Tool predicts how fast code will run on a chip

Tool predicts how fast code will run on a chip

Trending Technologies   January 06, 2020

The machine-learning system should enable developers to improve computing efficiency in a range of applications.

MIT researchers have invented a machine-learning tool that predicts how fast computer chips will execute code from various applications.  

To get code to run as fast as possible, developers and compilers — programs that translate programming language into machine-readable code — typically use performance models that run the code through a simulation of given chip architectures. 

Compilers use that information to automatically optimise code, and developers use it to tackle performance bottlenecks on the microprocessors that will run it. But performance models for machine code are handwritten by a relatively small group of experts and are not adequately validated. As a consequence, the simulated performance measurements often deviate from real-life results. 

In a series of conference papers, the researchers describe a novel machine-learning pipeline that automates this process, making it easier, faster, and more accurate. In a paper presented at the International Conference on Machine Learning in June, the researchers presented Ithemal, a neural-network model that trains on labelled data in the form of “basic blocks” — fundamental snippets of computing instructions — to automatically predict how long it takes a given chip to execute previously unseen basic blocks. Results suggest Ithemal performs far more accurately than traditional hand-tuned models. 

Then, at the November IEEE International Symposium on Workload Characterization, the researchers presented a benchmark suite of basic blocks from a variety of domains, including machine learning, compilers, cryptography, and graphics that can be used to validate performance models. They pooled more than 300,000 of the profiled blocks into an open-source dataset called BHive. During their evaluations, Ithemal predicted how fast Intel chips would run code even better than a performance model built by Intel itself. 

Ultimately, developers and compilers can use the tool to generate code that runs faster and more efficiently on an ever-growing number of diverse and “black box” chip designs. “Modern computer processors are opaque, horrendously complicated, and difficult to understand. It is also incredibly challenging to write computer code that executes as fast as possible for these processors,” says co-author Michael Carbin, an assistant professor in the Department of Electrical Engineering and Computer Science (EECS) and a researcher in the Computer Science and Artificial Intelligence Laboratory (CSAIL). “This tool is a big step forward toward fully modelling the performance of these chips for improved efficiency.”

Most recently, in a paper presented at the NeurIPS conference in December, the team proposed a new technique to automatically generate compiler optimisations.  Specifically, they automatically generate an algorithm, called Vemal, that converts certain code into vectors, which can be used for parallel computing. Vemal outperforms hand-crafted vectorisation algorithms used in the LLVM compiler — a popular compiler used in the industry.

Learning from data

Designing performance models by hand can be “a black art,” Carbin says. Intel provides extensive documentation of more than 3,000 pages describing its chips’ architectures. But there currently exists only a small group of experts who will build performance models that simulate the execution of code on those architectures. 

“Intel’s documents are neither error-free nor complete, and Intel will omit certain things, because it’s proprietary,” Mendis says. “However, when you use data, you don’t need to know the documentation. If there’s something hidden, you can learn it directly from the data.”

To do so, the researchers clocked the average number of cycles a given microprocessor takes to compute basic block instructions — basically, the sequence of boot-up, execute, and shut down — without human intervention. Automating the process enables rapid profiling of hundreds of thousands or millions of blocks. 

Domain-specific architectures

In training, the Ithemal model analyses millions of automatically profiled basic blocks to learn exactly how different chip architectures will execute computation. Importantly, Ithemal takes raw text as input and does not require manually adding features to the input data. In testing, Ithemal can be fed previously unseen basic blocks and a given chip and will generate a single number indicating how fast the chip will execute that code. 

The researchers found Ithemal cut error rates in accuracy — meaning the difference between the predicted speed versus real-world speed — by 50 per cent over traditional hand-crafted models. Further, in their next paper, they showed that Ithemal’s error rate was 10 per cent, while the Intel performance-prediction model’s error rate was 20 per cent on a variety of basic blocks across multiple different domains.

The tool now makes it easier to quickly learn performance speeds for any new chip architectures, Mendis says. For instance, domain-specific architectures, such as Google’s new Tensor Processing Unit used specifically for neural networks, are now being built but aren’t widely understood. “If you want to train a model on some new architecture, you just collect more data from that architecture, run it through our profiler, use that information to train Ithemal, and now you have a model that predicts performance,” Mendis says.

Next, the researchers are studying methods to make models interpretable. Much of machine learning is a black box, so it’s not really clear why a particular model made its predictions. “Our model is saying it takes a processor, say, 10 cycles to execute a basic block. Now, we’re trying to figure out why,” Carbin says. “That’s a fine level of granularity that would be amazing for these types of tools.”

They also hope to use Ithemal to enhance the performance of Vemal even further and achieve better performance automatically.

This article was originally published by MIT News. Follow this link to read related articles.

Specialized AI Chips Hold Both Promise and Peril for Developers

Specialized AI Chips Hold Both Promise and Peril for Developers

Trending Technologies   August 12, 2019

When it comes to the compute-intensive field of AI, hardware vendors are reviving the performance gains we enjoyed at the height of Moore’s Law. The gains come from a new generation of specialized chips for AI applications like deep learning. But the fragmented microchip marketplace that’s emerging will lead to some hard choices for developers. 

The new era of chip specialization for AI began when graphics processing units (GPUs), which were originally developed for gaming, were deployed for applications like deep learning. The same architecture that made GPUs render realistic images also enabled them to crunch data much more efficiently than central processing units (CPUs). A big step forward happened in 2007 when Nvidia released CUDA, a toolkit for making GPUs programmable in a general-purpose way.

AI researchers need every advantage they can get when dealing with the unprecedented computational requirements of deep learning. GPU processing power has advanced rapidly, and chips originally designed to render images have become the workhorses powering world-changing AI research and development. Many of the linear algebra routines that are necessary to make Fortnite run at 120 frames per second are now powering the neural networks at the heart of cutting-edge applications of computer vision, automated speech recognition, and natural language processing.  

Now, the trend toward microchip specialization is turning into an arms race. Gartner projects that specialized chip sales for AI will double to around the US $8 billion in 2019 and reach more than $34 billion by 2023. Nvidia’s internal projections place the market for data centre GPUs (which are almost solely used to power deep learning) at $50 billion in the same time frame. In the next five years, we’ll see massive investments in custom silicon come to fruition from Amazon, ARM, Apple, IBM, Intel, Google, Microsoft, Nvidia, Qualcomm. There is also a slew of startups in the mix. CrunchBase estimates that AI chip companies, including Cerebras, Graphcore, Groq, Mythic AI, SambaNova Systems, and Wave Computing, have collectively raised more than $1 billion. 

To be clear, specialized AI chips are both important and welcomed, as they’re catalysts for transforming cutting-edge AI research into real-world applications. However, the flood of new AI chips, each one faster and more specialized than the next, will also seem like a throwback to the rise of enterprise software. We can expect cut-throat sales deals and software specialization aimed at locking developers into working with just one vendor. 

Imagine if, 15 years ago, the cloud services AWS, Azure, Box, Dropbox, and GCP all came to market within 12 to 18 months. Their mission would have been to lock in as many businesses as possible—because once you’re on one platform, it’s hard to switch to another. This type of end-user gold rush is about to happen in AI, with tens of billions of dollars, and priceless research, at stake. 

Chipmakers won’t be short on promises, and the benefits will be real. But it’s important for AI developers to understand that new chips that require new architectures could make their products slower to market—even with faster performance. In most cases, AI models are not going to be portable between different chip makers. Developers are well aware of the vendor lock-in risk posed by adopting higher-level cloud APIs, but in the past, the actual compute substrate has been standardized and homogeneous. This situation is going to change dramatically in the world of AI development.

It's quite likely that more than half of the chip industry’s revenue will soon be driven by AI and deep learning applications. Just as software begets more software, AI begets more AI. We’ve seen it many times: Companies initially focus on one problem, but ultimately solve many. For example, major automakers are striving to bring autonomous cars to the road, and their cutting-edge work in deep learning and computer vision is already having a cascading effect; the research is leading to such offshoot projects as Ford’s delivery robots.

As specialized AI chips come to market, the current chip giants and major cloud companies will probably strike exclusive deals or acquire top performing startups. This trend will fragment the AI market rather than unifying it. All that AI developers can do now is understand what’s about to happen and plan how they’ll weigh the benefits of a faster chip with the costs of building on new architectures.

Evan Sparks is CEO of Determined AI. He holds a PhD in computer science from the University of California, Berkeley, where his research focused on distributed systems for data analysis and machine learning.

Drag-and-drop data analytics

Drag-and-drop data analytics

Trending Technologies   August 04, 2019

The system lets nonspecialists use machine-learning models to make predictions for medical research, sales, and more.

In the Iron Man movies, Tony Stark uses a holographic computer to project 3-D data into thin air, manipulate them with his hands, and find fixes to his superhero troubles. In the same vein, researchers from MIT and Brown University have now developed a system for interactive data analytics that runs on touchscreens and lets everyone — not just billionaire tech geniuses — tackle real-world issues.

For years, the researchers have been developing an interactive data-science system called Northstar, which runs in the cloud but has an interface that supports any touchscreen device, including smartphones and large interactive whiteboards. Users feed the system datasets, and manipulate, combine, and extract features on a user-friendly interface, using their fingers or a digital pen, to uncover trends and patterns.

In a paper being presented at the ACM SIGMOD conference, the researchers detail a new component of Northstar, called VDS for “virtual data scientist,” that instantly generates machine-learning models to run prediction tasks on their datasets. Doctors, for instance, can use the system to help predict which patients are more likely to have certain diseases, while business owners might want to forecast sales. If using an interactive whiteboard, everyone can also collaborate in real-time.

The aim is to democratize data science by making it easy to do complex analytics, quickly and accurately.

“Even a coffee shop owner who doesn’t know data science should be able to predict their sales over the next few weeks to figure out how much coffee to buy,” says co-author and long-time Northstar project lead Tim Kraska, an associate professor of electrical engineering and computer science in at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and founding co-director of the new Data System and AI Lab (DSAIL). “In companies that have data scientists, there’s a lot of back and forth between data scientists and nonexperts, so we can also bring them into one room to do analytics together.”

VDS is based on an increasingly popular technique in artificial intelligence called automated machine-learning (AutoML), which lets people with limited data-science know-how train AI models to make predictions based on their datasets. Currently, the tool leads the DARPA D3M Automatic Machine Learning competition, which every six months decide on the best-performing AutoML tool. 

Joining Kraska on the paper are: first author Zeyuan Shang, a graduate student, and Emanuel Zgraggen, a postdoc and main contributor of Northstar, both of EECS, CSAIL, and DSAIL; Benedetto Buratti, Yeounoh Chung, Philipp Eichmann, and Eli Upfal, all of Brown; and Carsten Binnig who recently moved from Brown to the Technical University of Darmstadt in Germany.

An “unbounded canvas” for analytics

The new work builds on years of collaboration on Northstar between researchers at MIT and Brown. Over four years, the researchers have published numerous papers detailing components of Northstar, including the interactive interface, operations on multiple platforms, accelerating results, and studies on user behavior.

Northstar starts as a blank, white interface. Users upload datasets into the system, which appear in a “datasets” box on the left. Any data labels will automatically populate a separate “attributes” box below. There’s also an “operators” box that contains various algorithms, as well as the new AutoML tool. All data are stored and analyzed in the cloud.

The researchers like to demonstrate the system on a public dataset that contains information on intensive care unit patients. Consider medical researchers who want to examine co-occurrences of certain diseases in certain age groups. They drag and drop into the middle of the interface a pattern-checking algorithm, which at first appears as a blank box. As input, they move into the box disease features labeled, say, “blood,” “infectious,” and “metabolic.” Percentages of those diseases in the dataset appear in the box. Then, they drag the “age” feature into the interface, which displays a bar chart of the patient’s age distribution. Drawing a line between the two boxes links them together. By circling age ranges, the algorithm immediately computes the co-occurrence of the three diseases among the age range.

“It’s like a big, unbounded canvas where you can layout how you want everything,” says Zgraggen, who is the key inventor of Northstar’s interactive interface. “Then, you can link things together to create more complex questions about your data.”

Approximating AutoML

With VDS, users can now also run predictive analytics on that data by getting models custom-fit to their tasks, such as data prediction, image classification, or analyzing complex graph structures.

Using the above example, say the medical researchers want to predict which patients may have blood disease based on all features in the dataset. They drag and drop “AutoML” from the list of algorithms. It’ll first produce a blank box, but with a “target” tab, under which they’d drop the “blood” feature. The system will automatically find best-performing machine-learning pipelines, presented as tabs with constantly updated accuracy percentages. Users can stop the process at any time, refine the search, and examine each model’s errors rates, structure, computations, and other things.

According to the researchers, VDS is the fastest interactive AutoML tool to date, thanks, in part, to their custom “estimation engine.” The engine sits between the interface and the cloud storage. The engine leverages automatically creates several representative samples of a dataset that can be progressively processed to produce high-quality results in seconds.

“Together with my co-authors, I spent two years designing VDS to mimic how a data scientist thinks,” Shang says, meaning it instantly identifies which models and preprocessing steps it should or shouldn’t run on certain tasks, based on various encoded rules. It first chooses from a large list of those possible machine-learning pipelines and runs simulations on the sample set. In doing so, it remembers results and refines its selection. After delivering fast approximated results, the system refines the results in the back end. But the final numbers are usually very close to the first approximation.

“For using a predictor, you don’t want to wait four hours to get your first results back. You want to already see what’s going on and, if you detect a mistake, you can immediately correct it. That’s normally not possible in any other system,” Kraska says. The researchers’ previous user study, in fact, “show that the moment you delay giving users results, they start to lose engagement with the system.”

The researchers evaluated the tool on 300 real-world datasets. Compared to other state-of-the-art AutoML systems, VDS’ approximations were as accurate but were generated within seconds, which is much faster than other tools, which operate in minutes to hours.

Next, the researchers are looking to add a feature that alerts users to potential data bias or errors. For instance, to protect patient privacy, sometimes researchers will label medical datasets with patients aged 0 (if they do not know the age) and 200 (if a patient is over 95 years old). But novices may not recognize such errors, which could completely throw off their analytics.

“If you’re a new user, you may get results and think they’re great,” Kraska says. “But we can warn people that there, in fact, maybe some outliers in the dataset that may indicate a problem.”

Face-Scanning AI Identifies Rare Genetic Disorders

Face-Scanning AI Identifies Rare Genetic Disorders

Research   January 13, 2019

Deep learning algorithms spot genetic disorders better than doctors can by analyzing a patient's facial features

The photograph is cropped close on the face of four-year-old Yael, who is smiling and looking as healthy as can be. But a computer analysis of her features says something’s not right. She has MR XL Bain Type, the computer predicts—a very rare syndrome that causes a wide range of health problems.

It turned out that the computer was right.

Yael is one of thousands of children who have contributed to the development of an artificial intelligence system called Deep Gestalt that can identify rare genetic disorders based on facial features alone. The system, built by Boston-based FDNA, analyzes photographs of faces using computer vision and deep learning algorithms.

In an article published this week in the journal Nature Medicine, FDNA announced the results of a study of DeepGestalt involving 17,000 children, with more than 200 different syndromes among them. The system outperformed clinicians’ ability to identify disease in three separate experiments.

In its best performance, the AI system correctly distinguished between different subtypes of the genetic disorder Noonan syndrome in 64 percent of the cases. Clinicians looking at images of people with Noonan syndrome in previous studies identified the disease correctly in only 20 percent of the cases.

Combined with DNA sequencing, DeepGestalt could prove useful in helping to identify disease, says Yaron Gurovich, chief technology officer at FDNA. “Some people call it deep phenotyping,” he says. “It’s the ability to get accurate and deep insights on a person and link them correctly to genes that were found as problematic in a [DNA] sequencing process.”

The tool could also help standardize the methods doctors use when looking for visual signs of disease, the study authors say. Trying to describe why a person’s facial features are phenotypic expressions of a disease can be challenging. “It’s like when you look at a child and you look at the mother and you know they’re related, but you’re not able to say why,” says Gurovich. “That’s the difference between [a doctor] looking at the facial features and our Gestalt algorithm. It finds a link that we can’t really describe.”

For example, people with Cornelia de Lange syndrome tend to have a small nose, arched eyebrows, and an atypical mouth. But other syndromes, such as the one Yael has, manifest in different ways or aren’t so readily apparent. (In psychology, Gestalt theory "emphasizes that the whole of anything is greater than its parts," according to Britannica.) Image: FDNA/Nature Medicine.This heat map illustrates facial features that influence the algorithms' predictions.

How the algorithms accomplish the task is a black box—a frustrating problem in many AI systems. To get a peek into the algorithms’ methods, the researchers created a color-coded map of the “hot” areas of the face—those that influence the computer’s predictions. It provides a “visualization for our users to try to look inside the black box and understand what the algorithm thinks and how it chose its results,” says Gurovich.

FDNA has analyzed more than 150,000 cases to date. The company amassed its database by building a community platform called Face2Gene that clinical geneticists can use for free. The doctors upload images into the system (with consent from the patient) and in return get to use the platform to help them narrow down the disease possibilities of their patients.

The system provides the doctors with a short list of about 10 possible syndromes the patient might have—not so much a diagnosis, but an aid to help the doctors narrow down the possibilities. Gurovich says 70 percent of clinical geneticists worldwide are using the Face2Gene system.

Those clinicians are getting reliable results, according to the new study. In a fourth experiment, Deep Gestalt analyzed 502 images and generated a suggested list of ten potential syndromes. The list included the patient’s actual disease 91 percent of the time, according to the researchers.

In the paper, Gurovich and his colleagues warn of potential for misuse of the images. “Unlike genomic data, facial images are easily accessible. Payers or employers could potentially analyze facial images and discriminate based on the probability of individuals having pre-existing conditions or developing medical complications,” the authors wrote. They suggest implementing monitoring strategies, such as recording digital footprints on a blockchain, to prevent abuse.

FDNA’s system did not require regulatory approval from the U.S. Food and Drug Administration because it’s considered a reference tool, according to the company.

Yael, who lives in Israel, has become the face of FDNA, with her picture on the home page of the company’s website. An FDNA spokesperson says Yael’s parents are actively searching for more people like Yael who have MR XL Bain Type. There is no treatment yet available for the disease.

One of the Fathers of AI Is Worried About Its Future

One of the Fathers of AI Is Worried About Its Future

Machine-Learning   November 21, 2018

Yoshua Bengio wants to stop talk of an AI arms race and make the technology more accessible to the developing world.

Yoshua Bengio is a grandmaster of modern artificial intelligence.

Alongside Geoff Hinton and Yan LeCun, Bengio is famous for championing a technique known as deep learning that in recent years has gone from an academic curiosity to one of the most powerful technologies on the planet.

Deep learning involves feeding data to large, crudely-simulated neural networks, and it has proven incredibly powerful and effective for all sorts of practical tasks, from voice recognition and image classification to controlling self-driving cars and automating business decisions.

Bengio has resisted the lure of any big tech company. While Hinton and LeCun joined Google and Facebook respectively, he remains a full-time professor at the University of Montreal. (He did, however, cofound Element AI in 2016, a company that built a very successful business helping big companies explore the commercial applications of AI research.)

Bengio met with MIT Technology Review’s senior editor for AI, Will Knight, at an MIT event recently.

What do you make of the idea that there’s an AI race between different countries?

I don’t like it. I don’t think it’s the right way to do it.

We could collectively participate in a race, but as a scientist and somebody who wants to think about the common good, I think we’re better off thinking about how to both build smarter machines and make sure AI is used for the wellbeing of as many people as possible.

Are there ways to foster more collaboration between countries?

We could make it easier for people from developing countries to come to here. It is a big problem right now. In Europe or the US or Canada it is very difficult for an African researcher to get a visa. It’s a lottery, and very often they will use any excuse to refuse access. This is totally unfair. It is already hard for them to do research with little resources, but in addition if they can’t have access to the community, I think that’s really unfair. As a way to counter some of that, we are going to have the ICLR conference [a major AI conference] in 2020 in Africa.

Inclusivity has to be more than a word we say to look good. The potential for AI to be useful in the developing world is even greater. They need to improve technology even more than we do, and they have different needs.

Are you worried about just a few AI companies, in the West and perhaps China, dominating the field of AI?

Yes, it’s another reason why we need to have more democracy in AI research. It’s that AI research by itself will tend to lead to concentrations of power, money, and researchers. The best students want to go to the best companies. They have much more money, they have much more data. And this is not healthy. Even in a democracy, it’s dangerous to have too much power concentrated in a few hands.

There has been a lot of controversy over military uses of AI. Where do you stand on that?

I stand very firmly against.

Even non-lethal uses of AI?

Well, I don’t want to prevent that. I think we need to make it immoral to have killer robots. We need to change the culture, and that includes changing laws and treaties. That can go a long way.

Of course, you’ll never completely prevent it, and people say, “some rogue country will develop these things.” My answer is that one, we want to make them feel guilty for doing it, and two, there’s nothing to stop us from building defensive technology. There’s a big difference between defensive weapons that will kill off drones, and offensive weapons that are targeting humans. Both can use AI.

Shouldn’t AI experts work with the military to ensure this happens?

If they had the right moral values, fine. But I don’t completely trust military organizations because they tend to put duty before morality. I wish it was different.

What are you most excited about in terms of new AI research?

I think we need to consider the hard challenges of AI and not be satisfied with short-term, incremental advances. I’m not saying I want to forget deep learning. On the contrary, I want to build on it. But we need to be able to extend it to do things like reasoning, learning causality, and exploring the world in order to learn and acquire information.

If we really want to approach human-level AI, it’s another ballgame. We need long-term investments and I think academia is the best place to carry that torch.

You mention causality — in other words grasping not just patterns in data by why something happens. Why is that important, and why is it so hard?

If you have a good causal model of the world you are dealing with, you can generalize even in unfamiliar situations. That’s crucial. We humans are able to project ourselves into situations that are very different from our day-to-day experience. Machines are not, because they don’t have these causal models.

We can hand-craft them but that’s not enough. We need machines that can discover causal models. To some extend it’s never going to be perfect. We don’t have a perfect causal model of the reality, that’s why we make a lot of mistakes. But we are much better off at doing this than other animals.

Right now, we don’t really have good algorithms for this, but I think if enough people work at it and consider it important, we will make advances.

Source: MIT Technology Review

New method peeks inside the 'black box' of artificial intelligence

New method peeks inside the 'black box' of artificial intelligence

Trending Technologies   November 04, 2018

Artificial intelligence specifically, machine learning is a part of daily life for computer and smartphone users. From autocorrecting typos to recommending new music, machine learning algorithms can help make life easier. They can also make mistakes.

It can be challenging for computer scientists to figure out what went wrong in such cases. This is because many machine learning algorithms learn from the information and make their predictions inside a virtual "black box," leaving few clues for researchers to follow.

A group of computer scientists at the University of Maryland has developed a promising new approach to interpreting machine learning algorithms. Unlike previous efforts, which typically sought to "break" the algorithms by removing key words from inputs to yield the wrong answer, the UMD group instead reduced the inputs to the bare minimum required to yield the correct answer. On average, the researchers got the correct answer with an input of fewer than three words.

In some cases, the researchers' model algorithms provided the correct answer based on a single word. Frequently, the input word or phrase appeared to have little obvious connection to the answer, revealing important insights into how some algorithms react to specific language. Because many algorithms are programmed to give an answer no matter what—even when prompted by a nonsensical input—the results could help computer scientists build more effective algorithms that can recognize their own limitations.

The researchers will present their work on November 4, 2018 at the 2018 Conference on Empirical Methods in Natural Language Processing.

"Black-box models do seem to work better than simpler models, such as decision trees, but even the people who wrote the initial code can't tell exactly what is happening," said Jordan Boyd-Graber, the senior author of the study and an associate professor of computer science at UMD. "When these models return incorrect or nonsensical answers, it's tough to figure out why. So instead, we tried to find the minimal input that would yield the correct result. The average input was about three words, but we could get it down to a single word in some cases."

In one example, the researchers entered a photo of a sunflower and the text-based question, "What colour is the flower?" as inputs into a model algorithm. These inputs yielded the correct answer of "yellow." After rephrasing the question into several different shorter combinations of words, the researchers found that they could get the same answer with "flower?" as the only text input for the algorithm.

In another, more complex example, the researchers used the prompt, "In 1899, John Jacob Astor IV invested $100,000 for Tesla to further develop and produce a new lighting system. Instead, Tesla used the money to fund his Colorado Springs experiments."

They then asked the algorithm, "What did Tesla spend Astor's money on?" and received the correct answer, "Colorado Springs experiments." Reducing this input to the single word "did" yielded the same correct answer.

The work reveals important insights about the rules that machine learning algorithms apply to problem-solving. Many real-world issues with algorithms result when an input that makes sense to humans results in a nonsensical answer. By showing that the opposite is also possible—that nonsensical inputs can also yield correct, sensible answers—Boyd-Graber and his colleagues demonstrate the need for algorithms that can recognize when they answer a nonsensical question with a high degree of confidence.

"The bottom line is that all this fancy machine learning stuff can actually be pretty stupid," said Boyd-Graber, who also has co-appointments at the University of Maryland Institute for Advanced Computer Studies (UMIACS) as well as UMD's College of Information Studies and Language Science Center. "When computer scientists train these models, we typically only show them real questions or real sentences. We don't show them nonsensical phrases or single words. The models don't know that they should be confused by these examples."

Most algorithms will force themselves to provide an answer, even with insufficient or conflicting data, according to Boyd-Graber. This could be at the heart of some of the incorrect or nonsensical outputs generated by machine learning algorithms—in model algorithms used for research, as well as real-world algorithms that help us by flagging spam email or offering alternate driving directions. Understanding more about these errors could help computer scientists find solutions and build more reliable algorithms.

"We show that models can be trained to know that they should be confused," Boyd-Graber said. "Then they can just come right out and say, 'You've shown me something I can't understand.'"

Provided by University of Maryland