with a novel normalized cost function.
A new approach to spectral clustering using random binning enables this important machine learning technique to scale to large datasets.
New data science course for Kenyan students funded by KDD Impact Program.
Could you make a critical health policy decision using four-year old data?
The Corpus Conversion Service is an AI-based cloud system that can ingest PDF documents at scale and extract their content using machine learning models.
Announcing the winners of the IBM Qiskit Developer Challenge.
IBM RXN for Chemistry is the first free AI web service for predicting chemical reactions.
An AI system for semantic analysis of code could help scientists analyze complex data sets efficiently by identifying existing relevant code.
The FEED 2018 workshop applies theory-guided data science to tackle challenging problems in sustainability of food, water, and energy.
A new release of the Adversarial Robustness Toolbox provides a method for defending against poisoning and "backdoor" attacks in machine learning models.
Quantum computing promises to be faster than classical computing in some, but not all computational problems.
Scientists from IBM, UCSF, Stanford University, and the University of Haifa are engineering a new antiviral agent against viruses such as polio.
A new interface and detailed process for annotation of question answering datasets, focusing on the knowledge and reasoning required for complex questions.
Search the two million mobile games on the app store, and you won’t find another quite like Hello Quantum.
A new approach that uses unsupervised text style transfer to translate offensive language in social media into corresponding non-offensive sentences.
Scientists publish a new approach to phase change memory using only a single chemical element—antimony—in Nature Materials.
Can we develop a watermarking technique to assure developers that their hard work in building AI, such as deep learning models, can be protected?
A new technique to fabricate microelectronic devices that allows molecular integration on conventional silicon chips and by standard semiconductor methods.
Operating a commercial building’s cooling system using machine learning techniques and IoT data can help to reduce energy consumption and costs.
A capacitor-based cross-point array for analog neural networks offers potential orders of magnitude improvements in deep learning computations.
IBM scientists developed an artificial synaptic architecture, a significant step towards large-scale and energy efficient neuromorphic computing technology.
IBM scientists present a programmable AI core for multi-domain deep learning training and inference on chip at 2018 VLSI Circuits Symposium.
IBM Research is committed to AI technologies developed without bias, including facial recognition technology that's built and trained responsibly.
IBM scientists developed a natural language processing tool for collaborative decision-making. It augments the decision process using a virtual facilitator.
IBM Research partnered with Tencent in China to produce a highlight reel from the NBA Finals in only 20 seconds using an AI-based video editing system.
A stereo vision system inspired by the human vision system that uses exclusively spiking neural network computation and can run on neuromorphic hardware.
Working with real quantum computers just got easier for experts in chemistry, artificial intelligence and optimization.
Could tiny ink spots, the size of a pinhead, fight counterfeiting medical tests? We believe the answer is yes.
Analog memory for accurate, faster, lower power neural network training - a major step on the path to hardware accelerators for the next AI breakthroughs.
IBM Research AI will present multiple technical papers and host two workshops at CVPR 2018. Join us in Salt Lake City at booth #845.
Can we do fast cloud-based signal processing on a very large (5 TB) dataset of fast-rate TRAPPIST-1 measurements? My team's new research brings SETI to the cloud with the IBM...
Recently, I wrote about the state of our climate and outlined our understanding of the science. Now, let’s talk about the economic impact of these changes, some of which the...
An algorithm that can provide contrastive explanations for models such as deep neural networks to improve the explanations generated by AI technologies.
IBM Crypto Anchor Verifier couples innovations in AI and optical imaging for object authentication. Our first client GIA is using it to evaluate diamonds.
The first commercial IBM Q hub in Asia, Keio University, announced its own members: JSR, MUFG Bank, Mizuho Financial Group, and Mitsubishi Chemical.
A cloud-based automated machine learning approach to provide decision support for non-specialist physicians in treating brain disorders.
IBM Research is collaborating with industry and academic partners to investigate how IoT, AI and cloud technologies can advance home-based integrated care for persons with multiple chronic diseases.
AI researchers are building the next generation of conversational AI interfaces by applying conversation science to user experience (UX) design.
A principled approach for disentanglement of hidden factors from a large pool of unlabeled observations via unsupervised learning.
Emory University, Far Eastern Federal University and Delft University have received grants to use the IBM World Community Grid to study research problems in climate and environmental sciences.
The recent surge of interest in quantum computing is largely due to the approach of “quantum advantage.”
An algorithm that extends the use of eigenoptions to settings with stochastic transitions and raw sensory data, demonstrated using Atari games.
A novel deep learning approach for graphs that reduces training cost but performs as well as state-of-the-art neural networks, making it scalable.
A reflection on the Hartree Centre's five years of innovation in high performance computing and data science from IBM researcher Kirk E. Jordan.
IBM Research developed a deep learning model to predict how many times someone will visit the emergency room, which may help to alleviate ER overcrowding.
New disease research from IBM quantifies the connection between agrochemicals and increased predicted schistosomiasis risk in humans for the first time.
IBM scientists in Johannesburg, Africa launched a free, web-based tool to help Africans adopt solar energy.
IBM researchers present an end-to-end framework to manage privacy risks - especially for large scale data sets - and execute a data privacy strategy.
IBM Research and Twiga Foods are extending access to microloans to 220 small food retailers across Kenya using a blockchain-based financing system.
Appearing today in the peer review journal Nature Electronics, IBM scientists introduce a novel hybrid concept called mixed-precision in-memory computing which combines a von Neumann machine with a computational memory unit....
IBM Research released the Adversarial Robustness Toolbox, an open-source software library to make AI systems more secure.
IBM scientists measured the energy levels of single molecules on insulators, for the first time.The research appeared in the journal Nature Nanotechnology.
Half of the 2018 class of IBM Fellows - IBM's highest technical honor - hails from IBM Research.
IBM scientists demonstrate a technique designed for the stop-and-go control of the flow of liquids in capillary-driven microfluidic chips, in the journal Applied Physical Letters.
IBM announced the first startups to join the IBM Q Network, which is dedicated to advancing quantum computing, at the first IBM Q Summit Silicon Valley.
IBM researchers published the first real world demonstration of a rocking Brownian motor for nanoparticles in the peer-review journal Science.
IBM scientists propose a linear-complexity search tool that avoids wasteful and repetitive computations and reduces the average time complexity to linear - potential applications include addressing fake news.
IBM researchers invented a microscopic reality system that is featured in this year's IBM 5 in 5 and can predict the health of the environment.
Today, at IBM THINK in Las Vegas, my colleagues and I are reporting a breakthrough in AI hardware performance using POWER9 with NVIDIA V100 GPUs.
Our mission at IBM is to help our clients change the way the world works. There’s no better example of that than IBM Research’s annual “5 in 5” technology predictions.
Working as the lead of IBM Research’s Strategic Initiative on Aging and being an IBM Master Inventor, my team and I think about how to best design autonomous vehicles through...
Scientists developed a new class of macromolecules designed to selectively kill cancer cells, while ignoring healthy ones.
Over the past decades, the aggressive scaling of transistors made on rigid silicon wafers has steadily boosted the performance of personal electronics and supercomputers. For emerging applications like real-time analytics...
More than 60 technical papers have been published based on experiments run on the IBM Q Experience.
As March 8, 2018 marks International Women's Day, this year's campaign is a #PressforProgress - focusing on gender parity in the community and in the workplace. Since early days at...
IBM scientists have developed an experimental offering named IBM PAIRS Geoscope, a cloud-centric information, discovery, and analytics service that rapidly explores and analyzes massive complex geospatiotemporal data sets.
In 2016, I had a bacterial infection that I acquired after a routine knee surgery. Fortunately, my immune system reacted favorably to antibiotics, but had the bacteria been a “superbug,”...
IBM Research – India and IBM Software Labs have been recognized by the Aegis Graham Bell Award in 2018 for their work in digital agriculture and innovation in IoT. The...
IBM researchers are presenting five papers at this week’s 2018 SPIE conference that show how much EUV can be extended to the 5nm, 3nm, and future nodes.
The Center for Advanced Knowledge Enablement (CAKE) now has a greater focus on AI with machine learning technology at the core of many research projects.
CellCycleTRACER is a supervised machine learning algorithm which classifies and sorts single-cell mass cytometry data according to their cell cycle, and enables correction for cell-cycle-state and cell-volume heterogeneity – essentially...
Cipcigan sees his role as the interface between data, physics and biology and his toolbox includes Big Data, computer simulation, AI and physics-based methods. Combined, these tools are a potent...
If research in silicon chips were a sport, the IEEE International Solid-State Circuits Conference would be its Olympics, and the Lewis Winner Award for Outstanding Paper the equivalent of the...