Tag Archives: how
#439628 How a Simple Crystal Could Help Pave the ...
Vaccine and drug development, artificial intelligence, transport and logistics, climate science—these are all areas that stand to be transformed by the development of a full-scale quantum computer. And there has been explosive growth in quantum computing investment over the past decade.
Yet current quantum processors are relatively small in scale, with fewer than 100 qubits— the basic building blocks of a quantum computer. Bits are the smallest unit of information in computing, and the term qubits stems from “quantum bits.”
While early quantum processors have been crucial for demonstrating the potential of quantum computing, realizing globally significant applications will likely require processors with upwards of a million qubits.
Our new research tackles a core problem at the heart of scaling up quantum computers: how do we go from controlling just a few qubits, to controlling millions? In research published today in Science Advances, we reveal a new technology that may offer a solution.
What Exactly Is a Quantum Computer?
Quantum computers use qubits to hold and process quantum information. Unlike the bits of information in classical computers, qubits make use of the quantum properties of nature, known as “superposition” and “entanglement,” to perform some calculations much faster than their classical counterparts.
Unlike a classical bit, which is represented by either 0 or 1, a qubit can exist in two states (that is, 0 and 1) at the same time. This is what we refer to as a superposition state.
Demonstrations by Google and others have shown even current, early-stage quantum computers can outperform the most powerful supercomputers on the planet for a highly specialized (albeit not particularly useful) task—reaching a milestone we call quantum supremacy.
Google’s quantum computer, built from superconducting electrical circuits, had just 53 qubits and was cooled to a temperature close to -273℃ in a high-tech refrigerator. This extreme temperature is needed to remove heat, which can introduce errors to the fragile qubits. While such demonstrations are important, the challenge now is to build quantum processors with many more qubits.
Major efforts are underway at UNSW Sydney to make quantum computers from the same material used in everyday computer chips: silicon. A conventional silicon chip is thumbnail-sized and packs in several billion bits, so the prospect of using this technology to build a quantum computer is compelling.
The Control Problem
In silicon quantum processors, information is stored in individual electrons, which are trapped beneath small electrodes at the chip’s surface. Specifically, the qubit is coded into the electron’s spin. It can be pictured as a small compass inside the electron. The needle of the compass can point north or south, which represents the 0 and 1 states.
To set a qubit in a superposition state (both 0 and 1), an operation that occurs in all quantum computations, a control signal must be directed to the desired qubit. For qubits in silicon, this control signal is in the form of a microwave field, much like the ones used to carry phone calls over a 5G network. The microwaves interact with the electron and cause its spin (compass needle) to rotate.
Currently, each qubit requires its own microwave control field. It is delivered to the quantum chip through a cable running from room temperature down to the bottom of the refrigerator at close to -273 degrees Celsius. Each cable brings heat with it, which must be removed before it reaches the quantum processor.
At around 50 qubits, which is state-of-the-art today, this is difficult but manageable. Current refrigerator technology can cope with the cable heat load. However, it represents a huge hurdle if we’re to use systems with a million qubits or more.
The Solution Is ‘Global’ Control
An elegant solution to the challenge of how to deliver control signals to millions of spin qubits was proposed in the late 1990s. The idea of “global control” was simple: broadcast a single microwave control field across the entire quantum processor.
Voltage pulses can be applied locally to qubit electrodes to make the individual qubits interact with the global field (and produce superposition states).
It’s much easier to generate such voltage pulses on-chip than it is to generate multiple microwave fields. The solution requires only a single control cable and removes obtrusive on-chip microwave control circuitry.
For more than two decades global control in quantum computers remained an idea. Researchers could not devise a suitable technology that could be integrated with a quantum chip and generate microwave fields at suitably low powers.
In our work we show that a component known as a dielectric resonator could finally allow this. The dielectric resonator is a small, transparent crystal which traps microwaves for a short period of time.
The trapping of microwaves, a phenomenon known as resonance, allows them to interact with the spin qubits longer and greatly reduces the power of microwaves needed to generate the control field. This was vital to operating the technology inside the refrigerator.
In our experiment, we used the dielectric resonator to generate a control field over an area that could contain up to four million qubits. The quantum chip used in this demonstration was a device with two qubits. We were able to show the microwaves produced by the crystal could flip the spin state of each one.
The Path to a Full-Scale Quantum Computer
There is still work to be done before this technology is up to the task of controlling a million qubits. For our study, we managed to flip the state of the qubits, but not yet produce arbitrary superposition states.
Experiments are ongoing to demonstrate this critical capability. We’ll also need to further study the impact of the dielectric resonator on other aspects of the quantum processor.
That said, we believe these engineering challenges will ultimately be surmountable— clearing one of the greatest hurdles to realizing a large-scale spin-based quantum computer.
This article is republished from The Conversation under a Creative Commons license. Read the original article.
Image Credit: Serwan Asaad/UNSW, Author provided Continue reading
Posted in Human Robots
#439592 Robot Shows How Simple Swimming Can Be
Lots of robots use bioinspiration in their design. Humanoids, quadrupeds, snake robots—if an animal has figured out a clever way of doing something, odds are there's a robot that's tried to duplicate it. But animals are often just a little too clever for the robots that we build that try to mimic them, which is why researchers at
Swiss Federal Institute of Technology Lausanne in Switzerland (EPFL) are using robots to learn about how animals themselves do what they do. In a paper published today in Science Robotics, roboticists from EPFL's Biorobotics Laboratory introduce a robotic eel that leverages sensory feedback from the water it swims through to coordinate its motion without the need for central control, suggesting a path towards simpler, more robust mobile robots.
The robotic eel—called AgnathaX—is a descendant of
AmphiBot, which has been swimming around at EPFL for something like two decades. AmphiBot's elegant motion in the water has come from the equivalent what are called central pattern generators (CPGs), which are sequences of neural circuits (the biological kind) that generate the sort of rhythms that you see in eel-like animals that rely on oscillations to move. It's possible to replicate these biological circuits using newfangled electronic circuits and software, leading to the same kind of smooth (albeit robotic) motion in AmphiBot.
Biological researchers had pretty much decided that CPGs explained the extent of wiggly animal motion, until it was discovered you can chop an eel's spinal cord in half, and it'll somehow maintain its coordinated undulatory swimming performance. Which is kinda nuts, right? Obviously, something else must be going on, but trying to futz with eels to figure out exactly what it was isn't, I would guess, pleasant for either researchers or their test subjects, which is where the robots come in. We can't make robotic eels that are exactly like the real thing, but we can duplicate some of their sensing and control systems well enough to understand how they do what they do.
AgnathaX exhibits the same smooth motions as the original version of AmphiBot, but it does so without having to rely on centralized programming that would be the equivalent of a biological CPG. Instead, it uses skin sensors that can detect pressure changes in the water around it, a feature also found on actual eels. By hooking these pressure sensors up to AgnathaX's motorized segments, the robot can generate swimming motions even if its segments aren't connected with each other—without a centralized nervous system, in other words. This spontaneous syncing up of disconnected moving elements is called entrainment, and the best demo of it that I've seen is this one, using metronomes:
UCLA Physics
The reason why this isn't just neat but also useful is that it provides a secondary method of control for robots. If the centralized control system of your swimming robot gets busted, you can rely on this water pressure-mediated local control to generate a swimming motion. And there are applications for modular robots as well, since you can potentially create a swimming robot out of a bunch of different physically connected modules that don't even have to talk to each other.
For more details, we spoke with
Robin Thandiackal and Kamilo Melo at EPFL, first authors on the Science Robotics paper.
IEEE Spectrum: Why do you need a robot to do this kind of research?
Thandiackal and Melo: From a more general perspective, with this kind of research we learn and understand how a system works by building it. This then allows us to modify and investigate the different components and understand their contribution to the system as a whole.
In a more specific context, it is difficult to separate the different components of the nervous system with respect to locomotion within a live animal. The central components are especially difficult to remove, and this is where a robot or also a simulated model becomes useful. We used both in our study. The robot has the unique advantage of using it within the real physics of the water, whereas these dynamics are approximated in simulation. However, we are confident in our simulations too because we validated them against the robot.
How is the robot model likely to be different from real animals? What can't you figure out using the robot, and how much could the robot be upgraded to fill that gap?
Thandiackal and Melo: The robot is by no means an exact copy of a real animal, only a first approximation. Instead, from observing and previous knowledge of real animals, we were able to create a mathematical representation of the neuromechanical control in real animals, and we implemented this mathematical representation of locomotion control on the robot to create a model. As the robot interacts with the real physics of undulatory swimming, we put a great effort in informing our design with the morphological and physiological characteristics of the real animal. This for example accounts for the scaling, the morphology and aspect ratio of the robot with respect to undulatory animals, and the muscle model that we used to approximately represent the viscoelastic characteristics of real muscles with a rotational joint.
Upgrading the robot is not going to be making it more “biological.” Again, the robot is part of the model, not a copy of the real biology. For the sake of this project, the robot was sufficient, and only a few things were missing in our design. You can even add other types of sensors and use the same robot base. However, if we would like to improve our robot for the future, it would be interesting to collect other fluid information like the surrounding fluid speed simultaneously with the force sensing, or to measure hydrodynamic pressure directly. Finally, we aim to test our model of undulatory swimming using a robot with three-dimensional capabilities, something which we are currently working on.
Upgrading the robot is not going to be making it more “biological.” The robot is part of the model, not a copy of the real biology.
What aspects of the function of a nervous system to generate undulatory motion in water aren't redundant with the force feedback from motion that you describe?
Thandiackal and Melo: Apart from the generation of oscillations and intersegmental coupling, which we found can be redundantly generated by the force feedback, the central nervous system still provides unique higher level commands like steering to regulate swimming direction. These commands typically originate in the brain (supraspinal) and are at the same time influenced by sensory signals. In many fish the lateral line organ, which directly connects to the brain, helps to inform the brain, e.g., to maintain position (rheotaxis) under variable flow conditions.
How can this work lead to robots that are more resilient?
Thandiackal and Melo: Robots that have our complete control architecture, with both peripheral and central components, are remarkably fault-tolerant and robust against damage in their sensors, communication buses, and control circuits. In principle, the robot should have the same fault-tolerance as demonstrated in simulation, with the ability to swim despite missing sensors, broken communication bus, or broken local microcontroller. Our control architecture offers very graceful degradation of swimming ability (as opposed to catastrophic failure).
Why is this discovery potentially important for modular robots?
Thandiackal and Melo: We showed that undulatory swimming can emerge in a self-organized manner by incorporating local force feedback without explicit communication between modules. In principle, we could create swimming robots of different sizes by simply attaching independent modules in a chain (e.g., without a communication bus between them). This can be useful for the design of modular swimming units with a high degree of reconfigurability and robustness, e.g. for search and rescue missions or environmental monitoring. Furthermore, the custom-designed sensing units provide a new way of accurate force sensing in water along the entirety of the body. We therefore hope that such units can help swimming robots to navigate through flow perturbations and enable advanced maneuvers in unsteady flows. Continue reading
Posted in Human Robots
#439543 How Robots Helped Out After the Surfside ...
Editor's Note: Along with Robin Murphy, the authors of this article include David Merrick, Justin Adams, Jarrett Broder, Austin Bush, Laura Hart, and Rayne Hawkins. This team is with the Florida State University's Disaster Incident Response Team, which was in Surfside for 24 days at the request of Florida US&R Task 1 (Miami Dade Fire Rescue Department).
On June 24, 2021, at 1:25AM portions of the 12 story Champlain Towers South condominium in Surfside, Florida collapsed, killing 98 people and injuring 11, making it the third largest fatal collapse in US history. The life-saving and mitigation Response Phase, the phase where responders from local, state, and federal agencies searched for survivors, spanned June 24 to July 7, 2021. This article summarizes what is known about the use of robots at Champlain Towers South, and offers insights into challenges for unmanned systems.
Small unmanned aerial systems (drones) were used immediately upon arrival by the Miami Dade Fire Rescue (MDFR) Department to survey the roughly 2.68 acre affected area. Drones, such as the DJI Mavic Enterprise Dual with a spotlight payload and thermal imaging, flew in the dark to determine the scope of the collapse and search for survivors. Regional and state emergency management drone teams were requested later that day to supplement the effort of flying day and night for tactical life-saving operations and to add flights for strategic operations to support managing the overall response.
View of a Phantom 4 Pro in use for mapping the collapse on July 2, 2021. Two other drones were also in the airspace conducting other missions but not visible. Photo: Robin R. Murphy
The teams brought at least 9 models of rotorcraft drones, including the DJI Mavic 2 Enterprise Dual, Mavic 2 Enterprise Advanced, DJI Mavic 2 Zoom, DJI Mavic Mini, DJI Phantom 4 Pro, DJI Matrice 210, Autel Dragonfish, and Autel EVO II Pro plus a tethered Fotokite drone. The picture above shows a DJI Phantom 4 Pro in use, with one of the multiple cranes in use on the site visible. The number of flights for tactical operations were not recorded, but drones were flown for 304 missions for strategic operations alone, making the Surfside collapse the largest and longest use of drones recorded for a disaster, exceeding the records set by Hurricane Harvey (112) and Hurricane Florence (260).
Unmanned ground bomb squad robots were reportedly used on at least two occasions in the standing portion of the structure during the response, once to investigate and document the garage and once on July 9 to hold a repeater for a drone flying in the standing portion of the garage. Note that details about the ground robots are not yet available and there may have been more missions, though not on the order of magnitude of the drone use. Bomb squad robots tend to be too large for use in areas other than the standing portions of the collapse.
We concentrate on the use of the drones for tactical and strategic operations, as the authors were directly involved in those operations. It offers a preliminary analysis of the lessons learned. The full details of the response will not be available for many months due to the nature of an active investigation into the causes of the collapse and due to privacy of the victims and their families.
Drone Use for Tactical Operations
Tactical operations were carried out primarily by MDFR with other drone teams supporting when necessary to meet the workload. Drones were first used by the MDFR drone team, which arrived within minutes of the collapse as part of the escalating calls. The drone effort started with night operations for direct life-saving and mitigation activities. Small DJI Mavic 2 Enterprise Dual drones with thermal camera and spotlight payloads were used for general situation awareness to help responders understand the extent of the collapse beyond what could be seen from the street side. The built-in thermal imager was used but did not have the resolution and was unable to show details as much of the material was the same temperature and heat emissions were fuzzy. The spotlight with the standard visible light camera was more effective, though the view was constricted. The drones were also used to look for survivors or trapped victims, help determine safety hazards to responders, and provide task force leaders with overwatch of the responders. During daylight, DJI Mavic Zoom drones were added because of their higher camera resolution zoom. When fires started in the rubble, drones with a streaming connection to bucket truck operators were used to help optimize position of water. Drones were also used to locate civilians entering the restricted area or flying drones to taking pictures.
In a novel use of drones for physical interaction, MDFR squads flew drones to attempt to find and pick up items in the standing portion of the structure with immense value to survivors.
As the response evolved, the use of drones was expanded to missions where the drones would fly in close proximity to structures and objects, fly indoors, and physically interact with the environment. For example, drones were used to read license plates to help identify residents, search for pets, and document belongings inside parts of the standing structure for families. In a novel use of drones for physical interaction, MDFR squads flew drones to attempt to find and pick up items in the standing portion of the structure with immense value to survivors. Before the demolition of the standing portion of the tower, MDFR used a drone to remove an American flag that had been placed on the structure during the initial search.
Drone Use for Strategic Operations
An orthomosiac of the collapse constructed from imagery collected by a drone on July 1, 2021.
Strategic operations were carried out by the Disaster Incident Research Team (DIRT) from the Florida State University Center for Disaster Risk Policy. The DIRT team is a state of Florida asset and was requested by Florida Task Force 1 when it was activated to assist later on June 24. FSU supported tactical operations but was solely responsible for collecting and processing imagery for use in managing the response. This data was primarily orthomosiac maps (a single high resolution image of the collapse created from stitching together individual high resolution imagers, as in the image above) and digital elevation maps (created from structure from motion, below).
Digital elevation map constructed from imagery collected by a drone on 27 June, 2021.Photo: Robin R. Murphy
These maps were collected every two to four hours during daylight, with FSU flying an average of 15.75 missions per day for the first two weeks of the response. The latest orthomosaic maps were downloaded at the start of a shift by the tactical responders for use as base maps on their mobile devices. In addition, a 3D reconstruction of the state of the collapse on July 4 was flown the afternoon before the standing portion was demolished, shown below.
GeoCam 3D reconstruction of the collapse on July 4, 2021. Photo: Robin R. Murphy
The mapping functions are notable because they require specialized software for data collection and post-processing, plus the speed of post-processing software relied on wireless connectivity. In order to stitch and fuse images without gaps or major misalignments, dedicated software packages are used to generate flight paths and autonomously fly and trigger image capture with sufficient coverage of the collapse and overlap between images.
Coordination of Drones on Site
The aerial assets were loosely coordinated through social media. All drones teams and Federal Aviation Administration (FAA) officials shared a WhatsApp group chat managed by MDFR. WhatsApp offered ease of use, compatibility with everyone's smartphones and mobile devices, and ease of adding pilots. Ease of adding pilots was important because many were not from MDFR and thus would not be in any personnel-oriented coordination system. The pilots did not have physical meetings or briefings as a whole, though the tactical and strategic operations teams did share a common space (nicknamed “Drone Zone”) while the National Institute of Standards and Technology teams worked from a separate staging location. If a pilot was approved by MDFR drone captain who served as the “air boss,” they were invited to the WhatsApp group chat and could then begin flying immediately without physically meeting the other pilots.
The teams flew concurrently and independently without rigid, pre-specified altitude or area restrictions. One team would post that they were taking off to fly at what area of the collapse and at what altitude and then post when they landed. The easiest solution was for the pilots to be aware of each others' drones and adjust their missions, pause, or temporarily defer flights. If a pilot forgot to post, someone would send a teasing chat eliciting a rapid apology.
Incursions by civilian manned and unmanned aircraft in the restricted airspace did occur. If FAA observers or other pilots saw a drone flying that was not accounted for in the chat, i.e., that five drones were visible over the area but only four were posted, or if a drone pilot saw a drone in an unexpected area, they would post a query asking if someone had forgotten to post or update a flight. If the drone remained unaccounted for, the FAA would assume that a civilian drone had violated the temporary flight restrictions and search the surrounding area for the offending pilot.
Preliminary Lessons LearnedWhile the drone data and performance is still being analyzed, some lessons learned have emerged that may be of value to the robotics, AI, and engineering communities.
Tactical and strategic operations during the response phase favored small, inexpensive, easy to carry platforms with cameras supporting coarse structure from motion rather than larger, more expensive lidar systems. The added accuracy of lidar systems was not needed for those missions, though the greater accuracy and resolution of such systems were valuable for the forensic structural analysis. For tactical and strategic operations, the benefits of lidar was not worth the capital costs and logistical burden. Indeed, general purpose consumer/prosumer drones that could fly day or night, indoors and outdoors, and for both mapping and first person view missions were highly preferred over specialized drones. The reliability of a drone was another major factor in choosing a specific model to field, again favoring consumer/prosumer drones as they typically have hundreds of thousand hours of flight time more than specialized or novel drones. Tethered drones offer some advantages for overwatch but many tactical operations missions require a great deal of mobility. Strategic mapping necessitates flying directly over the entire area being mapped.
While small, inexpensive general purpose drones offered many advantages, they could be further improved for flying at night and indoors. A wider area of lighting would be helpful. A 360 degree (spherical) area of coverage for obstacle avoidance for working indoors or at low altitudes and close proximity to irregular work envelopes and near people, especially as night, would also be useful. Systems such as the Flyability ELIOS 2 are designed to fly in narrow and highly cluttered indoor areas, but no models were available for the immediate response. Drone camera systems need to be able to look straight up to inspect the underside of structures or ceilings. Mechanisms for determining the accurate GPS location of a pixel in an image, not just the GPS location of the drone, is becoming increasing desirable.
Other technologies could be of benefit to the enterprise but face challenges. Computer vision/machine learning (CV/ML) for searching for victims in rubble is often mentioned as a possible goal, but a search for victims who are not on the surface of the collapse is not visually directed. The portions of victims that are not covered by rubble are usually camouflaged with gray dust, so searches tend to favor canines using scent. Another challenge for CV/ML methods is the lack of access to training data. Privacy and ethical concerns poses barriers to the research community gaining access to imagery with victims in the rubble, but simulations may not have sufficient fidelity.
The collapse supplies motivation for how informatics research and human-computer interaction and human-robot interaction can contribute to the effective use of robots during a disaster, and illustrates that a response does not follow a strictly centralized, hierarchical command structure and the agencies and members of the response are not known in advance. Proposed systems must be flexible, robust, and easy to use. Furthermore, it is not clear that responders will accept a totally new software app versus making do with a general purpose app such as WhatsApp that the majority routinely use for other purposes.
The biggest lesson learned is that robots are helpful and warrant more investment, particular as many US states are proposing to terminate purchase of the very models of drones that were so effective over cybersecurity concerns.
However, the biggest lesson learned is that robots are helpful and warrant more investment, particular as many US states are proposing to terminate purchase of the very models of drones that were so effective over cybersecurity concerns. There remains much to work to be done by researchers, manufacturers, and emergency management to make these critical technologies more useful for extreme environments. Our current work is focusing on creating open source datasets and documentation and conducting a more thorough analysis to accelerate the process.
Value of Drones The pervasive use of the drones indicates their implicit value to responding to, and documenting, the disaster. It is difficult to quantify the impact of drones, similar to the difficulties in quantifying the impact of a fire truck on firefighting or the use of mobile devices in general. Simply put, drones would not have been used beyond a few flights if they were not valuable.
The impact of the drones on tactical operations was immediate, as upon arrival MDFR flew drones to assess the extent of the collapse. Lighting on fire trucks primarily illuminated the street side of the standing portion of the building, while the drones, unrestricted by streets or debris, quickly expanded situation awareness of the disaster. The drones were used optimize placement of water to suppress the fires in the debris. The impact of the use of drones for other tactical activities is harder to quantify, but the frequent flights and pilots remaining on stand-by 24/7 indicate their value.
The impact of the drones on strategic operations was also considerable. The data collected by the drones and then processed into 2D maps and 3D models became a critical part of the US&R operations as well as one part of the nascent investigation into why the building failed. During initial operations, DIRT provided 2D maps to the US&R teams four times per day. These maps became the base layers for the mobile apps used on the pile to mark the locations of human remains, structural members of the building, personal effects, or other identifiable information. Updated orthophotos were critical to the accuracy of these reports. These apps running on mobile devices suffered from GPS accuracy issues, often with errors as high as ten meters. By having base imagery that was only hours old, mobile app users where able to 'drag the pin' on the mobile app to a more accurate report location on the pile – all by visualizing where they were standing compared to fresh UAS imagery. Without this capability, none of the GPS field data would be of use to US&R or investigators looking at why the structural collapse occurred. In addition to serving a base layer on mobile applications, the updated map imagery was used in all tactical, operational, and strategic dashboards by the individual US&R teams as well as the FEMA US&R Incident Support Team (IST) on site to assist in the management of the incident.
Aside from the 2D maps and orthophotos, 3D models were created from the drone data and used by structural experts to plan operations, including identifying areas with high probabilities of finding survivors or victims. Three-dimensional data created through post-processing also supported the demand for up-to-date volumetric estimates – how much material was being removed from the pile, and how much remained. These metrics provided clear indications of progress throughout the operations.
Acknowledgments Portions of this work were supported by NSF grants IIS-1945105 and CMMI- 2140451. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.
The authors express their sincere condolences to the families of the victims. Continue reading
Posted in Human Robots
#439374 A model to predict how much humans and ...
Researchers at University of Michigan have recently developed a bi-directional model that can predict how much both humans and robotic agents can be trusted in situations that involve human-robot collaboration. This model, presented in a paper published in IEEE Robotics and Automation Letters, could help to allocate tasks to different agents more reliably and efficiently. Continue reading
Posted in Human Robots
#439357 How the Financial Industry Can Apply AI ...
iStockphoto
THE INSTITUTE Artificial intelligence is transforming the financial services industry. The technology is being used to determine creditworthiness, identify money laundering, and detect fraud.
AI also is helping to personalize services and recommend new offerings by developing a better understanding of customers. Chatbots and other AI assistants have made it easier for clients to get answers to their questions, 24/7.
Although confidence in financial institutions is high, according to the Banking Exchange, that’s not the case with AI. Many people have raised concerns about bias, discrimination, privacy, surveillance, and transparency.
Regulations are starting to take shape to address such concerns. In April the European Commission released the first legal framework to govern use of the technology, as reported in IEEE Spectrum. The proposed European regulations cover a variety of AI applications including credit checks, chatbots, and social credit scoring, which assesses an individual’s creditworthiness based on behavior. The U.S. Federal Trade Commission in April said it expects AI to be used truthfully, fairly, and equitably when it comes to decisions about credit, insurance, and other services.
To ensure the financial industry is addressing such issues, IEEE recently launched a free guide, “Trusted Data and Artificial Intelligence Systems (AIS) for Financial Services.” The authors of the nearly 100-page playbook want to ensure that those involved in developing the technologies are not neglecting human well-being and ethical considerations.
More than 50 representatives from major banks, credit unions, pension funds, and legal and compliance groups in Canada, the United Kingdom, and the United States provided input, as did AI experts from academia and technology companies.
“This IEEE finance playbook is a milestone achievement and provides a much-needed practical road map for organizations globally to develop their trusted data and ethical AI systems.”
“We are in the business of trust. A primary goal of financial services organizations is to use client and member data to generate new products and services that deliver value,” Sami Ahmed says. He is a member of IEEE industry executive steering committee that oversaw the playbook’s creation.
Ahmed is senior vice president of data and advanced analytics of OMERS, Ontario’s municipal government employees’ pension fund and one of the largest institutional investors in Canada.
“Best-in-class guidance assembled from industry experts in IEEE’s finance playbook,” he says, “addresses emerging risks such as bias, fairness, explainability, and privacy in our data and algorithms to inform smarter business decisions and uphold that trust.”
The playbook includes a road map to help organizations develop their systems. To provide a theoretical framework, the document incorporates IEEE’s “Ethically Aligned Design” report, the IEEE 7000 series of AI standards and projects, and the Ethics Certification Program for Autonomous and Intelligent Systems.
“Design looks completely different when a product has already been developed or is in prototype form,” says John C.Havens, executive director of the IEEE Global Initiative on Ethics of Autonomous and Intelligent Systems. “The primary message of ethically aligned design is to use the methodology outlined in the document to address these issues at the outset.”
Havens adds that although IEEE isn’t well known by financial services regulatory bodies, it does have a lot of credibility in harnessing the technical community and creating consensus-based material.
“That is why IEEE is the right place to publish this playbook, which sets the groundwork for standards development in the future,” he says.
IEEE Member Pavel Abdur-Rahman, chair of the IEEE industry executive steering committee, says the document was necessary to accomplish three things. One was to “verticalize the discussion within financial services for a very industry-specific capability building dialog. Another was to involve industry participants in the cocreation of this playbook, not only to curate best practices but also to develop and drive adoption of the IEEE standards into their organizations.” Lastly, he says, “it’s the first step toward creating recommended practices for MLOps [machine-learning operations], data cooperatives, and data products and marketplaces.
Abdur-Rahman is the head of trusted data and AI at IBM Canada.
ROAD MAP AND RESOURCES
The playbook has two sections, a road map for how to build trusted AI systems and resources from experts.
The road map helps organizations identify where they are in the process of adopting responsible ethically aligned design: early, developing, advanced, or mature stage. This section also outlines 20 ways that trusted data and AI can provide value to operating units within a financial organization. Called use cases, the examples include cybersecurity, loan and deposit pricing, improving operational efficiency, and talent acquisition. Graphs are used to break down potential ethical concerns for each use case.
The key resources section includes best practices, educational videos, guidelines, and reports on codes of conduct, ethical challenges, building bots responsibly, and other topics. Among the groups contributing resources are the European Commission, IBM, the IEEE Standards Association, Microsoft, and the World Economic Forum. Also included is a report on the impact the coronavirus pandemic has had on the financial services industry in Canada. Supplemental information includes a list of 84 documents on ethical guidelines.
“We are at a critical junction of industrial-scale AI adoption and acceleration,” says Amy Shi-Nash, a member of the steering committee and the global head of analytics and data science for HSBC. “This IEEE finance playbook is a milestone achievement and provides a much-needed practical road map for organizations globally to develop their trusted data and ethical AI systems.”
To get an evaluation of the readiness of your organization’s AI system, you can anonymously take a 20-minute survey.
IEEE membership offers a wide range of benefits and opportunities for those who share a common interest in technology. If you are not already a member, consider joining IEEE and becoming part of a worldwide network of more than 400,000 students and professionals. Continue reading
Posted in Human Robots