School of Engineering

Subterms

    More stories

    • 138 Shares199 Views

      in Data Management & Statistics

      New hope for early pancreatic cancer intervention via AI-based risk prediction

      by

      The first documented case of pancreatic cancer dates back to the 18th century. Since then, researchers have undertaken a protracted and challenging odyssey to understand the elusive and deadly disease. To date, there is no better cancer treatment than early intervention. Unfortunately, the pancreas, nestled deep within the abdomen, is particularly elusive for early detection. 

      MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) scientists, alongside Limor Appelbaum, a staff scientist in the Department of Radiation Oncology at Beth Israel Deaconess Medical Center (BIDMC), were eager to better identify potential high-risk patients. They set out to develop two machine-learning models for early detection of pancreatic ductal adenocarcinoma (PDAC), the most common form of the cancer. To access a broad and diverse database, the team synced up with a federated network company, using electronic health record data from various institutions across the United States. This vast pool of data helped ensure the models’ reliability and generalizability, making them applicable across a wide range of populations, geographical locations, and demographic groups.

      The two models — the “PRISM” neural network, and the logistic regression model (a statistical technique for probability), outperformed current methods. The team’s comparison showed that while standard screening criteria identify about 10 percent of PDAC cases using a five-times higher relative risk threshold, Prism can detect 35 percent of PDAC cases at this same threshold. 

      Using AI to detect cancer risk is not a new phenomena — algorithms analyze mammograms, CT scans for lung cancer, and assist in the analysis of Pap smear tests and HPV testing, to name a few applications. “The PRISM models stand out for their development and validation on an extensive database of over 5 million patients, surpassing the scale of most prior research in the field,” says Kai Jia, an MIT PhD student in electrical engineering and computer science (EECS), MIT CSAIL affiliate, and first author on an open-access paper in eBioMedicine outlining the new work. “The model uses routine clinical and lab data to make its predictions, and the diversity of the U.S. population is a significant advancement over other PDAC models, which are usually confined to specific geographic regions, like a few health-care centers in the U.S. Additionally, using a unique regularization technique in the training process enhanced the models’ generalizability and interpretability.” 

      “This report outlines a powerful approach to use big data and artificial intelligence algorithms to refine our approach to identifying risk profiles for cancer,” says David Avigan, a Harvard Medical School professor and the cancer center director and chief of hematology and hematologic malignancies at BIDMC, who was not involved in the study. “This approach may lead to novel strategies to identify patients with high risk for malignancy that may benefit from focused screening with the potential for early intervention.” 

      Prismatic perspectives

      The journey toward the development of PRISM began over six years ago, fueled by firsthand experiences with the limitations of current diagnostic practices. “Approximately 80-85 percent of pancreatic cancer patients are diagnosed at advanced stages, where cure is no longer an option,” says senior author Appelbaum, who is also a Harvard Medical School instructor as well as radiation oncologist. “This clinical frustration sparked the idea to delve into the wealth of data available in electronic health records (EHRs).”The CSAIL group’s close collaboration with Appelbaum made it possible to understand the combined medical and machine learning aspects of the problem better, eventually leading to a much more accurate and transparent model. “The hypothesis was that these records contained hidden clues — subtle signs and symptoms that could act as early warning signals of pancreatic cancer,” she adds. “This guided our use of federated EHR networks in developing these models, for a scalable approach for deploying risk prediction tools in health care.”Both PrismNN and PrismLR models analyze EHR data, including patient demographics, diagnoses, medications, and lab results, to assess PDAC risk. PrismNN uses artificial neural networks to detect intricate patterns in data features like age, medical history, and lab results, yielding a risk score for PDAC likelihood. PrismLR uses logistic regression for a simpler analysis, generating a probability score of PDAC based on these features. Together, the models offer a thorough evaluation of different approaches in predicting PDAC risk from the same EHR data.

      One paramount point for gaining the trust of physicians, the team notes, is better understanding how the models work, known in the field as interpretability. The scientists pointed out that while logistic regression models are inherently easier to interpret, recent advancements have made deep neural networks somewhat more transparent. This helped the team to refine the thousands of potentially predictive features derived from EHR of a single patient to approximately 85 critical indicators. These indicators, which include patient age, diabetes diagnosis, and an increased frequency of visits to physicians, are automatically discovered by the model but match physicians’ understanding of risk factors associated with pancreatic cancer. 

      The path forward

      Despite the promise of the PRISM models, as with all research, some parts are still a work in progress. U.S. data alone are the current diet for the models, necessitating testing and adaptation for global use. The path forward, the team notes, includes expanding the model’s applicability to international datasets and integrating additional biomarkers for more refined risk assessment.

      “A subsequent aim for us is to facilitate the models’ implementation in routine health care settings. The vision is to have these models function seamlessly in the background of health care systems, automatically analyzing patient data and alerting physicians to high-risk cases without adding to their workload,” says Jia. “A machine-learning model integrated with the EHR system could empower physicians with early alerts for high-risk patients, potentially enabling interventions well before symptoms manifest. We are eager to deploy our techniques in the real world to help all individuals enjoy longer, healthier lives.” 

      Jia wrote the paper alongside Applebaum and MIT EECS Professor and CSAIL Principal Investigator Martin Rinard, who are both senior authors of the paper. Researchers on the paper were supported during their time at MIT CSAIL, in part, by the Defense Advanced Research Projects Agency, Boeing, the National Science Foundation, and Aarno Labs. TriNetX provided resources for the project, and the Prevent Cancer Foundation also supported the team. More

    • 75 Shares189 Views

      in Data Management & Statistics

      Self-powered sensor automatically harvests magnetic energy

      by

      MIT researchers have developed a battery-free, self-powered sensor that can harvest energy from its environment.

      Because it requires no battery that must be recharged or replaced, and because it requires no special wiring, such a sensor could be embedded in a hard-to-reach place, like inside the inner workings of a ship’s engine. There, it could automatically gather data on the machine’s power consumption and operations for long periods of time.

      The researchers built a temperature-sensing device that harvests energy from the magnetic field generated in the open air around a wire. One could simply clip the sensor around a wire that carries electricity — perhaps the wire that powers a motor — and it will automatically harvest and store energy which it uses to monitor the motor’s temperature.

      “This is ambient power — energy that I don’t have to make a specific, soldered connection to get. And that makes this sensor very easy to install,” says Steve Leeb, the Emanuel E. Landsman Professor of Electrical Engineering and Computer Science (EECS) and professor of mechanical engineering, a member of the Research Laboratory of Electronics, and senior author of a paper on the energy-harvesting sensor.

      In the paper, which appeared as the featured article in the January issue of the IEEE Sensors Journal, the researchers offer a design guide for an energy-harvesting sensor that lets an engineer balance the available energy in the environment with their sensing needs.

      The paper lays out a roadmap for the key components of a device that can sense and control the flow of energy continually during operation.

      The versatile design framework is not limited to sensors that harvest magnetic field energy, and can be applied to those that use other power sources, like vibrations or sunlight. It could be used to build networks of sensors for factories, warehouses, and commercial spaces that cost less to install and maintain.

      “We have provided an example of a battery-less sensor that does something useful, and shown that it is a practically realizable solution. Now others will hopefully use our framework to get the ball rolling to design their own sensors,” says lead author Daniel Monagle, an EECS graduate student.

      Monagle and Leeb are joined on the paper by EECS graduate student Eric Ponce.

      John Donnal, an associate professor of weapons and controls engineering at the U.S. Naval Academy who was not involved with this work, studies techniques to monitor ship systems. Getting access to power on a ship can be difficult, he says, since there are very few outlets and strict restrictions as to what equipment can be plugged in.

      “Persistently measuring the vibration of a pump, for example, could give the crew real-time information on the health of the bearings and mounts, but powering a retrofit sensor often requires so much additional infrastructure that the investment is not worthwhile,” Donnal adds. “Energy-harvesting systems like this could make it possible to retrofit a wide variety of diagnostic sensors on ships and significantly reduce the overall cost of maintenance.”

      A how-to guide

      The researchers had to meet three key challenges to develop an effective, battery-free, energy-harvesting sensor.

      First, the system must be able to cold start, meaning it can fire up its electronics with no initial voltage. They accomplished this with a network of integrated circuits and transistors that allow the system to store energy until it reaches a certain threshold. The system will only turn on once it has stored enough power to fully operate.

      Second, the system must store and convert the energy it harvests efficiently, and without a battery. While the researchers could have included a battery, that would add extra complexities to the system and could pose a fire risk.

      “You might not even have the luxury of sending out a technician to replace a battery. Instead, our system is maintenance-free. It harvests energy and operates itself,” Monagle adds.

      To avoid using a battery, they incorporate internal energy storage that can include a series of capacitors. Simpler than a battery, a capacitor stores energy in the electrical field between conductive plates. Capacitors can be made from a variety of materials, and their capabilities can be tuned to a range of operating conditions, safety requirements, and available space.

      The team carefully designed the capacitors so they are big enough to store the energy the device needs to turn on and start harvesting power, but small enough that the charge-up phase doesn’t take too long.

      In addition, since a sensor might go weeks or even months before turning on to take a measurement, they ensured the capacitors can hold enough energy even if some leaks out over time.

      Finally, they developed a series of control algorithms that dynamically measure and budget the energy collected, stored, and used by the device. A microcontroller, the “brain” of the energy management interface, constantly checks how much energy is stored and infers whether to turn the sensor on or off, take a measurement, or kick the harvester into a higher gear so it can gather more energy for more complex sensing needs.

      “Just like when you change gears on a bike, the energy management interface looks at how the harvester is doing, essentially seeing whether it is pedaling too hard or too soft, and then it varies the electronic load so it can maximize the amount of power it is harvesting and match the harvest to the needs of the sensor,” Monagle explains.

      Self-powered sensor

      Using this design framework, they built an energy management circuit for an off-the-shelf temperature sensor. The device harvests magnetic field energy and uses it to continually sample temperature data, which it sends to a smartphone interface using Bluetooth.

      The researchers used super-low-power circuits to design the device, but quickly found that these circuits have tight restrictions on how much voltage they can withstand before breaking down. Harvesting too much power could cause the device to explode.

      To avoid that, their energy harvester operating system in the microcontroller automatically adjusts or reduces the harvest if the amount of stored energy becomes excessive.

      They also found that communication — transmitting data gathered by the temperature sensor — was by far the most power-hungry operation.

      “Ensuring the sensor has enough stored energy to transmit data is a constant challenge that involves careful design,” Monagle says.

      In the future, the researchers plan to explore less energy-intensive means of transmitting data, such as using optics or acoustics. They also want to more rigorously model and predict how much energy might be coming into a system, or how much energy a sensor might need to take measurements, so a device could effectively gather even more data.

      “If you only make the measurements you think you need, you may miss something really valuable. With more information, you might be able to learn something you didn’t expect about a device’s operations. Our framework lets you balance those considerations,” Leeb says.  

      “This paper is well-documented regarding what a practical self-powered sensor node should internally entail for realistic scenarios. The overall design guidelines, particularly on the cold-start issue, are very helpful,” says Jinyeong Moon, an assistant professor of electrical and computer engineering at Florida State University College of Engineering who was not involved with this work. “Engineers planning to design a self-powering module for a wireless sensor node will greatly benefit from these guidelines, easily ticking off traditionally cumbersome cold-start-related checklists.”

      The work is supported, in part, by the Office of Naval Research and The Grainger Foundation. More

    • 125 Shares169 Views

      in Data Management & Statistics

      Multiple AI models help robots execute complex plans more transparently

      by

      Your daily to-do list is likely pretty straightforward: wash the dishes, buy groceries, and other minutiae. It’s unlikely you wrote out “pick up the first dirty dish,” or “wash that plate with a sponge,” because each of these miniature steps within the chore feels intuitive. While we can routinely complete each step without much thought, a robot requires a complex plan that involves more detailed outlines.

      MIT’s Improbable AI Lab, a group within the Computer Science and Artificial Intelligence Laboratory (CSAIL), has offered these machines a helping hand with a new multimodal framework: Compositional Foundation Models for Hierarchical Planning (HiP), which develops detailed, feasible plans with the expertise of three different foundation models. Like OpenAI’s GPT-4, the foundation model that ChatGPT and Bing Chat were built upon, these foundation models are trained on massive quantities of data for applications like generating images, translating text, and robotics.Unlike RT2 and other multimodal models that are trained on paired vision, language, and action data, HiP uses three different foundation models each trained on different data modalities. Each foundation model captures a different part of the decision-making process and then works together when it’s time to make decisions. HiP removes the need for access to paired vision, language, and action data, which is difficult to obtain. HiP also makes the reasoning process more transparent.

      What’s considered a daily chore for a human can be a robot’s “long-horizon goal” — an overarching objective that involves completing many smaller steps first — requiring sufficient data to plan, understand, and execute objectives. While computer vision researchers have attempted to build monolithic foundation models for this problem, pairing language, visual, and action data is expensive. Instead, HiP represents a different, multimodal recipe: a trio that cheaply incorporates linguistic, physical, and environmental intelligence into a robot.

      “Foundation models do not have to be monolithic,” says NVIDIA AI researcher Jim Fan, who was not involved in the paper. “This work decomposes the complex task of embodied agent planning into three constituent models: a language reasoner, a visual world model, and an action planner. It makes a difficult decision-making problem more tractable and transparent.”The team believes that their system could help these machines accomplish household chores, such as putting away a book or placing a bowl in the dishwasher. Additionally, HiP could assist with multistep construction and manufacturing tasks, like stacking and placing different materials in specific sequences.Evaluating HiP

      The CSAIL team tested HiP’s acuity on three manipulation tasks, outperforming comparable frameworks. The system reasoned by developing intelligent plans that adapt to new information.

      First, the researchers requested that it stack different-colored blocks on each other and then place others nearby. The catch: Some of the correct colors weren’t present, so the robot had to place white blocks in a color bowl to paint them. HiP often adjusted to these changes accurately, especially compared to state-of-the-art task planning systems like Transformer BC and Action Diffuser, by adjusting its plans to stack and place each square as needed.

      Another test: arranging objects such as candy and a hammer in a brown box while ignoring other items. Some of the objects it needed to move were dirty, so HiP adjusted its plans to place them in a cleaning box, and then into the brown container. In a third demonstration, the bot was able to ignore unnecessary objects to complete kitchen sub-goals such as opening a microwave, clearing a kettle out of the way, and turning on a light. Some of the prompted steps had already been completed, so the robot adapted by skipping those directions.

      A three-pronged hierarchy

      HiP’s three-pronged planning process operates as a hierarchy, with the ability to pre-train each of its components on different sets of data, including information outside of robotics. At the bottom of that order is a large language model (LLM), which starts to ideate by capturing all the symbolic information needed and developing an abstract task plan. Applying the common sense knowledge it finds on the internet, the model breaks its objective into sub-goals. For example, “making a cup of tea” turns into “filling a pot with water,” “boiling the pot,” and the subsequent actions required.

      “All we want to do is take existing pre-trained models and have them successfully interface with each other,” says Anurag Ajay, a PhD student in the MIT Department of Electrical Engineering and Computer Science (EECS) and a CSAIL affiliate. “Instead of pushing for one model to do everything, we combine multiple ones that leverage different modalities of internet data. When used in tandem, they help with robotic decision-making and can potentially aid with tasks in homes, factories, and construction sites.”

      These models also need some form of “eyes” to understand the environment they’re operating in and correctly execute each sub-goal. The team used a large video diffusion model to augment the initial planning completed by the LLM, which collects geometric and physical information about the world from footage on the internet. In turn, the video model generates an observation trajectory plan, refining the LLM’s outline to incorporate new physical knowledge.This process, known as iterative refinement, allows HiP to reason about its ideas, taking in feedback at each stage to generate a more practical outline. The flow of feedback is similar to writing an article, where an author may send their draft to an editor, and with those revisions incorporated in, the publisher reviews for any last changes and finalizes.

      In this case, the top of the hierarchy is an egocentric action model, or a sequence of first-person images that infer which actions should take place based on its surroundings. During this stage, the observation plan from the video model is mapped over the space visible to the robot, helping the machine decide how to execute each task within the long-horizon goal. If a robot uses HiP to make tea, this means it will have mapped out exactly where the pot, sink, and other key visual elements are, and begin completing each sub-goal.Still, the multimodal work is limited by the lack of high-quality video foundation models. Once available, they could interface with HiP’s small-scale video models to further enhance visual sequence prediction and robot action generation. A higher-quality version would also reduce the current data requirements of the video models.That being said, the CSAIL team’s approach only used a tiny bit of data overall. Moreover, HiP was cheap to train and demonstrated the potential of using readily available foundation models to complete long-horizon tasks. “What Anurag has demonstrated is proof-of-concept of how we can take models trained on separate tasks and data modalities and combine them into models for robotic planning. In the future, HiP could be augmented with pre-trained models that can process touch and sound to make better plans,” says senior author Pulkit Agrawal, MIT assistant professor in EECS and director of the Improbable AI Lab. The group is also considering applying HiP to solving real-world long-horizon tasks in robotics.Ajay and Agrawal are lead authors on a paper describing the work. They are joined by MIT professors and CSAIL principal investigators Tommi Jaakkola, Joshua Tenenbaum, and Leslie Pack Kaelbling; CSAIL research affiliate and MIT-IBM AI Lab research manager Akash Srivastava; graduate students Seungwook Han and Yilun Du ’19; former postdoc Abhishek Gupta, who is now assistant professor at University of Washington; and former graduate student Shuang Li PhD ’23.

      The team’s work was supported, in part, by the National Science Foundation, the U.S. Defense Advanced Research Projects Agency, the U.S. Army Research Office, the U.S. Office of Naval Research Multidisciplinary University Research Initiatives, and the MIT-IBM Watson AI Lab. Their findings were presented at the 2023 Conference on Neural Information Processing Systems (NeurIPS). More

    • 63 Shares169 Views

      in Data Management & Statistics

      Inclusive research for social change

      by

      Pair a decades-old program dedicated to creating research opportunities for underrepresented minorities and populations with a growing initiative committed to tackling the very issues at the heart of such disparities, and you’ll get a transformative partnership that only MIT can deliver. 

      Since 1986, the MIT Summer Research Program (MSRP) has led an institutional effort to prepare underrepresented students (minorities, women in STEM, or students with low socioeconomic status) for doctoral education by pairing them with MIT labs and research groups. For the past three years, the Initiative on Combatting Systemic Racism (ICSR), a cross-disciplinary research collaboration led by MIT’s Institute for Data, Systems, and Society (IDSS), has joined them in their mission, helping bring the issue full circle by providing MSRP students with the opportunity to use big data and computational tools to create impactful changes toward racial equity.

      “ICSR has further enabled our direct engagement with undergrads, both within and outside of MIT,” says Fotini Christia, the Ford International Professor of the Social Sciences, associate director of IDSS, and co-organizer for the initiative. “We’ve found that this line of research has attracted students interested in examining these topics with the most rigorous methods.”

      The initiative fits well under the IDSS banner, as IDSS research seeks solutions to complex societal issues through a multidisciplinary approach that includes statistics, computation, modeling, social science methodologies, human behavior, and an understanding of complex systems. With the support of faculty and researchers from all five schools and the MIT Schwarzman College of Computing, the objective of ICSR is to work on an array of different societal aspects of systemic racism through a set of verticals including policing, housing, health care, and social media.

      Where passion meets impact

      Grinnell senior Mia Hines has always dreamed of using her love for computer science to support social justice. She has experience working with unhoused people and labor unions, and advocating for Indigenous peoples’ rights. When applying to college, she focused her essay on using technology to help Syrian refugees.

      “As a Black woman, it’s very important to me that we focus on these areas, especially on how we can use technology to help marginalized communities,” Hines says. “And also, how do we stop technology or improve technology that is already hurting marginalized communities?”   

      Through MSRP, Hines was paired with research advisor Ufuoma Ovienmhada, a fourth-year doctoral student in the Department of Aeronautics and Astronautics at MIT. A member of Professor Danielle Wood’s Space Enabled research group at MIT’s Media Lab, Ovienmhada received funding from an ICSR Seed Grant and NASA’s Applied Sciences Program to support her ongoing research measuring environmental injustice and socioeconomic disparities in prison landscapes. 

      “I had been doing satellite remote sensing for environmental challenges and sustainability, starting out looking at coastal ecosystems, when I learned about an issue called ‘prison ecology,’” Ovienmhada explains. “This refers to the intersection of mass incarceration and environmental justice.”

      Ovienmhada’s research uses satellite remote sensing and environmental data to characterize exposures to different environmental hazards such as air pollution, extreme heat, and flooding. “This allows others to use these datasets for real-time advocacy, in addition to creating public awareness,” she says.

      Focused especially on extreme heat, Hines used satellite remote sensing to monitor the fluctuation of temperature to assess the risk being imposed on prisoners, including death, especially in states like Texas, where 75 percent of prisons either don’t have full air conditioning or have none at all.

      “Before this project I had done little to no work with geospatial data, and as a budding data scientist, getting to work with and understanding different types of data and resources is really helpful,” Hines says. “I was also funded and afforded the flexibility to take advantage of IDSS’s Data Science and Machine Learning online course. It was really great to be able to do that and learn even more.”

      Filling the gap

      Much like Hines, Harvey Mudd senior Megan Li was specifically interested in the IDSS-supported MSRP projects. She was drawn to the interdisciplinary approach, and she seeks in her own work to apply computational methods to societal issues and to make computer science more inclusive, considerate, and ethical. 

      Working with Aurora Zhang, a grad student in IDSS’s Social and Engineering Systems PhD program, Li used county-level data on income and housing prices to quantify and visualize how affordability based on income alone varies across the United States. She then expanded the analysis to include assets and debt to determine the most common barriers to home ownership.

      “I spent my day-to-day looking at census data and writing Python scripts that could work with it,” reports Li. “I also reached out to the Census Bureau directly to learn a little bit more about how they did their data collection, and discussed questions related to some of their previous studies and working papers that I had reviewed.” 

      Outside of actual day-to-day research, Li says she learned a lot in conversations with fellow researchers, particularly changing her “skeptical view” of whether or not mortgage lending algorithms would help or hurt home buyers in the approval process. “I think I have a little bit more faith now, which is a good thing.”

      “Harvey Mudd is undergraduate-only, and while professors do run labs here, my specific research areas are not well represented,” Li says. “This opportunity was enormous in that I got the experience I need to see if this research area is actually something that I want to do long term, and I got more mirrors into what I would be doing in grad school from talking to students and getting to know faculty.”

      Closing the loop

      While participating in MSRP offered crucial research experience to Hines, the ICSR projects enabled her to engage in topics she’s passionate about and work that could drive tangible societal change.

      “The experience felt much more concrete because we were working on these very sophisticated projects, in a supportive environment where people were very excited to work with us,” she says.

      A significant benefit for Li was the chance to steer her research in alignment with her own interests. “I was actually given the opportunity to propose my own research idea, versus supporting a graduate student’s work in progress,” she explains. 

      For Ovienmhada, the pairing of the two initiatives solidifies the efforts of MSRP and closes a crucial loop in diversity, equity, and inclusion advocacy. 

      “I’ve participated in a lot of different DEI-related efforts and advocacy and one thing that always comes up is the fact that it’s not just about bringing people in, it’s also about creating an environment and opportunities that align with people’s values,” Ovienmhada says. “Programs like MSRP and ICSR create opportunities for people who want to do work that’s aligned with certain values by providing the needed mentoring and financial support.” More

    • 63 Shares159 Views

      in Data Management & Statistics

      Leveraging language to understand machines

      by

      Natural language conveys ideas, actions, information, and intent through context and syntax; further, there are volumes of it contained in databases. This makes it an excellent source of data to train machine-learning systems on. Two master’s of engineering students in the 6A MEng Thesis Program at MIT, Irene Terpstra ’23 and Rujul Gandhi ’22, are working with mentors in the MIT-IBM Watson AI Lab to use this power of natural language to build AI systems.

      As computing is becoming more advanced, researchers are looking to improve the hardware that they run on; this means innovating to create new computer chips. And, since there is literature already available on modifications that can be made to achieve certain parameters and performance, Terpstra and her mentors and advisors Anantha Chandrakasan, MIT School of Engineering dean and the Vannevar Bush Professor of Electrical Engineering and Computer Science, and IBM’s researcher Xin Zhang, are developing an AI algorithm that assists in chip design.

      “I’m creating a workflow to systematically analyze how these language models can help the circuit design process. What reasoning powers do they have, and how can it be integrated into the chip design process?” says Terpstra. “And then on the other side, if that proves to be useful enough, [we’ll] see if they can automatically design the chips themselves, attaching it to a reinforcement learning algorithm.”

      To do this, Terpstra’s team is creating an AI system that can iterate on different designs. It means experimenting with various pre-trained large language models (like ChatGPT, Llama 2, and Bard), using an open-source circuit simulator language called NGspice, which has the parameters of the chip in code form, and a reinforcement learning algorithm. With text prompts, researchers will be able to query how the physical chip should be modified to achieve a certain goal in the language model and produced guidance for adjustments. This is then transferred into a reinforcement learning algorithm that updates the circuit design and outputs new physical parameters of the chip.

      “The final goal would be to combine the reasoning powers and the knowledge base that is baked into these large language models and combine that with the optimization power of the reinforcement learning algorithms and have that design the chip itself,” says Terpstra.

      Rujul Gandhi works with the raw language itself. As an undergraduate at MIT, Gandhi explored linguistics and computer sciences, putting them together in her MEng work. “I’ve been interested in communication, both between just humans and between humans and computers,” Gandhi says.

      Robots or other interactive AI systems are one area where communication needs to be understood by both humans and machines. Researchers often write instructions for robots using formal logic. This helps ensure that commands are being followed safely and as intended, but formal logic can be difficult for users to understand, while natural language comes easily. To ensure this smooth communication, Gandhi and her advisors Yang Zhang of IBM and MIT assistant professor Chuchu Fan are building a parser that converts natural language instructions into a machine-friendly form. Leveraging the linguistic structure encoded by the pre-trained encoder-decoder model T5, and a dataset of annotated, basic English commands for performing certain tasks, Gandhi’s system identifies the smallest logical units, or atomic propositions, which are present in a given instruction.

      “Once you’ve given your instruction, the model identifies all the smaller sub-tasks you want it to carry out,” Gandhi says. “Then, using a large language model, each sub-task can be compared against the available actions and objects in the robot’s world, and if any sub-task can’t be carried out because a certain object is not recognized, or an action is not possible, the system can stop right there to ask the user for help.”

      This approach of breaking instructions into sub-tasks also allows her system to understand logical dependencies expressed in English, like, “do task X until event Y happens.” Gandhi uses a dataset of step-by-step instructions across robot task domains like navigation and manipulation, with a focus on household tasks. Using data that are written just the way humans would talk to each other has many advantages, she says, because it means a user can be more flexible about how they phrase their instructions.

      Another of Gandhi’s projects involves developing speech models. In the context of speech recognition, some languages are considered “low resource” since they might not have a lot of transcribed speech available, or might not have a written form at all. “One of the reasons I applied to this internship at the MIT-IBM Watson AI Lab was an interest in language processing for low-resource languages,” she says. “A lot of language models today are very data-driven, and when it’s not that easy to acquire all of that data, that’s when you need to use the limited data efficiently.” 

      Speech is just a stream of sound waves, but humans having a conversation can easily figure out where words and thoughts start and end. In speech processing, both humans and language models use their existing vocabulary to recognize word boundaries and understand the meaning. In low- or no-resource languages, a written vocabulary might not exist at all, so researchers can’t provide one to the model. Instead, the model can make note of what sound sequences occur together more frequently than others, and infer that those might be individual words or concepts. In Gandhi’s research group, these inferred words are then collected into a pseudo-vocabulary that serves as a labeling method for the low-resource language, creating labeled data for further applications.

      The applications for language technology are “pretty much everywhere,” Gandhi says. “You could imagine people being able to interact with software and devices in their native language, their native dialect. You could imagine improving all the voice assistants that we use. You could imagine it being used for translation or interpretation.” More

    • 113 Shares119 Views

      in Data Management & Statistics

      Image recognition accuracy: An unseen challenge confounding today’s AI

      by

      Imagine you are scrolling through the photos on your phone and you come across an image that at first you can’t recognize. It looks like maybe something fuzzy on the couch; could it be a pillow or a coat? After a couple of seconds it clicks — of course! That ball of fluff is your friend’s cat, Mocha. While some of your photos could be understood in an instant, why was this cat photo much more difficult?

      MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) researchers were surprised to find that despite the critical importance of understanding visual data in pivotal areas ranging from health care to transportation to household devices, the notion of an image’s recognition difficulty for humans has been almost entirely ignored. One of the major drivers of progress in deep learning-based AI has been datasets, yet we know little about how data drives progress in large-scale deep learning beyond that bigger is better.

      In real-world applications that require understanding visual data, humans outperform object recognition models despite the fact that models perform well on current datasets, including those explicitly designed to challenge machines with debiased images or distribution shifts. This problem persists, in part, because we have no guidance on the absolute difficulty of an image or dataset. Without controlling for the difficulty of images used for evaluation, it’s hard to objectively assess progress toward human-level performance, to cover the range of human abilities, and to increase the challenge posed by a dataset.

      To fill in this knowledge gap, David Mayo, an MIT PhD student in electrical engineering and computer science and a CSAIL affiliate, delved into the deep world of image datasets, exploring why certain images are more difficult for humans and machines to recognize than others. “Some images inherently take longer to recognize, and it’s essential to understand the brain’s activity during this process and its relation to machine learning models. Perhaps there are complex neural circuits or unique mechanisms missing in our current models, visible only when tested with challenging visual stimuli. This exploration is crucial for comprehending and enhancing machine vision models,” says Mayo, a lead author of a new paper on the work.

      This led to the development of a new metric, the “minimum viewing time” (MVT), which quantifies the difficulty of recognizing an image based on how long a person needs to view it before making a correct identification. Using a subset of ImageNet, a popular dataset in machine learning, and ObjectNet, a dataset designed to test object recognition robustness, the team showed images to participants for varying durations from as short as 17 milliseconds to as long as 10 seconds, and asked them to choose the correct object from a set of 50 options. After over 200,000 image presentation trials, the team found that existing test sets, including ObjectNet, appeared skewed toward easier, shorter MVT images, with the vast majority of benchmark performance derived from images that are easy for humans.

      The project identified interesting trends in model performance — particularly in relation to scaling. Larger models showed considerable improvement on simpler images but made less progress on more challenging images. The CLIP models, which incorporate both language and vision, stood out as they moved in the direction of more human-like recognition.

      “Traditionally, object recognition datasets have been skewed towards less-complex images, a practice that has led to an inflation in model performance metrics, not truly reflective of a model’s robustness or its ability to tackle complex visual tasks. Our research reveals that harder images pose a more acute challenge, causing a distribution shift that is often not accounted for in standard evaluations,” says Mayo. “We released image sets tagged by difficulty along with tools to automatically compute MVT, enabling MVT to be added to existing benchmarks and extended to various applications. These include measuring test set difficulty before deploying real-world systems, discovering neural correlates of image difficulty, and advancing object recognition techniques to close the gap between benchmark and real-world performance.”

      “One of my biggest takeaways is that we now have another dimension to evaluate models on. We want models that are able to recognize any image even if — perhaps especially if — it’s hard for a human to recognize. We’re the first to quantify what this would mean. Our results show that not only is this not the case with today’s state of the art, but also that our current evaluation methods don’t have the ability to tell us when it is the case because standard datasets are so skewed toward easy images,” says Jesse Cummings, an MIT graduate student in electrical engineering and computer science and co-first author with Mayo on the paper.

      From ObjectNet to MVT

      A few years ago, the team behind this project identified a significant challenge in the field of machine learning: Models were struggling with out-of-distribution images, or images that were not well-represented in the training data. Enter ObjectNet, a dataset comprised of images collected from real-life settings. The dataset helped illuminate the performance gap between machine learning models and human recognition abilities, by eliminating spurious correlations present in other benchmarks — for example, between an object and its background. ObjectNet illuminated the gap between the performance of machine vision models on datasets and in real-world applications, encouraging use for many researchers and developers — which subsequently improved model performance.

      Fast forward to the present, and the team has taken their research a step further with MVT. Unlike traditional methods that focus on absolute performance, this new approach assesses how models perform by contrasting their responses to the easiest and hardest images. The study further explored how image difficulty could be explained and tested for similarity to human visual processing. Using metrics like c-score, prediction depth, and adversarial robustness, the team found that harder images are processed differently by networks. “While there are observable trends, such as easier images being more prototypical, a comprehensive semantic explanation of image difficulty continues to elude the scientific community,” says Mayo.

      In the realm of health care, for example, the pertinence of understanding visual complexity becomes even more pronounced. The ability of AI models to interpret medical images, such as X-rays, is subject to the diversity and difficulty distribution of the images. The researchers advocate for a meticulous analysis of difficulty distribution tailored for professionals, ensuring AI systems are evaluated based on expert standards, rather than layperson interpretations.

      Mayo and Cummings are currently looking at neurological underpinnings of visual recognition as well, probing into whether the brain exhibits differential activity when processing easy versus challenging images. The study aims to unravel whether complex images recruit additional brain areas not typically associated with visual processing, hopefully helping demystify how our brains accurately and efficiently decode the visual world.

      Toward human-level performance

      Looking ahead, the researchers are not only focused on exploring ways to enhance AI’s predictive capabilities regarding image difficulty. The team is working on identifying correlations with viewing-time difficulty in order to generate harder or easier versions of images.

      Despite the study’s significant strides, the researchers acknowledge limitations, particularly in terms of the separation of object recognition from visual search tasks. The current methodology does concentrate on recognizing objects, leaving out the complexities introduced by cluttered images.

      “This comprehensive approach addresses the long-standing challenge of objectively assessing progress towards human-level performance in object recognition and opens new avenues for understanding and advancing the field,” says Mayo. “With the potential to adapt the Minimum Viewing Time difficulty metric for a variety of visual tasks, this work paves the way for more robust, human-like performance in object recognition, ensuring that models are truly put to the test and are ready for the complexities of real-world visual understanding.”

      “This is a fascinating study of how human perception can be used to identify weaknesses in the ways AI vision models are typically benchmarked, which overestimate AI performance by concentrating on easy images,” says Alan L. Yuille, Bloomberg Distinguished Professor of Cognitive Science and Computer Science at Johns Hopkins University, who was not involved in the paper. “This will help develop more realistic benchmarks leading not only to improvements to AI but also make fairer comparisons between AI and human perception.” 

      “It’s widely claimed that computer vision systems now outperform humans, and on some benchmark datasets, that’s true,” says Anthropic technical staff member Simon Kornblith PhD ’17, who was also not involved in this work. “However, a lot of the difficulty in those benchmarks comes from the obscurity of what’s in the images; the average person just doesn’t know enough to classify different breeds of dogs. This work instead focuses on images that people can only get right if given enough time. These images are generally much harder for computer vision systems, but the best systems are only a bit worse than humans.”

      Mayo, Cummings, and Xinyu Lin MEng ’22 wrote the paper alongside CSAIL Research Scientist Andrei Barbu, CSAIL Principal Research Scientist Boris Katz, and MIT-IBM Watson AI Lab Principal Researcher Dan Gutfreund. The researchers are affiliates of the MIT Center for Brains, Minds, and Machines.

      The team is presenting their work at the 2023 Conference on Neural Information Processing Systems (NeurIPS). More

    • 125 Shares129 Views

      in Data Management & Statistics

      Three MIT students selected as inaugural MIT-Pillar AI Collective Fellows

      by

      MIT-Pillar AI Collective has announced three inaugural fellows for the fall 2023 semester. With support from the program, the graduate students, who are in their final year of a master’s or PhD program, will conduct research in the areas of artificial intelligence, machine learning, and data science with the aim of commercializing their innovations.

      Launched by MIT’s School of Engineering and Pillar VC in 2022, the MIT-Pillar AI Collective supports faculty, postdocs, and students conducting research on AI, machine learning, and data science. Supported by a gift from Pillar VC and administered by the MIT Deshpande Center for Technological Innovation, the mission of the program is to advance research toward commercialization.

      The fall 2023 MIT-Pillar AI Collective Fellows are:

      Alexander Andonian SM ’21 is a PhD candidate in electrical engineering and computer science whose research interests lie in computer vision, deep learning, and artificial intelligence. More specifically, he is focused on building a generalist, multimodal AI scientist driven by generative vision-language model agents capable of proposing scientific hypotheses, running computational experiments, evaluating supporting evidence, and verifying conclusions in the same way as a human researcher or reviewer. Such an agent could be trained to optimally distill and communicate its findings for human consumption and comprehension. Andonian’s work holds the promise of creating a concrete foundation for rigorously building and holistically testing the next-generation autonomous AI agent for science. In addition to his research, Andonian is the CEO and co-founder of Reelize, a startup that offers a generative AI video tool that effortlessly turns long videos into short clips — and originated from his business coursework and was supported by MIT Sandbox. Andonian is also a founding AI researcher at Poly AI, an early-stage YC-backed startup building AI design tools. Andonian earned an SM from MIT and a BS in neuroscience, physics, and mathematics from Bates College.

      Daniel Magley is a PhD candidate in the Harvard-MIT Program in Health Sciences and Technology who is passionate about making a healthy, fully functioning mind and body a reality for all. His leading-edge research is focused on developing a swallowable wireless thermal imaging capsule that could be used in treating and monitoring inflammatory bowel diseases and their manifestations, such as Crohn’s disease. Providing increased sensitivity and eliminating the need for bowel preparation, the capsule has the potential to vastly improve treatment efficacy and overall patient experience in routine monitoring. The capsule has completed animal studies and is entering human studies at Mass General Brigham, where Magley leads a team of engineers in the hospital’s largest translational research lab, the Tearney Lab. Following the human pilot studies, the largest technological and regulatory risks will be cleared for translation. Magley will then begin focusing on a multi-site study to get the device into clinics, with the promise of benefiting patients across the country. Magley earned a BS in electrical engineering from Caltech.

      Madhumitha Ravichandra is a PhD candidate interested in advancing heat transfer and surface engineering techniques to enhance the safety and performance of nuclear energy systems and reduce their environmental impacts. Leveraging her deep knowledge of the integration of explainable AI with high-throughput autonomous experimentation, she seeks to transform the development of radiation-hardened (rad-hard) sensors, which could potentially withstand and function amidst radiation levels that would render conventional sensors useless. By integrating explainable AI with high-throughput autonomous experimentation, she aims to rapidly iterate designs, test under varied conditions, and ensure that the final product is both robust and transparent in its operations. Her work in this space could shift the paradigm in rad-hard sensor development, addressing a glaring void in the market and redefining standards, ensuring that nuclear and space applications are safer, more efficient, and at the cutting edge of technological progress. Ravichandran earned a BTech in mechanical engineering from SASTRA University, India. More

    • 138 Shares169 Views

      in Data Management & Statistics

      Automated system teaches users when to collaborate with an AI assistant

      by

      Artificial intelligence models that pick out patterns in images can often do so better than human eyes — but not always. If a radiologist is using an AI model to help her determine whether a patient’s X-rays show signs of pneumonia, when should she trust the model’s advice and when should she ignore it?

      A customized onboarding process could help this radiologist answer that question, according to researchers at MIT and the MIT-IBM Watson AI Lab. They designed a system that teaches a user when to collaborate with an AI assistant.

      In this case, the training method might find situations where the radiologist trusts the model’s advice — except she shouldn’t because the model is wrong. The system automatically learns rules for how she should collaborate with the AI, and describes them with natural language.

      During onboarding, the radiologist practices collaborating with the AI using training exercises based on these rules, receiving feedback about her performance and the AI’s performance.

      The researchers found that this onboarding procedure led to about a 5 percent improvement in accuracy when humans and AI collaborated on an image prediction task. Their results also show that just telling the user when to trust the AI, without training, led to worse performance.

      Importantly, the researchers’ system is fully automated, so it learns to create the onboarding process based on data from the human and AI performing a specific task. It can also adapt to different tasks, so it can be scaled up and used in many situations where humans and AI models work together, such as in social media content moderation, writing, and programming.

      “So often, people are given these AI tools to use without any training to help them figure out when it is going to be helpful. That’s not what we do with nearly every other tool that people use — there is almost always some kind of tutorial that comes with it. But for AI, this seems to be missing. We are trying to tackle this problem from a methodological and behavioral perspective,” says Hussein Mozannar, a graduate student in the Social and Engineering Systems doctoral program within the Institute for Data, Systems, and Society (IDSS) and lead author of a paper about this training process.

      The researchers envision that such onboarding will be a crucial part of training for medical professionals.

      “One could imagine, for example, that doctors making treatment decisions with the help of AI will first have to do training similar to what we propose. We may need to rethink everything from continuing medical education to the way clinical trials are designed,” says senior author David Sontag, a professor of EECS, a member of the MIT-IBM Watson AI Lab and the MIT Jameel Clinic, and the leader of the Clinical Machine Learning Group of the Computer Science and Artificial Intelligence Laboratory (CSAIL).

      Mozannar, who is also a researcher with the Clinical Machine Learning Group, is joined on the paper by Jimin J. Lee, an undergraduate in electrical engineering and computer science; Dennis Wei, a senior research scientist at IBM Research; and Prasanna Sattigeri and Subhro Das, research staff members at the MIT-IBM Watson AI Lab. The paper will be presented at the Conference on Neural Information Processing Systems.

      Training that evolves

      Existing onboarding methods for human-AI collaboration are often composed of training materials produced by human experts for specific use cases, making them difficult to scale up. Some related techniques rely on explanations, where the AI tells the user its confidence in each decision, but research has shown that explanations are rarely helpful, Mozannar says.

      “The AI model’s capabilities are constantly evolving, so the use cases where the human could potentially benefit from it are growing over time. At the same time, the user’s perception of the model continues changing. So, we need a training procedure that also evolves over time,” he adds.

      To accomplish this, their onboarding method is automatically learned from data. It is built from a dataset that contains many instances of a task, such as detecting the presence of a traffic light from a blurry image.

      The system’s first step is to collect data on the human and AI performing this task. In this case, the human would try to predict, with the help of AI, whether blurry images contain traffic lights.

      The system embeds these data points onto a latent space, which is a representation of data in which similar data points are closer together. It uses an algorithm to discover regions of this space where the human collaborates incorrectly with the AI. These regions capture instances where the human trusted the AI’s prediction but the prediction was wrong, and vice versa.

      Perhaps the human mistakenly trusts the AI when images show a highway at night.

      After discovering the regions, a second algorithm utilizes a large language model to describe each region as a rule, using natural language. The algorithm iteratively fine-tunes that rule by finding contrasting examples. It might describe this region as “ignore AI when it is a highway during the night.”

      These rules are used to build training exercises. The onboarding system shows an example to the human, in this case a blurry highway scene at night, as well as the AI’s prediction, and asks the user if the image shows traffic lights. The user can answer yes, no, or use the AI’s prediction.

      If the human is wrong, they are shown the correct answer and performance statistics for the human and AI on these instances of the task. The system does this for each region, and at the end of the training process, repeats the exercises the human got wrong.

      “After that, the human has learned something about these regions that we hope they will take away in the future to make more accurate predictions,” Mozannar says.

      Onboarding boosts accuracy

      The researchers tested this system with users on two tasks — detecting traffic lights in blurry images and answering multiple choice questions from many domains (such as biology, philosophy, computer science, etc.).

      They first showed users a card with information about the AI model, how it was trained, and a breakdown of its performance on broad categories. Users were split into five groups: Some were only shown the card, some went through the researchers’ onboarding procedure, some went through a baseline onboarding procedure, some went through the researchers’ onboarding procedure and were given recommendations of when they should or should not trust the AI, and others were only given the recommendations.

      Only the researchers’ onboarding procedure without recommendations improved users’ accuracy significantly, boosting their performance on the traffic light prediction task by about 5 percent without slowing them down. However, onboarding was not as effective for the question-answering task. The researchers believe this is because the AI model, ChatGPT, provided explanations with each answer that convey whether it should be trusted.

      But providing recommendations without onboarding had the opposite effect — users not only performed worse, they took more time to make predictions.

      “When you only give someone recommendations, it seems like they get confused and don’t know what to do. It derails their process. People also don’t like being told what to do, so that is a factor as well,” Mozannar says.

      Providing recommendations alone could harm the user if those recommendations are wrong, he adds. With onboarding, on the other hand, the biggest limitation is the amount of available data. If there aren’t enough data, the onboarding stage won’t be as effective, he says.

      In the future, he and his collaborators want to conduct larger studies to evaluate the short- and long-term effects of onboarding. They also want to leverage unlabeled data for the onboarding process, and find methods to effectively reduce the number of regions without omitting important examples.

      “People are adopting AI systems willy-nilly, and indeed AI offers great potential, but these AI agents still sometimes makes mistakes. Thus, it’s crucial for AI developers to devise methods that help humans know when it’s safe to rely on the AI’s suggestions,” says Dan Weld, professor emeritus at the Paul G. Allen School of Computer Science and Engineering at the University of Washington, who was not involved with this research. “Mozannar et al. have created an innovative method for identifying situations where the AI is trustworthy, and (importantly) to describe them to people in a way that leads to better human-AI team interactions.”

      This work is funded, in part, by the MIT-IBM Watson AI Lab. More