More stories

  • in

    Large language models are biased. Can logic help save them?

    Turns out, even language models “think” they’re biased. When prompted in ChatGPT, the response was as follows: “Yes, language models can have biases, because the training data reflects the biases present in society from which that data was collected. For example, gender and racial biases are prevalent in many real-world datasets, and if a language model is trained on that, it can perpetuate and amplify these biases in its predictions.” A well-known but dangerous problem. 

    Humans (typically) can dabble with both logical and stereotypical reasoning when learning. Still, language models mainly mimic the latter, an unfortunate narrative we’ve seen play out ad nauseam when the ability to employ reasoning and critical thinking is absent. So would injecting logic into the fray be enough to mitigate such behavior? 

    Scientists from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) had an inkling that it might, so they set off to examine if logic-aware language models could significantly avoid more harmful stereotypes. They trained a language model to predict the relationship between two sentences, based on context and semantic meaning, using a dataset with labels for text snippets detailing if a second phrase “entails,” “contradicts,” or is neutral with respect to the first one. Using this dataset — natural language inference — they found that the newly trained models were significantly less biased than other baselines, without any extra data, data editing, or additional training algorithms.

    For example, with the premise “the person is a doctor” and the hypothesis “the person is masculine,” using these logic-trained models, the relationship would be classified as “neutral,” since there’s no logic that says the person is a man. With more common language models, two sentences might seem to be correlated due to some bias in training data, like “doctor” might be pinged with “masculine,” even when there’s no evidence that the statement is true. 

    At this point, the omnipresent nature of language models is well-known: Applications in natural language processing, speech recognition, conversational AI, and generative tasks abound. While not a nascent field of research, growing pains can take a front seat as they increase in complexity and capability. 

    “Current language models suffer from issues with fairness, computational resources, and privacy,” says MIT CSAIL postdoc Hongyin Luo, the lead author of a new paper about the work. “Many estimates say that the CO2 emission of training a language model can be higher than the lifelong emission of a car. Running these large language models is also very expensive because of the amount of parameters and the computational resources they need. With privacy, state-of-the-art language models developed by places like ChatGPT or GPT-3 have their APIs where you must upload your language, but there’s no place for sensitive information regarding things like health care or finance. To solve these challenges, we proposed a logical language model that we qualitatively measured as fair, is 500 times smaller than the state-of-the-art models, can be deployed locally, and with no human-annotated training samples for downstream tasks. Our model uses 1/400 the parameters compared with the largest language models, has better performance on some tasks, and significantly saves computation resources.” 

    This model, which has 350 million parameters, outperformed some very large-scale language models with 100 billion parameters on logic-language understanding tasks. The team evaluated, for example, popular BERT pretrained language models with their “textual entailment” ones on stereotype, profession, and emotion bias tests. The latter outperformed other models with significantly lower bias, while preserving the language modeling ability. The “fairness” was evaluated with something called ideal context association (iCAT) tests, where higher iCAT scores mean fewer stereotypes. The model had higher than 90 percent iCAT scores, while other strong language understanding models ranged between 40 to 80. 

    Luo wrote the paper alongside MIT Senior Research Scientist James Glass. They will present the work at the Conference of the European Chapter of the Association for Computational Linguistics in Croatia. 

    Unsurprisingly, the original pretrained language models the team examined were teeming with bias, confirmed by a slew of reasoning tests demonstrating how professional and emotion terms are significantly biased to the feminine or masculine words in the gender vocabulary. 

    With professions, a language model (which is biased) thinks that “flight attendant,” “secretary,” and “physician’s assistant” are feminine jobs, while “fisherman,” “lawyer,” and “judge” are masculine. Concerning emotions, a language model thinks that “anxious,” “depressed,” and “devastated” are feminine.

    While we may still be far away from a neutral language model utopia, this research is ongoing in that pursuit. Currently, the model is just for language understanding, so it’s based on reasoning among existing sentences. Unfortunately, it can’t generate sentences for now, so the next step for the researchers would be targeting the uber-popular generative models built with logical learning to ensure more fairness with computational efficiency. 

    “Although stereotypical reasoning is a natural part of human recognition, fairness-aware people conduct reasoning with logic rather than stereotypes when necessary,” says Luo. “We show that language models have similar properties. A language model without explicit logic learning makes plenty of biased reasoning, but adding logic learning can significantly mitigate such behavior. Furthermore, with demonstrated robust zero-shot adaptation ability, the model can be directly deployed to different tasks with more fairness, privacy, and better speed.” More

  • in

    Report: CHIPS Act just the first step in addressing threats to US leadership in advanced computing

    When Liu He, a Chinese economist, politician, and “chip czar,” was tapped to lead the charge in a chipmaking arms race with the United States, his message lingered in the air, leaving behind a dewy glaze of tension: “For our country, technology is not just for growth… it is a matter of survival.”

    Once upon a time, the United States’ early technological prowess positioned the nation to outpace foreign rivals and cultivate a competitive advantage for domestic businesses. Yet, 30 years later, America’s lead in advanced computing is continuing to wane. What happened?

    A new report from an MIT researcher and two colleagues sheds light on the decline in U.S. leadership. The scientists looked at high-level measures to examine the shrinkage: overall capabilities, supercomputers, applied algorithms, and semiconductor manufacturing. Through their analysis, they found that not only has China closed the computing gap with the U.S., but nearly 80 percent of American leaders in the field believe that their Chinese competitors are improving capabilities faster — which, the team says, suggests a “broad threat to U.S. competitiveness.”

    To delve deeply into the fray, the scientists conducted the Advanced Computing Users Survey, sampling 120 top-tier organizations, including universities, national labs, federal agencies, and industry. The team estimates that this group comprises one-third and one-half of all the most significant computing users in the United States.

    “Advanced computing is crucial to scientific improvement, economic growth and the competitiveness of U.S. companies,” says Neil Thompson, director of the FutureTech Research Project at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), who helped lead the study.

    Thompson, who is also a principal investigator at MIT’s Initiative on the Digital Economy, wrote the paper with Chad Evans, executive vice president and secretary and treasurer to the board at the Council on Competitiveness, and Daniel Armbrust, who is the co-founder, initial CEO, and member of the board of directors at Silicon Catalyst and former president of SEMATECH, the semiconductor consortium that developed industry roadmaps.

    The semiconductor, supercomputer, and algorithm bonanza

    Supercomputers — the room-sized, “giant calculators” of the hardware world — are an industry no longer dominated by the United States. Through 2015, about half of the most powerful computers were sitting firmly in the U.S., and China was growing slowly from a very slow base. But in the past six years, China has swiftly caught up, reaching near parity with America.

    This disappearing lead matters. Eighty-four percent of U.S. survey respondents said they’re computationally constrained in running essential programs. “This result was telling, given who our respondents are: the vanguard of American research enterprises and academic institutions with privileged access to advanced national supercomputing resources,” says Thompson. 

    With regards to advanced algorithms, historically, the U.S. has fronted the charge, with two-thirds of all significant improvements dominated by U.S.-born inventors. But in recent decades, U.S. dominance in algorithms has relied on bringing in foreign talent to work in the U.S., which the researchers say is now in jeopardy. China has outpaced the U.S. and many other countries in churning out PhDs in STEM fields since 2007, with one report postulating a near-distant future (2025) where China will be home to nearly twice as many PhDs than in the U.S. China’s rise in algorithms can also be seen with the “Gordon Bell Prize,” an achievement for outstanding work in harnessing the power of supercomputers in varied applications. U.S. winners historically dominated the prize, but China has now equaled or surpassed Americans’ performance in the past five years.

    While the researchers note the CHIPS and Science Act of 2022 is a critical step in re-establishing the foundation of success for advanced computing, they propose recommendations to the U.S. Office of Science and Technology Policy. 

    First, they suggest democratizing access to U.S. supercomputing by building more mid-tier systems that push boundaries for many users, as well as building tools so users scaling up computations can have less up-front resource investment. They also recommend increasing the pool of innovators by funding many more electrical engineers and computer scientists being trained with longer-term US residency incentives and scholarships. Finally, in addition to this new framework, the scientists urge taking advantage of what already exists, via providing the private sector access to experimentation with high-performance computing through supercomputing sites in academia and national labs.

    All that and a bag of chips

    Computing improvements depend on continuous advances in transistor density and performance, but creating robust, new chips necessitate a harmonious blend of design and manufacturing.

    Over the last six years, China was not known as the savants of noteworthy chips. In fact, in the past five decades, the U.S. designed most of them. But this changed in the past six years when China created the HiSilicon Kirin 9000, propelling itself to the international frontier. This success was mainly obtained through partnerships with leading global chip designers that began in the 2000s. Now, China now has 14 companies among the world’s top 50 fabless designers. A decade ago, there was only one. 

    Competitive semiconductor manufacturing has been more mixed, where U.S.-led policies and internal execution issues have slowed China’s rise, but as of July 2022, the Semiconductor Manufacturing International Corporation (SMIC) has evidence of 7 nanometer logic, which was not expected until much later. However, with extreme ultraviolet export restrictions, progress below 7 nm means domestic technology development would be expensive. Currently, China is only at parity or better in two out of 12 segments of the semiconductor supply chain. Still, with government policy and investments, the team expects a whopping increase to seven segments in 10 years. So, for the moment, the U.S. retains leadership in hardware manufacturing, but with fewer dimensions of advantage.

    The authors recommend that the White House Office of Science and Technology Policy work with key national agencies, such as the U.S. Department of Defense, U.S. Department of Energy, and the National Science Foundation, to define initiatives to build the hardware and software systems needed for important computing paradigms and workloads critical for economic and security goals. “It is crucial that American enterprises can get the benefit of faster computers,” says Thompson. “With Moore’s Law slowing down, the best way to do this is to create a portfolio of specialized chips (or “accelerators”) that are customized to our needs.”

    The scientists further believe that to lead the next generation of computing, four areas must be addressed. First, by issuing grand challenges to the CHIPS Act National Semiconductor Technology Center, researchers and startups would be motivated to invest in research and development and to seek startup capital for new technologies in areas such as spintronics, neuromorphics, optical and quantum computing, and optical interconnect fabrics. By supporting allies in passing similar acts, overall investment in these technologies would increase, and supply chains would become more aligned and secure. Establishing test beds for researchers to test algorithms on new computing architectures and hardware would provide an essential platform for innovation and discovery. Finally, planning for post-exascale systems that achieve higher levels of performance through next-generation advances would ensure that current commercial technologies don’t limit future computing systems.

    “The advanced computing landscape is in rapid flux — technologically, economically, and politically, with both new opportunities for innovation and rising global rivalries,” says Daniel Reed, Presidential Professor and professor of computer science and electrical and computer engineering at the University of Utah. “The transformational insights from both deep learning and computational modeling depend on both continued semiconductor advances and their instantiation in leading edge, large-scale computing systems — hyperscale clouds and high-performance computing systems. Although the U.S. has historically led the world in both advanced semiconductors and high-performance computing, other nations have recognized that these capabilities are integral to 21st century economic competitiveness and national security, and they are investing heavily.”

    The research was funded, in part, through Thompson’s grant from Good Ventures, which supports his FutureTech Research Group. The paper is being published by the Georgetown Public Policy Review. More

  • in

    MIT community members elected to the National Academy of Engineering for 2023

    Seven MIT researchers are among the 106 new members and 18 international members elected to the National Academy of Engineering (NAE) this week. Fourteen additional MIT alumni, including one member of the MIT Corporation, were also elected as new members.

    One of the highest professional distinctions for engineers, membership to the NAE is given to individuals who have made outstanding contributions to “engineering research, practice, or education, including, where appropriate, significant contributions to the engineering literature” and to “the pioneering of new and developing fields of technology, making major advancements in traditional fields of engineering, or developing/implementing innovative approaches to engineering education.”

    The seven MIT researchers elected this year include:

    Regina Barzilay, the School of Engineering Distinguished Professor for AI and Health in the Department of Electrical Engineering and Computer Science, principal investigator at the Computer Science and Artificial Intelligence Laboratory, and faculty lead for the MIT Abdul Latif Jameel Clinic for Machine Learning in Health, for machine learning models that understand structures in text, molecules, and medical images.

    Markus J. Buehler, the Jerry McAfee (1940) Professor in Engineering from the Department of Civil and Environmental Engineering, for implementing the use of nanomechanics to model and design fracture-resistant bioinspired materials.

    Elfatih A.B. Eltahir SM ’93, ScD ’93, the H.M. King Bhumibol Professor in the Department of Civil and Environmental Engineering, for advancing understanding of how climate and land use impact water availability, environmental and human health, and vector-borne diseases.

    Neil Gershenfeld, director of the Center for Bits and Atoms, for eliminating boundaries between digital and physical worlds, from quantum computing to digital materials to the internet of things.

    Roger D. Kamm SM ’73, PhD ’77, the Cecil and Ida Green Distinguished Professor of Biological and Mechanical Engineering, for contributions to the understanding of mechanics in biology and medicine, and leadership in biomechanics.

    David W. Miller ’82, SM ’85, ScD ’88, the Jerome C. Hunsaker Professor in the Department of Aeronautics and Astronautics, for contributions in control technology for space-based telescope design, and leadership in cross-agency guidance of space technology.

    David Simchi-Levi, professor of civil and environmental engineering, core faculty member in the Institute for Data, Systems, and Society, and principal investigator at the Laboratory for Information and Decision Systems, for contributions using optimization and stochastic modeling to enhance supply chain management and operations.

    Fariborz Maseeh ScD ’90, life member of the MIT Corporation and member of the School of Engineering Dean’s Advisory Council, was also elected as a member for leadership and advances in efficient design, development, and manufacturing of microelectromechanical systems, and for empowering engineering talent through public service.

    Thirteen additional alumni were elected to the National Academy of Engineering this year. They are: Mark George Allen SM ’86, PhD ’89; Shorya Awtar ScD ’04; Inderjit Chopra ScD ’77; David Huang ’85, SM ’89, PhD ’93; Eva Lerner-Lam SM ’78; David F. Merrion SM ’59; Virginia Norwood ’47; Martin Gerard Plys ’80, SM ’81, ScD ’84; Mark Prausnitz PhD ’94; Anil Kumar Sachdev ScD ’77; Christopher Scholz PhD ’67; Melody Ann Swartz PhD ’98; and Elias Towe ’80, SM ’81, PhD ’87.

    “I am delighted that seven members of MIT’s faculty and many members of the wider MIT community were elected to the National Academy of Engineering this year,” says Anantha Chandrakasan, the dean of the MIT School of Engineering and the Vannevar Bush Professor of Electrical Engineering and Computer Science. “My warmest congratulations on this recognition of their many contributions to engineering research and education.”

    Including this year’s inductees, 156 members of the National Academy of Engineering are current or retired members of the MIT faculty and staff, or members of the MIT Corporation. More

  • in

    Helping companies deploy AI models more responsibly

    Companies today are incorporating artificial intelligence into every corner of their business. The trend is expected to continue until machine-learning models are incorporated into most of the products and services we interact with every day.

    As those models become a bigger part of our lives, ensuring their integrity becomes more important. That’s the mission of Verta, a startup that spun out of MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL).

    Verta’s platform helps companies deploy, monitor, and manage machine-learning models safely and at scale. Data scientists and engineers can use Verta’s tools to track different versions of models, audit them for bias, test them before deployment, and monitor their performance in the real world.

    “Everything we do is to enable more products to be built with AI, and to do that safely,” Verta founder and CEO Manasi Vartak SM ’14, PhD ’18 says. “We’re already seeing with ChatGPT how AI can be used to generate data, artefacts — you name it — that look correct but aren’t correct. There needs to be more governance and control in how AI is being used, particularly for enterprises providing AI solutions.”

    Verta is currently working with large companies in health care, finance, and insurance to help them understand and audit their models’ recommendations and predictions. It’s also working with a number of high-growth tech companies looking to speed up deployment of new, AI-enabled solutions while ensuring those solutions are used appropriately.

    Vartak says the company has been able to decrease the time it takes customers to deploy AI models by orders of magnitude while ensuring those models are explainable and fair — an especially important factor for companies in highly regulated industries.

    Health care companies, for example, can use Verta to improve AI-powered patient monitoring and treatment recommendations. Such systems need to be thoroughly vetted for errors and biases before they’re used on patients.

    “Whether it’s bias or fairness or explainability, it goes back to our philosophy on model governance and management,” Vartak says. “We think of it like a preflight checklist: Before an airplane takes off, there’s a set of checks you need to do before you get your airplane off the ground. It’s similar with AI models. You need to make sure you’ve done your bias checks, you need to make sure there’s some level of explainability, you need to make sure your model is reproducible. We help with all of that.”

    From project to product

    Before coming to MIT, Vartak worked as a data scientist for a social media company. In one project, after spending weeks tuning machine-learning models that curated content to show in people’s feeds, she learned an ex-employee had already done the same thing. Unfortunately, there was no record of what they did or how it affected the models.

    For her PhD at MIT, Vartak decided to build tools to help data scientists develop, test, and iterate on machine-learning models. Working in CSAIL’s Database Group, Vartak recruited a team of graduate students and participants in MIT’s Undergraduate Research Opportunities Program (UROP).

    “Verta would not exist without my work at MIT and MIT’s ecosystem,” Vartak says. “MIT brings together people on the cutting edge of tech and helps us build the next generation of tools.”

    The team worked with data scientists in the CSAIL Alliances program to decide what features to build and iterated based on feedback from those early adopters. Vartak says the resulting project, named ModelDB, was the first open-source model management system.

    Vartak also took several business classes at the MIT Sloan School of Management during her PhD and worked with classmates on startups that recommended clothing and tracked health, spending countless hours in the Martin Trust Center for MIT Entrepreneurship and participating in the center’s delta v summer accelerator.

    “What MIT lets you do is take risks and fail in a safe environment,” Vartak says. “MIT afforded me those forays into entrepreneurship and showed me how to go about building products and finding first customers, so by the time Verta came around I had done it on a smaller scale.”

    ModelDB helped data scientists train and track models, but Vartak quickly saw the stakes were higher once models were deployed at scale. At that point, trying to improve (or accidentally breaking) models can have major implications for companies and society. That insight led Vartak to begin building Verta.

    “At Verta, we help manage models, help run models, and make sure they’re working as expected, which we call model monitoring,” Vartak explains. “All of those pieces have their roots back to MIT and my thesis work. Verta really evolved from my PhD project at MIT.”

    Verta’s platform helps companies deploy models more quickly, ensure they continue working as intended over time, and manage the models for compliance and governance. Data scientists can use Verta to track different versions of models and understand how they were built, answering questions like how data were used and which explainability or bias checks were run. They can also vet them by running them through deployment checklists and security scans.

    “Verta’s platform takes the data science model and adds half a dozen layers to it to transform it into something you can use to power, say, an entire recommendation system on your website,” Vartak says. “That includes performance optimizations, scaling, and cycle time, which is how quickly you can take a model and turn it into a valuable product, as well as governance.”

    Supporting the AI wave

    Vartak says large companies often use thousands of different models that influence nearly every part of their operations.

    “An insurance company, for example, will use models for everything from underwriting to claims, back-office processing, marketing, and sales,” Vartak says. “So, the diversity of models is really high, there’s a large volume of them, and the level of scrutiny and compliance companies need around these models are very high. They need to know things like: Did you use the data you were supposed to use? Who were the people who vetted it? Did you run explainability checks? Did you run bias checks?”

    Vartak says companies that don’t adopt AI will be left behind. The companies that ride AI to success, meanwhile, will need well-defined processes in place to manage their ever-growing list of models.

    “In the next 10 years, every device we interact with is going to have intelligence built in, whether it’s a toaster or your email programs, and it’s going to make your life much, much easier,” Vartak says. “What’s going to enable that intelligence are better models and software, like Verta, that help you integrate AI into all of these applications very quickly.” More

  • in

    Putting clear bounds on uncertainty

    In science and technology, there has been a long and steady drive toward improving the accuracy of measurements of all kinds, along with parallel efforts to enhance the resolution of images. An accompanying goal is to reduce the uncertainty in the estimates that can be made, and the inferences drawn, from the data (visual or otherwise) that have been collected. Yet uncertainty can never be wholly eliminated. And since we have to live with it, at least to some extent, there is much to be gained by quantifying the uncertainty as precisely as possible.

    Expressed in other terms, we’d like to know just how uncertain our uncertainty is.

    That issue was taken up in a new study, led by Swami Sankaranarayanan, a postdoc at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), and his co-authors — Anastasios Angelopoulos and Stephen Bates of the University of California at Berkeley; Yaniv Romano of Technion, the Israel Institute of Technology; and Phillip Isola, an associate professor of electrical engineering and computer science at MIT. These researchers succeeded not only in obtaining accurate measures of uncertainty, they also found a way to display uncertainty in a manner the average person could grasp.

    Their paper, which was presented in December at the Neural Information Processing Systems Conference in New Orleans, relates to computer vision — a field of artificial intelligence that involves training computers to glean information from digital images. The focus of this research is on images that are partially smudged or corrupted (due to missing pixels), as well as on methods — computer algorithms, in particular — that are designed to uncover the part of the signal that is marred or otherwise concealed. An algorithm of this sort, Sankaranarayanan explains, “takes the blurred image as the input and gives you a clean image as the output” — a process that typically occurs in a couple of steps.

    First, there is an encoder, a kind of neural network specifically trained by the researchers for the task of de-blurring fuzzy images. The encoder takes a distorted image and, from that, creates an abstract (or “latent”) representation of a clean image in a form — consisting of a list of numbers — that is intelligible to a computer but would not make sense to most humans. The next step is a decoder, of which there are a couple of types, that are again usually neural networks. Sankaranarayanan and his colleagues worked with a kind of decoder called a “generative” model. In particular, they used an off-the-shelf version called StyleGAN, which takes the numbers from the encoded representation (of a cat, for instance) as its input and then constructs a complete, cleaned-up image (of that particular cat). So the entire process, including the encoding and decoding stages, yields a crisp picture from an originally muddied rendering.

    But how much faith can someone place in the accuracy of the resultant image? And, as addressed in the December 2022 paper, what is the best way to represent the uncertainty in that image? The standard approach is to create a “saliency map,” which ascribes a probability value — somewhere between 0 and 1 — to indicate the confidence the model has in the correctness of every pixel, taken one at a time. This strategy has a drawback, according to Sankaranarayanan, “because the prediction is performed independently for each pixel. But meaningful objects occur within groups of pixels, not within an individual pixel,” he adds, which is why he and his colleagues are proposing an entirely different way of assessing uncertainty.

    Their approach is centered around the “semantic attributes” of an image — groups of pixels that, when taken together, have meaning, making up a human face, for example, or a dog, or some other recognizable thing. The objective, Sankaranarayanan maintains, “is to estimate uncertainty in a way that relates to the groupings of pixels that humans can readily interpret.”

    Whereas the standard method might yield a single image, constituting the “best guess” as to what the true picture should be, the uncertainty in that representation is normally hard to discern. The new paper argues that for use in the real world, uncertainty should be presented in a way that holds meaning for people who are not experts in machine learning. Rather than producing a single image, the authors have devised a procedure for generating a range of images — each of which might be correct. Moreover, they can set precise bounds on the range, or interval, and provide a probabilistic guarantee that the true depiction lies somewhere within that range. A narrower range can be provided if the user is comfortable with, say, 90 percent certitude, and a narrower range still if more risk is acceptable.

    The authors believe their paper puts forth the first algorithm, designed for a generative model, which can establish uncertainty intervals that relate to meaningful (semantically-interpretable) features of an image and come with “a formal statistical guarantee.” While that is an important milestone, Sankaranarayanan considers it merely a step toward “the ultimate goal. So far, we have been able to do this for simple things, like restoring images of human faces or animals, but we want to extend this approach into more critical domains, such as medical imaging, where our ‘statistical guarantee’ could be especially important.”

    Suppose that the film, or radiograph, of a chest X-ray is blurred, he adds, “and you want to reconstruct the image. If you are given a range of images, you want to know that the true image is contained within that range, so you are not missing anything critical” — information that might reveal whether or not a patient has lung cancer or pneumonia. In fact, Sankaranarayanan and his colleagues have already begun working with a radiologist to see if their algorithm for predicting pneumonia could be useful in a clinical setting.

    Their work may also have relevance in the law enforcement field, he says. “The picture from a surveillance camera may be blurry, and you want to enhance that. Models for doing that already exist, but it is not easy to gauge the uncertainty. And you don’t want to make a mistake in a life-or-death situation.” The tools that he and his colleagues are developing could help identify a guilty person and help exonerate an innocent one as well.

    Much of what we do and many of the things happening in the world around us are shrouded in uncertainty, Sankaranarayanan notes. Therefore, gaining a firmer grasp of that uncertainty could help us in countless ways. For one thing, it can tell us more about exactly what it is we do not know.

    Angelopoulos was supported by the National Science Foundation. Bates was supported by the Foundations of Data Science Institute and the Simons Institute. Romano was supported by the Israel Science Foundation and by a Career Advancement Fellowship from Technion. Sankaranarayanan’s and Isola’s research for this project was sponsored by the U.S. Air Force Research Laboratory and the U.S. Air Force Artificial Intelligence Accelerator and was accomplished under Cooperative Agreement Number FA8750-19-2- 1000. MIT SuperCloud and the Lincoln Laboratory Supercomputing Center also provided computing resources that contributed to the results reported in this work. More

  • in

    Unpacking the “black box” to build better AI models

    When deep learning models are deployed in the real world, perhaps to detect financial fraud from credit card activity or identify cancer in medical images, they are often able to outperform humans.

    But what exactly are these deep learning models learning? Does a model trained to spot skin cancer in clinical images, for example, actually learn the colors and textures of cancerous tissue, or is it flagging some other features or patterns?

    These powerful machine-learning models are typically based on artificial neural networks that can have millions of nodes that process data to make predictions. Due to their complexity, researchers often call these models “black boxes” because even the scientists who build them don’t understand everything that is going on under the hood.

    Stefanie Jegelka isn’t satisfied with that “black box” explanation. A newly tenured associate professor in the MIT Department of Electrical Engineering and Computer Science, Jegelka is digging deep into deep learning to understand what these models can learn and how they behave, and how to build certain prior information into these models.

    “At the end of the day, what a deep-learning model will learn depends on so many factors. But building an understanding that is relevant in practice will help us design better models, and also help us understand what is going on inside them so we know when we can deploy a model and when we can’t. That is critically important,” says Jegelka, who is also a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL) and the Institute for Data, Systems, and Society (IDSS).

    Jegelka is particularly interested in optimizing machine-learning models when input data are in the form of graphs. Graph data pose specific challenges: For instance, information in the data consists of both information about individual nodes and edges, as well as the structure — what is connected to what. In addition, graphs have mathematical symmetries that need to be respected by the machine-learning model so that, for instance, the same graph always leads to the same prediction. Building such symmetries into a machine-learning model is usually not easy.

    Take molecules, for instance. Molecules can be represented as graphs, with vertices that correspond to atoms and edges that correspond to chemical bonds between them. Drug companies may want to use deep learning to rapidly predict the properties of many molecules, narrowing down the number they must physically test in the lab.

    Jegelka studies methods to build mathematical machine-learning models that can effectively take graph data as an input and output something else, in this case a prediction of a molecule’s chemical properties. This is particularly challenging since a molecule’s properties are determined not only by the atoms within it, but also by the connections between them.  

    Other examples of machine learning on graphs include traffic routing, chip design, and recommender systems.

    Designing these models is made even more difficult by the fact that data used to train them are often different from data the models see in practice. Perhaps the model was trained using small molecular graphs or traffic networks, but the graphs it sees once deployed are larger or more complex.

    In this case, what can researchers expect this model to learn, and will it still work in practice if the real-world data are different?

    “Your model is not going to be able to learn everything because of some hardness problems in computer science, but what you can learn and what you can’t learn depends on how you set the model up,” Jegelka says.

    She approaches this question by combining her passion for algorithms and discrete mathematics with her excitement for machine learning.

    From butterflies to bioinformatics

    Jegelka grew up in a small town in Germany and became interested in science when she was a high school student; a supportive teacher encouraged her to participate in an international science competition. She and her teammates from the U.S. and Singapore won an award for a website they created about butterflies, in three languages.

    “For our project, we took images of wings with a scanning electron microscope at a local university of applied sciences. I also got the opportunity to use a high-speed camera at Mercedes Benz — this camera usually filmed combustion engines — which I used to capture a slow-motion video of the movement of a butterfly’s wings. That was the first time I really got in touch with science and exploration,” she recalls.

    Intrigued by both biology and mathematics, Jegelka decided to study bioinformatics at the University of Tübingen and the University of Texas at Austin. She had a few opportunities to conduct research as an undergraduate, including an internship in computational neuroscience at Georgetown University, but wasn’t sure what career to follow.

    When she returned for her final year of college, Jegelka moved in with two roommates who were working as research assistants at the Max Planck Institute in Tübingen.

    “They were working on machine learning, and that sounded really cool to me. I had to write my bachelor’s thesis, so I asked at the institute if they had a project for me. I started working on machine learning at the Max Planck Institute and I loved it. I learned so much there, and it was a great place for research,” she says.

    She stayed on at the Max Planck Institute to complete a master’s thesis, and then embarked on a PhD in machine learning at the Max Planck Institute and the Swiss Federal Institute of Technology.

    During her PhD, she explored how concepts from discrete mathematics can help improve machine-learning techniques.

    Teaching models to learn

    The more Jegelka learned about machine learning, the more intrigued she became by the challenges of understanding how models behave, and how to steer this behavior.

    “You can do so much with machine learning, but only if you have the right model and data. It is not just a black-box thing where you throw it at the data and it works. You actually have to think about it, its properties, and what you want the model to learn and do,” she says.

    After completing a postdoc at the University of California at Berkeley, Jegelka was hooked on research and decided to pursue a career in academia. She joined the faculty at MIT in 2015 as an assistant professor.

    “What I really loved about MIT, from the very beginning, was that the people really care deeply about research and creativity. That is what I appreciate the most about MIT. The people here really value originality and depth in research,” she says.

    That focus on creativity has enabled Jegelka to explore a broad range of topics.

    In collaboration with other faculty at MIT, she studies machine-learning applications in biology, imaging, computer vision, and materials science.

    But what really drives Jegelka is probing the fundamentals of machine learning, and most recently, the issue of robustness. Often, a model performs well on training data, but its performance deteriorates when it is deployed on slightly different data. Building prior knowledge into a model can make it more reliable, but understanding what information the model needs to be successful and how to build it in is not so simple, she says.

    She is also exploring methods to improve the performance of machine-learning models for image classification.

    Image classification models are everywhere, from the facial recognition systems on mobile phones to tools that identify fake accounts on social media. These models need massive amounts of data for training, but since it is expensive for humans to hand-label millions of images, researchers often use unlabeled datasets to pretrain models instead.

    These models then reuse the representations they have learned when they are fine-tuned later for a specific task.

    Ideally, researchers want the model to learn as much as it can during pretraining, so it can apply that knowledge to its downstream task. But in practice, these models often learn only a few simple correlations — like that one image has sunshine and one has shade — and use these “shortcuts” to classify images.

    “We showed that this is a problem in ‘contrastive learning,’ which is a standard technique for pre-training, both theoretically and empirically. But we also show that you can influence the kinds of information the model will learn to represent by modifying the types of data you show the model. This is one step toward understanding what models are actually going to do in practice,” she says.

    Researchers still don’t understand everything that goes on inside a deep-learning model, or details about how they can influence what a model learns and how it behaves, but Jegelka looks forward to continue exploring these topics.

    “Often in machine learning, we see something happen in practice and we try to understand it theoretically. This is a huge challenge. You want to build an understanding that matches what you see in practice, so that you can do better. We are still just at the beginning of understanding this,” she says.

    Outside the lab, Jegelka is a fan of music, art, traveling, and cycling. But these days, she enjoys spending most of her free time with her preschool-aged daughter. More

  • in

    Simulating discrimination in virtual reality

    Have you ever been advised to “walk a mile in someone else’s shoes?” Considering another person’s perspective can be a challenging endeavor — but recognizing our errors and biases is key to building understanding across communities. By challenging our preconceptions, we confront prejudice, such as racism and xenophobia, and potentially develop a more inclusive perspective about others.

    To assist with perspective-taking, MIT researchers have developed “On the Plane,” a virtual reality role-playing game (VR RPG) that simulates discrimination. In this case, the game portrays xenophobia directed against a Malaysian America woman, but the approach can be generalized. Situated on an airplane, players can take on the role of characters from different backgrounds, engaging in dialogue with others while making in-game choices to a series of prompts. In turn, players’ decisions control the outcome of a tense conversation between the characters about cultural differences.

    As a VR RPG, “On the Plane” encourages players to take on new roles that may be outside of their personal experiences in the first person, allowing them to confront in-group/out-group bias by incorporating new perspectives into their understanding of different cultures. Players engage with three characters: Sarah, a first-generation Muslim American of Malaysian ancestry who wears a hijab; Marianne, a white woman from the Midwest with little exposure to other cultures and customs; or a flight attendant. Sarah represents the out group, Marianne is a member of the in group, and the flight staffer is a bystander witnessing an exchange between the two passengers.“This project is part of our efforts to harness the power of virtual reality and artificial intelligence to address social ills, such as discrimination and xenophobia,” says Caglar Yildirim, an MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) research scientist who is a co-author and co-game designer on the project. “Through the exchange between the two passengers, players experience how one passenger’s xenophobia manifests itself and how it affects the other passenger. The simulation engages players in critical reflection and seeks to foster empathy for the passenger who was ‘othered’ due to her outfit being not so ‘prototypical’ of what an American should look like.”

    Yildirim worked alongside the project’s principal investigator, D. Fox Harrell, MIT professor of digital media and AI at CSAIL, the Program in Comparative Media Studies/Writing (CMS), and the Institute for Data, Systems, and Society (IDSS) and founding director of the MIT Center for Advanced Virtuality. “It is not possible for a simulation to give someone the life experiences of another person, but while you cannot ‘walk in someone else’s shoes’ in that sense, a system like this can help people recognize and understand the social patterns at work when it comes to issue like bias,” says Harrell, who is also co-author and designer on this project. “An engaging, immersive, interactive narrative can also impact people emotionally, opening the door for users’ perspectives to be transformed and broadened.” This simulation also utilizes an interactive narrative engine that creates several options for responses to in-game interactions based on a model of how people are categorized socially. The tool grants players a chance to alter their standing in the simulation through their reply choices to each prompt, affecting their affinity toward the other two characters. For example, if you play as the flight attendant, you can react to Marianne’s xenophobic expressions and attitudes toward Sarah, changing your affinities. The engine will then provide you with a different set of narrative events based on your changes in standing with others.

    To animate each avatar, “On the Plane” incorporates artificial intelligence knowledge representation techniques controlled by probabilistic finite state machines, a tool commonly used in machine learning systems for pattern recognition. With the help of these machines, characters’ body language and gestures are customizable: if you play as Marianne, the game will customize her mannerisms toward Sarah based on user inputs, impacting how comfortable she appears in front of a member of a perceived out group. Similarly, players can do the same from Sarah or the flight attendant’s point of view.In a 2018 paper based on work done in a collaboration between MIT CSAIL and the Qatar Computing Research Institute, Harrell and co-author Sercan Şengün advocated for virtual system designers to be more inclusive of Middle Eastern identities and customs. They claimed that if designers allowed users to customize virtual avatars more representative of their background, it might empower players to engage in a more supportive experience. Four years later, “On the Plane” accomplishes a similar goal, incorporating a Muslim’s perspective into an immersive environment.

    “Many virtual identity systems, such as avatars, accounts, profiles, and player characters, are not designed to serve the needs of people across diverse cultures. We have used statistical and AI methods in conjunction with qualitative approaches to learn where the gaps are,” they note. “Our project helps engender perspective transformation so that people will treat each other with respect and enhanced understanding across diverse cultural avatar representations.”

    Harrell and Yildirim’s work is part of the MIT IDSS’s Initiative on Combatting Systemic Racism (ICSR). Harrell is on the initiative’s steering committee and is the leader of the newly forming Antiracism, Games, and Immersive Media vertical, who study behavior, cognition, social phenomena, and computational systems related to race and racism in video games and immersive experiences.

    The researchers’ latest project is part of the ICSR’s broader goal to launch and coordinate cross-disciplinary research that addresses racially discriminatory processes across American institutions. Using big data, members of the research initiative develop and employ computing tools that drive racial equity. Yildirim and Harrell accomplish this goal by depicting a frequent, problematic scenario that illustrates how bias creeps into our everyday lives.“In a post-9/11 world, Muslims often experience ethnic profiling in American airports. ‘On the Plane’ builds off of that type of in-group favoritism, a well-established finding in psychology,” says MIT Professor Fotini Christia, director of the Sociotechnical Systems Research Center (SSRC) and associate director or IDSS. “This game also takes a novel approach to analyzing hardwired bias by utilizing VR instead of field experiments to simulate prejudice. Excitingly, this research demonstrates that VR can be used as a tool to help us better measure bias, combating systemic racism and other forms of discrimination.”“On the Plane” was developed on the Unity game engine using the XR Interaction Toolkit and Harrell’s Chimeria platform for authoring interactive narratives that involve social categorization. The game will be deployed for research studies later this year on both desktop computers and the standalone, wireless Meta Quest headsets. A paper on the work was presented in December at the 2022 IEEE International Conference on Artificial Intelligence and Virtual Reality. More

  • in

    Subtle biases in AI can influence emergency decisions

    It’s no secret that people harbor biases — some unconscious, perhaps, and others painfully overt. The average person might suppose that computers — machines typically made of plastic, steel, glass, silicon, and various metals — are free of prejudice. While that assumption may hold for computer hardware, the same is not always true for computer software, which is programmed by fallible humans and can be fed data that is, itself, compromised in certain respects.

    Artificial intelligence (AI) systems — those based on machine learning, in particular — are seeing increased use in medicine for diagnosing specific diseases, for example, or evaluating X-rays. These systems are also being relied on to support decision-making in other areas of health care. Recent research has shown, however, that machine learning models can encode biases against minority subgroups, and the recommendations they make may consequently reflect those same biases.

    A new study by researchers from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and the MIT Jameel Clinic, which was published last month in Communications Medicine, assesses the impact that discriminatory AI models can have, especially for systems that are intended to provide advice in urgent situations. “We found that the manner in which the advice is framed can have significant repercussions,” explains the paper’s lead author, Hammaad Adam, a PhD student at MIT’s Institute for Data Systems and Society. “Fortunately, the harm caused by biased models can be limited (though not necessarily eliminated) when the advice is presented in a different way.” The other co-authors of the paper are Aparna Balagopalan and Emily Alsentzer, both PhD students, and the professors Fotini Christia and Marzyeh Ghassemi.

    AI models used in medicine can suffer from inaccuracies and inconsistencies, in part because the data used to train the models are often not representative of real-world settings. Different kinds of X-ray machines, for instance, can record things differently and hence yield different results. Models trained predominately on white people, moreover, may not be as accurate when applied to other groups. The Communications Medicine paper is not focused on issues of that sort but instead addresses problems that stem from biases and on ways to mitigate the adverse consequences.

    A group of 954 people (438 clinicians and 516 nonexperts) took part in an experiment to see how AI biases can affect decision-making. The participants were presented with call summaries from a fictitious crisis hotline, each involving a male individual undergoing a mental health emergency. The summaries contained information as to whether the individual was Caucasian or African American and would also mention his religion if he happened to be Muslim. A typical call summary might describe a circumstance in which an African American man was found at home in a delirious state, indicating that “he has not consumed any drugs or alcohol, as he is a practicing Muslim.” Study participants were instructed to call the police if they thought the patient was likely to turn violent; otherwise, they were encouraged to seek medical help.

    The participants were randomly divided into a control or “baseline” group plus four other groups designed to test responses under slightly different conditions. “We want to understand how biased models can influence decisions, but we first need to understand how human biases can affect the decision-making process,” Adam notes. What they found in their analysis of the baseline group was rather surprising: “In the setting we considered, human participants did not exhibit any biases. That doesn’t mean that humans are not biased, but the way we conveyed information about a person’s race and religion, evidently, was not strong enough to elicit their biases.”

    The other four groups in the experiment were given advice that either came from a biased or unbiased model, and that advice was presented in either a “prescriptive” or a “descriptive” form. A biased model would be more likely to recommend police help in a situation involving an African American or Muslim person than would an unbiased model. Participants in the study, however, did not know which kind of model their advice came from, or even that models delivering the advice could be biased at all. Prescriptive advice spells out what a participant should do in unambiguous terms, telling them they should call the police in one instance or seek medical help in another. Descriptive advice is less direct: A flag is displayed to show that the AI system perceives a risk of violence associated with a particular call; no flag is shown if the threat of violence is deemed small.  

    A key takeaway of the experiment is that participants “were highly influenced by prescriptive recommendations from a biased AI system,” the authors wrote. But they also found that “using descriptive rather than prescriptive recommendations allowed participants to retain their original, unbiased decision-making.” In other words, the bias incorporated within an AI model can be diminished by appropriately framing the advice that’s rendered. Why the different outcomes, depending on how advice is posed? When someone is told to do something, like call the police, that leaves little room for doubt, Adam explains. However, when the situation is merely described — classified with or without the presence of a flag — “that leaves room for a participant’s own interpretation; it allows them to be more flexible and consider the situation for themselves.”

    Second, the researchers found that the language models that are typically used to offer advice are easy to bias. Language models represent a class of machine learning systems that are trained on text, such as the entire contents of Wikipedia and other web material. When these models are “fine-tuned” by relying on a much smaller subset of data for training purposes — just 2,000 sentences, as opposed to 8 million web pages — the resultant models can be readily biased.  

    Third, the MIT team discovered that decision-makers who are themselves unbiased can still be misled by the recommendations provided by biased models. Medical training (or the lack thereof) did not change responses in a discernible way. “Clinicians were influenced by biased models as much as non-experts were,” the authors stated.

    “These findings could be applicable to other settings,” Adam says, and are not necessarily restricted to health care situations. When it comes to deciding which people should receive a job interview, a biased model could be more likely to turn down Black applicants. The results could be different, however, if instead of explicitly (and prescriptively) telling an employer to “reject this applicant,” a descriptive flag is attached to the file to indicate the applicant’s “possible lack of experience.”

    The implications of this work are broader than just figuring out how to deal with individuals in the midst of mental health crises, Adam maintains.  “Our ultimate goal is to make sure that machine learning models are used in a fair, safe, and robust way.” More