An increasing number of medical devices incorporate artificial intelligence (AI) capabilities to support therapeutic and diagnostic applications. In spite of the risks connected with this innovative technology, the applicable regulatory framework does not specify any requirements for this class of medical devices. To make matters even more complicated for manufacturers, there are no standards, guidance documents or common specifications for medical devices on how to demonstrate conformity with the essential requirements.

The term artificial intelligence (AI) describes the capability of algorithms to take over tasks and decisions by mimicking human intelligence.1 Many experts believe that machine learning, a subset of artificial intelligence, will play a significant role in the medtech sector.2,3 Machine learning is the term used to describe algorithms capable of learning directly from a large volume of training data. The algorithm builds a model based on training data and applies the experience, it has gained from the training to make predictions and decisions on new, unknown data. Artificial neural networks are a subset of machine learning methods, which have evolved from the idea of simulating the human brain.22 Neural networks are information-processing systems used for machine learning and comprise multiple layers of neurons. Between the input layer, which receives information, and the output layer, there are numerous hidden layers of neurons. In simple terms, neural networks comprise neurons also known as nodes which receive external information or information from other connected nodes, modify this information, and pass it on, either to the next neuron layer or to the output layer as the final result.5 Deep learning is a variation of artificial neural networks, which consist of multiple hidden neural network layers between the input and output layers. The inner layers are designed to extract higher-level features from the raw external data.

The role of artificial intelligence and machine learning in the health sector was already the topic of debate well before the coronavirus pandemic.6 As shown in an excerpt from PubMed several approaches for AI in medical devices have already been implemented in the past (see Figure 1). However, the number of publications on artificial intelligence and medical devices has grown exponentially since roughly 2005.

Artificial intelligence in the medtech sector is at the beginning of a growth phase. However, expectations for this technology are already high, and consequently prospects for the digital future of the medical sector are auspicious. In the future, artificial intelligence may be able to support health professionals in critical tasks, controlling and automating complex processes. This will enable diagnosis, therapy and care to be optimally aligned to patients individual needs, thereby increasing treatment efficiency, which in turn will ensure an effective and affordable healthcare sector in the future.4

However, some AI advocates tend to overlook some of the obstacles and risks encountered when artificial intelligence is implemented in clinical practice. This is particularly true for the upcoming regulation of this innovative technology. The risks of incorporating artificial intelligence in medical devices include faulty or manipulated training data, attacks on AI such as adversarial attacks, violation of privacy and lack of trust in technology. In spite of these technology-related risks, the applicable standards and regulatory frameworks do not include any specific requirements for the use of artificial intelligence in medical devices. After years of negotiations in the European Parliament, Regulation (EU) 2017/745 on medical devices and Regulation (EU) 2017/746 on in-vitro diagnostic medical devices entered into force on May 25, 2017. In contrast to Directives, EU Regulations enter directly into force in the EU Member States and do not have to be transferred into national law. The new regulations impose strict demands on medical device manufacturers and the Notified Bodies, which manufacturers must involve in the certification process of medical devices and in-vitro diagnostic medical devices (excluding class I medical devices and nonsterile class A in-vitro diagnostic medical devices, for which the manufacturers self-declaration will be sufficient).

Annex I to both the EU Regulation on medical devices (MDR) and the EU Regulation on in vitro diagnostic medical devices (IVDR) define general safety and performance requirements for medical devices and in-vitro diagnostics. However, these general requirements do not address the specific requirements related to artificial intelligence. To make matters even more complicated for manufacturers, there are no standards, guidance documents or common specifications on how to demonstrate conformity with the general requirements. To place a medical device on the European market, manufacturers must meet various criteria, including compliance with the essential requirements and completion of the appropriate conformity assessment procedure. By complying with the requirements, manufacturers ensure that their medical devices fulfill the high levels of safety and health protection required by the respective regulations.

To ensure the safety and performance of artificial intelligence in medical devices and in-vitro diagnostics, certain minimum requirements must be fulfilled. However, the above regulations define only general requirements for software. According to the general safety and performance requirements, software must be developed and manufactured in keeping with the state of the art. Factors to be taken into account include the software lifecycle process and risk management. Beyond the above, repeatability, reliability and performance in line with the intended use of the medical device must be ensured. This implicitly requires artificial intelligence to be repeatable, performant, reliable and predictable. However, this is only possible with a verified and validated model. Due to the absence of relevant regulatory requirements and standards, manufacturers and Notified Bodies are determining the state of the art for developing and testing artificial intelligence in medical devices, respectively. During the development, assessment and testing of AI, fundamental differences between artificial intelligence (particularly machine learning) and conventional software algorithms become apparent.

Towards the end of 2019, and thus just weeks before the World Health Organizations (WHO) warning of an epidemic in China, a Canadian company (BlueDot) specializing in AI-based monitoring of the spread of infectious diseases alerted its customers to the same risk. To achieve this the companys AI combed through news reports and databases of animal and plant diseases. By accessing global flight ticketing data, the AI system correctly forecast the spread of the virus in the days after it emerged. This example shows the high level of performance that can already be achieved with artificial intelligence today.7 However, it also reveals one of the fundamental problems encountered with artificial intelligence: Despite the distribution of information of the outbreak to various health organizations in different countries, international responses were few. One reason for this lack of response to the AI-based warning is the lack of trust in technology that we do not understand, which plays a particularly significant role in medical applications.

In clinical applications, artificial intelligence is predominantly used for diagnostic purposes. Analysis of medical images is the area where the development of AI models is most advanced. Artificial intelligence is successfully used in radiology, oncology, ophthalmology, dermatology and other medical disciplines.2 The advantages of using artificial intelligence in medical applications include the speed of data analysis and the capability of identifying patterns invisible to the human eye.

Take the diagnosis of osteoarthritis, for example. Although medical imaging enables healthcare professionals to identify osteoarthritis, this is generally at a late stage after the disease has already caused some cartilage breakdown. Using an artificial-intelligence system, a research team led by Dr. Shinjini Kundu analyzed magnetic resonance tomography (MRT) images. The team was able to predict osteoarthritis three years before the first symptoms manifested themselves.8 However, the team members were unable to explain how the AI system arrived at its diagnosis. In other words, the system was not explainable. The question now is whether patients will undergo treatment such as surgery, based on a diagnosis made by an AI system, which no doctor can either explain or confirm.

Further investigations revealed that the AI system identified diffusion of water into cartilage. It detected a symptom invisible to the human eye and, even more important, a pattern that had previously been unknown to science. This example again underlines the importance of trust in the decision of artificial intelligence, particularly in the medtech sector. Justification of decisions is one of the cornerstones of a doctor-patient (or AI-patient) relationship based on mutual trust. However, to do so the AI system must be explainable, understandable and transparent. Patients, doctors and other users will only trust in AI systems if their decisions can be explained and understood.

Many medical device manufacturers wonder why assessment and development of artificial intelligence must follow a different approach to that of conventional software. The reason is based on the principles of how artificial intelligence is developed and how it performs. Conventional software algorithms take an input variable X, process it using a defined algorithm and supply the result Y as the output variable (if X, then Y). The algorithm is programmed, and its correct function can be verified and validated. The requirements for software development, validation and verification are described in the two standards IEC 62304 and IEC 82304-1. However, there are fundamental differences between conventional software and artificial intelligence implementing a machine learning algorithm. Machine learning is based on using data to train a model without explicitly programming the data flow line by line. As described above, machine learning is trained using an automated appraisal of existing information (training data). Given this, both the development and conformity assessment of artificial intelligence necessitate different standards. The following sections provide a brief overview of the typical pitfalls.

A major disadvantage of artificial intelligence, in particular machine learning based on neural networks, is the complexity of the algorithms. This makes them highly non-transparent, hence their designation of black-box AI (see Figure 2). The complex nature of AI algorithms not only concerns their mathematical description but alsoin the case of deep-learning algorithmstheir high level of dimensionality and abstraction. For these classes of AI, the extent to which input information contributes to a specific decision is mostly impossible to determine. This is why AI is often referred to as black box AI. Can we trust the prediction of the AI system in such a case and, in a worst-case scenario, can we identify a failure of the system or a misdiagnosis?

A world-famous example of the result of a black-box AI was the match between AlphaGo, the artificial intelligence system made by DeepMind (Google) and the Go world champion, Lee Sedol. In the match, which was watched by an audience of 60 million including experts, move 37 showed the significance of these particular artificial intelligence characteristics. The experts described the move as a mistake, predicting that AlphaGo would lose the match since in their opinion the move made no sense at all. In fact, they went even further and said, Its not a human move. Ive never seen a human play this move9.

None of them understood the level of creativity behind AlphaGos move, which proved to be critical for winning the match. While understanding the decision made by the artificial intelligence system would certainly not change the outcome of the match, it still shows the significance of the explainability and transparency of artificial intelligence, particularly in the medical field. AlphaGo could also have been wrong!

One example of AI with an intended medical use was the application of artificial intelligence for determining a patients risk of pneumonia. This example shows the risk of black-box AI in the MedTech sector. The system in question surprisingly identified the high-risk patients as non-significant.10 Rich Caruana, one of the leading AI experts at Microsoft, who was also one of the developers of the system, advised against the use of the artificial intelligence he had developed: I said no. I said we dont understand what it does inside. I said I was afraid.11

In this context, it is important to note that open or explainable artificial intelligence, also referred to as white box, is by no means inferior to black-box AI. While there have been no standard methods for opening the black box, there are promising approaches for ensuring the plausibility of the predictions made by AI models. Some approaches try to achieve explainability based on individual predictions on input data. Others, by contrast, try to limit the range of input pixels that impact the decisions of artificial intelligence.12

Medical devices and their manufacturers must comply with further regulatory requirements in addition to the Medical Device Regulation (MDR) and the In-vitro Diagnostics Regulation (IVDR). The EUs General Data Protection Regulation (GDPR), for instance, is of particular relevance for the explainability of artificial intelligence. It describes the rules that apply to the processing of personal data and is aimed at ensuring their protection. Article 110 of the Medical Device Regulation (MDR) explicitly requires measures to be taken to protect personal data, referencing the predecessor of the General Data Protection Regulation.

AI systems which influence decisions that might concern an individual person must comply with the requirements of Articles 13, 22 and 35 of the GDPR.

Where personal data are collected, the controller shall provide.the following information: the existence of automated decision-making and at least in those cases, meaningful information of the logic involved13

In simple terms this means, that patients who are affected by automated decision-making must be able to understand this decision and have the possibility to take legal action against it. However, this is precisely the type of understanding which is not possible in the case of black box AI. Is a medical product implemented as black-box AI eligible for certification as a medical device? The exact interpretation of the requirements specified in the General Data Protection Regulation is currently the subject of legal debate.14

The Medical Device Regulation places manufacturers under the obligation to ensure the safety of medical devices. Among other specifications, Annex I to the regulation includes, , requirements concerning the repeatability, reliability and performance of medical devices (both for stand-alone software and software embedded into a medical device):

Devices that incorporate electronic programmable systems, including software, shall be designed to ensure repeatability, reliability and performance in line with their intended use. (MDR Annex I, 17.1)15

Compliance with general safety and performance requirements can be demonstrated by utilizing harmonized standards. Adherence to a harmonized standard leads to the assumption of conformity, whereby the requirements of the regulation are deemed to be fulfilled. Manufacturers can thus validate artificial intelligence models in accordance with the ISO 13485:2016 standard, which, among other requirements, describes the processes for the validation of design and development in clause 7.3.7.

For machine learning two independent sets of data must be considered. In the first step, one set of data is needed to train the AI model. Subsequently, another set of data is necessary to validate the model. Validation of the model should use independent data and can also be performed by cross-validation in the meaning of the combined use of both data sets. However, it must be noted that AI models can only be validated using an independent data set. Now, which ratio is recommended for the two sets of data? This is not an easy question to answer without more detailed information about the characteristics of the AI model. A look at the published literature (state of the art) recommends a ratio of approx. 80% training data to approx. 20% validation data. However, the ratio being used depends on many factors and is not set in stone. The notified bodies will continue to monitor the state of the art in this area and, within the scope of conformity assessment, also request the reasons underlying the ratio used.

Another important question concerns the number of data sets. As the number of data sets depends on the following factors, this issue is difficult to assess, depending on:

Generally, the larger the number of data the more performant the model can be assumed to work. In their publication on speech recognition, Banko and Brill from Microsoft state, After throwing more than one billion words within context at the problem, any algorithm starts to perform incredibly well16

At the other end of the scale, i.e. the minimum number of data sets required, computational learning theory offers approaches for estimating the lower threshold. However, general answers to this question are not yet known and these approaches are based on ideal assumptions and only valid for simple algorithms.

Manufacturers need to look not only at the number of data, but also at the statistical distribution of both sets of data. To prevent bias, the data used for training and validation must represent the statistical distribution of the environment of application. Training with data that are not representative will result in bias. The U.S. healthcare system, for example, uses artificial intelligence algorithms to identify and help patients with complex health needs. However, it soon became evident that where patients had the same level of health risks, the model suggested Afro-Americans less often for enrolment in these special high-risk care management programs.17 Studies carried out by Obermeyer, et al. showed the cause for this to be racial bias in training data. Bias in training data not only involves ethical and moral aspects that need to be considered by manufacturers: it can also affect the safety and performance of a medical device. Bias in training data could, for example, result in certain indications going undetected on fair skin.

Many deep learning models rely on a supervised learning approach, in which AI models are trained using labelled data. In cases involving labelled data, the bottleneck is not the number of data, but the rate and accuracy at which data are labeled. This renders labeling a critical process in model development. At the same time, data labelling is error-prone and frequently subjective, as it is mostly done by humans. Humans also tend to make mistakes in repetitive tasks (such as labelling thousands of images).

Labeling of large data volumes and selection of suitable identifiers is a time- and cost-intensive process. In many cases, only a very minor amount of the data will be processed manually. These data are used to train an AI system. Subsequently the AI system is instructed to label the remaining data itselfa process that is not always error-free, which in turn means that errors will be reproduced.7 Nevertheless, the performance of artificial intelligence combined with machine learning very much depends on data quality. This is where the accepted principle of garbage in, garbage out becomes evident. If a model is trained using data of inferior quality, the developer will also obtain a model of the same quality.

Other properties of artificial intelligence that manufacturers need to take into account are adversarial learning problems and instabilities of deep learning algorithms. Generally, the assumption in most machine learning algorithms is that training and test data are governed by identical distributions. However, this statistical assumption can be influenced by an adversary (i.e., an attacker that attempts to fool the model by providing deceptive input). Such attackers aim to destabilize the model and to cause the AI to make false predictions. The introduction of certain adversarial patterns to the input data that are invisible to the human eye causes major errors of detection to be made by the AI system. In 2020, for example, the security company McAfee demonstrated their ability to trick Teslas Mobileye EyeQ3 AI System into driving 80 km/h over the speed limit, simply by adding a 5 cm strip of black tape to a speed limit sign.24

AI methods used in the reconstruction of MRT and CT images have also proved unstable in practice time and again. A study investigating six of the most common AI methods used in the reconstruction of MRT and CT images proved these methods to be highly unstable. Even minor changes in the input images, invisible to the human eye, result in completely distorted reconstructed image.18 The distorted images included artifacts such as removal of tissue structures, which might result in misdiagnosis. Such an attack may cause artificial intelligence to reconstruct a tumor at a location where there is none in reality or even remove cancerous tissue from the real image. These artifacts are not present when manipulated images are reconstructed using conventional algorithms.18

Another vulnerability of artificial intelligence concerns image-scaling attacks. This vulnerability has been known since as long ago as 2019.19 Image-scaling attacks enable the attacker to manipulate the input data in such a way that artificial intelligence models with machine learning and image scaling can be brought under the attackers control. Xiao et al., for example, succeeded in manipulating the well-known machine-learning scaling library, TensorFlow, in such a manner that attackers could even replace complete images.19 An example of such an image-scaling attack is shown in Figure 3. In this scaling operation, the image of a cat is replaced by an image of a dog. Image-scaling attacks are particularly critical as they can both distort training of artificial intelligence and influence the decisions of artificial intelligence trained using manipulated images.

Adversarial attacks and stability issues pose significant threats to the safety and performance of medical devices incorporating artificial intelligence. Especially concerning is the fact that the conditions of when and where the attacks could occur, are difficult to predict. Furthermore, the response of AI to adversarial attacks is difficult to specify. If, for instance, a conventional surgical robot is attacked, it can still rely on other sensors. However, changing the policy of the AI in a surgical robot might lead to unpredictable behavior and by this to catastrophic (from a human perspective) responses of the system. Methods to address the above vulnerabilities and reduce susceptibility to errors do exist. For example, the models can be subjected to safety training, making them more resilient to the vulnerabilities. Defense techniques such as adversarial training and defense distillation have already been practiced successfully in image reconstruction algorithms.21 Further methods include human-in-the-loop approaches, as humans performance is strongly robust against adversarial attacks targeting AI systems. However, this approach has limited applicability in instances where humans can be directly involved.25

Although many medical devices using artificial intelligence have already been approved, the regulatory pathways in the medtech sector are still open. At present no laws, common specifications or harmonized standards exist to regulate AI application in medical devices. In contrast to the EU authorities, the FDA published a discussion paper on a proposed regulatory framework for artificial intelligence in medical devices in 2019. The document is based on the principle of risk management, software-change management, guidance on the clinical evaluation of software and a best-practice approach to the software lifecycle.20 in 2021, the FDA published their action plan on furthering AI in medical devices. The action plan consists of five next steps, with the foremost being to develop a regulatory framework explicitly for change control of AI, a good machine learning practice and new methods to evaluate algorithm bias and robustness 26

In 2020 the European Union also published a position paper on the regulation of artificial intelligence and medical devices. The EU is currently working on future regulation, with a first draft expected in 2021.

Chinas National Medical Products Administration (NMPA) published the Technical Guiding Principles of Real-World Data for Clinical Evaluation of Medical Devices guidance document. It specifies obligations concerning requirements-analysis, data collection and processing, model definition, verification, and validation as well as post-market surveillance.

Japans Ministry of Health, Labour and Welfare is working on a regional standard for artificial intelligence in medical devices. However, to date this standard is available in Japanese only. Key points of assessment are plasticity the predictability of models, quality of data and degree of autonomy. 27

In Germany, the Notified Bodies have developed their own guidance for artificial intelligence. The guidance document was prepared by the Interest Group of the Notified Bodies for Medical Devices in Germany (IG-NB) and is aimed at providing guidance to Notified Bodies, manufacturers and interested third parties. The guidance follows the principle that the safety of AI-based medical devices can only be achieved by means of a process-focused approach that covers all relevant processes throughout the whole life cycle of a medical device. Consequently, the guidance does not define specific requirements for products, but for processes.

The World Health Organization, too, is currently working on a guideline addressing artificial intelligence in health care.

Artificial intelligence is already used in the medtech sector, albeit currently somewhat sporadically. However, at the same time the number of AI algorithms certified as medical devices has increased significantly over the last years.28 Artificial intelligence is expected to play a significant role in all stages of patient care. According to the requirements defined in the Medical Device Regulation, any medical device, including those incorporating AI, must be designed in such a way as to ensure repeatability, reliability and performance according to its intended use. In the event of a fault condition (single fault condition), the manufacturer must implement measures to minimize unacceptable risks and reduction of the performance of the medical device (MDR Annex I, 17.1). However, this requires validation and verification of the AI model.

Many of the AI models used are black-box models. In other words, there is no transparency in how these models arrive at their decisions. This poses a problem where interpretability and trustworthiness of the systems are concerned. Without transparent and explainable AI predictions, the medical validity of a decision might be doubted. Some current errors of AI in pre-clinical applications might fuel doubts further. Explainable and approvable AI decisions are a prerequisite for the safe use of AI on actual patients. This is the only way to inspire trust and maintain it in the long term.

The General Data Protection Regulation demands a high level of protection of personal data. Its strict legal requirements also apply to processing of sensitive health data in the development or verification of artificial intelligence.

Adversarial attacks aim at influencing artificial intelligence, both during the training of the model and in the classification decision. These risks must be kept under control by taking suitable measures.

Impartiality and fairness are important, safety-relevant, moral and ethical aspects of artificial intelligence. To safeguard these aspects, experts must take steps to prevent bias when training the system.

Another important question concerns the responsibility and accountability of artificial intelligence. Medical errors made by human doctors can generally be traced back to the individuals, who can be held accountable if necessary. However, if artificial intelligence makes a mistake the lines of responsibility become blurred. For medical devices on the other hand, the question is straightforward. The legal manufacturer of the medical device incorporating artificial intelligence must ensure the safety and security of the medical device and assume liability for possible damage.

Regulation of artificial intelligence is likewise still at the beginning of development involving various approaches. All major regulators around the globe have defined or are starting to define requirements for artificial intelligence in medical devices. A high level of safety in medical devices will only be possible with suitable measures in place to regulate and control artificial intelligencebut this must not impair the development of technical innovation.

Follow to Page 2 for References.

The Chinese government is investing heavily in the development of new technologies that leverage AI.

If youre looking to market your medical device, there are many tasks to complete.

The term "Big Data" is a few years old, but its implications for medical devices

The race to apply AI to medical treatment is rapidly accelerating in China and Japan.

Go here to see the original:
AI in MedTech: Risks and Opportunities of Innovative Technologies in Medical Applications - MedTech Intelligence

Servers run inside the Facebook New Albany Data Center on Thursday, February 6, 2020 in New Albany, Ohio. Photo via Joshua A. Bickel/Dispatch and Reuters.

Artificial intelligence (AI) is causing significant structural changes to global competition and economic growth. AI may generate trillions of dollars in new value over the next decade, but this value will not be easily captured or evenly distributed across nations. Much of it will depend on how governments invest in the underlying computational infrastructure that makes AI possible.

Yet early signs point to a blind spota lack of understanding, measurement, and planning. It comes in the form of compute divides that throttle innovation across academia, startups, and industry. Policymakers must make AI compute a core component of their strategic planning in order to fully realize the anticipated economic windfall.

AI compute refers to a specialized stack of hardware and software optimized for AI applications or workloads. This computer can be located and accessed in different ways (public clouds, private data centers, individual workstations, etc.) and leveraged to solve complex problems across domains from astrophysics to e-commerce and to autonomous vehicles.

Conventional information technology (IT) infrastructure is now widely available as a utility through public cloud service providers. This idea of infrastructure as a service includes computing for AI, so the public cloud is duly credited with democratizing access. At the same time, it has created a state of complacency that AI compute will be there when we need it. AI, however, is not the same as IT.

Today, when nations plan for AI, they gloss over AI compute, with policies focused almost exclusively on data and algorithms. No government leader can answer three fundamental questions: How much domestic AI compute capacity do we have? How does this compare to other nations? And do we have enough capacity to support our national AI ambitions? This lack of uniform data, definitions, and benchmarks leaves government leaders (and their scientific advisors) unequipped to formulate a comprehensive plan for AI compute investments.

Recognizing this blind spot, the Organisation for Economic Cooperation and Development (OECD) recently established a new task force to tackle the issue. Theres nothing that helps our member countries assess what [AI compute] they need and what they have, and so some of them are making large but not necessarily well-informed investments in AI compute, Karine Perset, head of the OECD AI Policy Observatory, told VentureBeat in January.

Measuring domestic AI compute capacity is a complex task compounded by a paucity of good data. While there are widely accepted standards for measuring the performance of individual AI systems, they use highly technical metrics and dont apply well to nations as a whole. There is also the nagging question of capacity versus effective use. What if a nation acquires sufficient AI compute capacity but lacks the skills and ecosystem to effectively use it? In this case, more public investment may not lead to more public benefit.

Nonetheless, more than sixty national AI strategies were formulated and published from June 2017 to December 2020. These plans average sixty-five pages and coalesce around a common set of topics including transforming legacy industries, expanding opportunities for AI education, advancing public-sector AI adoption, and promoting responsible or trustworthy AI principles. Many national plans are aspirational and not detailed enough to be operational.

In conducting a survey of forty national AI plans, and analyzing the scope and depth of content around AI compute, we found that on average national AI plans had 23,202 words while sections on AI compute averaged only 269 words. Put simply, 98.8 percent of the content of national AI strategies focuses on vision, and only 1.2 percent covers the infrastructure needed to execute this vision.

This would be the equivalent of a national transportation strategy that devotes less than 2 percent of its recommendations to roads, bridges, and highways. Just like transportation investments, AI compute capacity will be a crucial driver of the future wealth of nations. And yet, national AI plans generally exclude detailed recommendations on this topic. There are exceptions, of course. France, Norway, India, and Singapore had particularly comprehensive sections on AI infrastructure, some with detailed recommendations for AI compute requirements, performance, and system size.

Exponential growth in computational power has led to astonishing improvements in AI capabilities, but it also raises questions of inequality in compute distribution. Fairness, inclusion, and ethics are now center stage in AI policy discussionsand these apply to compute, even though it is often ignored or relegated to side workshops at major conferences. Unequal access to computational power as well as the environmental cost of training computationally complex models require greater attention.

Some of the more popular AI models are large neural networks that run on state-of-the-art machines. For example, AlphaGo Zero and GPT-3 require millions of dollars in AI compute. This has led to AI research being dominated and shaped by a few actors mostly affiliated with big tech companies or elite universities. Governments may have to step up and reduce the compute divide by developing national research clouds. Initiatives in Europe and China are underway to develop indigenous high-performance computing (HPC) technologies and related computing supply chains that reduce their dependence on foreign sources and promote technology sovereignty.

Governments in the United States, Europe, China, and Japan have been making substantial investments in the exascale race, a contest to develop supercomputers capable of one billion billion (1018) calculations per second. The rationale behind investments in HPC is that the benefits are now going beyond scientific publications and prestige, as computing capabilities become a necessary instrument for scientific discovery and innovation. The machines are the engine for economic prosperity.

Understanding compute requirements is a non-trivial measurement challenge. At a micro level, it is important to scientifically understand the relative contribution of AI compute to driving AI progress on, for example, natural-language understanding, computer vision, and drug discovery. At a macro level, nations and companies need to assess compute requirements using a data-driven approach that accounts for future industrial pathways. Wouldnt it be eye-opening to benchmark how much compute power is being used at the company or national level and to help evaluate future compute needs?

In 2019, Stanford University organized a workshop with over one hundred interdisciplinary experts to better understand opportunities and challenges related to measurement in AI policy. The participants unanimously agreed that growth in computational power is leading to measurable improvements in AI capabilities, while also raising issues of the efficiency of AI models and the inherent inequality of compute distribution. Getting measurement right will be a stepping stone to addressing this blind spot in national AI policymaking. Going forward, we recommend that governments pivot to develop a national AI compute plan with detailed sections on capacity, effectiveness, and resilience.

The completeness of a national AI strategy forecasts that nations ability to compete in the digital global economy. Few national AI strategies, however, reflect a robust understanding of domestic AI compute capacity, how to use it effectively, and how to structure it in a resilient manner. Nations need to take action by measuring and planning for the computational infrastructure needed to advance their AI ambitions. The future of their economies is at stake.

Saurabh Mishra is an economist and former researcher at Stanford Universitys Institute for Human-Centered Artificial Intelligence.

Keith Strier is the chair of the AI Compute Taskforce at the Organization for Economic Cooperation and Development, and vice president for Worldwide AI Initiatives at the NVIDIA corporation.

Wed, Mar 24, 2021

Cyber conflict has not yet escalated from a fight inside cyberspace to a more traditional armed attack because of cyberspace. In part, this is because countries understand there are some tacit upper limits to escalation above which the response from the offended country will be war. Unfortunately, this happy state may not last.

New AtlanticistbyJason Healey and Robert Jervis

Tue, Mar 16, 2021

Theres a silver lining to parenting in a pandemic: Its an education in the core concepts of international relations, as well as a useful reminder that were all operating in a condition of anarchy. Heres a moms primer on deterrence, coercion, credibility, and reassurance.

New AtlanticistbyEmma Ashford, Erica Borghard

Read the original:
Computing to win: Addressing the policy blind spot that threatens national AI ambitions - Atlantic Council

Introduction

The March 2016 defeat of Go world champion Lee Sedol by DeepMind, Alphabets artificially intelligent AlphaGo algorithm, will be remembered as a crucial turning point in the U.S.-China relationship. Oxfords Future of Humanity Institute branded the event Chinas Sputnik moment: a moment of realization among its political and military leaders that artificial intelligence (A.I.) could be Chinas key to achieving global hegemony and dominance over the United States.

Since then, Chinas government has plowed ahead in developing its A.I. capabilities, with President Xi Jinping calling for his country to become a world leader[1] as fast as possible. Given that A.I. technologies could contribute an estimated $112 billion to the Chinese economy by 2030[2], it is no surprise that Beijing believes A.I. to be a new focus of international competition.[3]

For the simple reason that China has focused on pragmatic, collaborative policies rather than restrictive, unilateral ones, it is currently on track to overtake the United States in the A.I. race. A.I.s potentially devastating military applications make the A.I. race not just a struggle for economic dominance, but also a national security threat for whichever State loses the advantage.

While the Biden administrations readiness to boost research and development (R&D) spending and reverse the previous administrations assault on U.S. alliances is a promising first step towards combating this technological challenge, much more is necessary to ensure that the United States technological capabilities do not fall behind those of rising powers.

To hold the line, the United States must leverage its historic alliances with Europe, Australia, and Southeast Asia to pool R&D funding into a multilateral A.I. research group. By creating incentives for scientists across the globe to collaborate together on U.S.-led A.I. development, the United States can ensure that its allies and partners maintain a technological edge over China long into the future.

A.I. Scare

In addition to its commercial benefits, the rapid development of A.I. will also advantage China by adding a potentially devastating tool to its cyberwar arsenal. Concerns about weaponized A.I. have recently been raised by the United Kingdoms Government Communications Headquarters. Their 2020 report claimed that military A.I. would facilitate the rapid adaptation of malware and require a speed of response far greater than human decision-making allows, thus making it difficult for countries to defend against it with current software. The conclusion that many experts have drawn is that the threat of A.I. cyberattacks necessitates the development of defensive A.I. by countries at risk of being targeted.

This threat is not merely theoretical; indeed, China has repeatedly indicated its intention to leverage new technologies like A.I. for offensive purposes. From a 2010 hack of Google by a group with ties to Chinas Peoples Liberation Army to a suspected cyberattack on Australian political institutions in 2020, it is clear that China will not shy away from utilizing the military applications of its emerging technologies.

In fact, Chinas 2017 Next Generation Artificial Intelligence Development Plan made this explicit. The report called for enhancing A.I. civil-military integration by establishing lines of communication and coordination between research institutions, private companies, and the military. Given that any future A.I. cyberattacks could be aimed at U.S. allies and interests, it is vital that the United States prioritizes the development of its own A.I. capabilities to defend against novel techniques.

Unfortunately, research shows that the United States is somewhat unprepared for incoming attacks. While China funneled an estimated $70 billion into A.I. in 2020 (up from $12 billion in 2017), the United States government devoted only $4.9 billiona quarter of what was allocated to the Chinese port of Tianjin for A.I. development alone. It was encouraging to see the Trump administration unveil its American A.I. initiative in response to Chinas 2017 plan albeit with a 19 month delay yet this was only a first step in the right direction. A multilateral strategy is also necessary to prevent China from overtaking the United States in a crucial sector which has the potential to tip the global balance of power.

The Xi Doctrine

The forward-leaning policies initiated by President Xi Jinping have led to many advancements, accelerating Chinas A.I. program and imperiling U.S. national security in the process. One of Xis most effective initiatives has been the so-called thousand talents plan, which offers high salaries and tempting benefits to scientists and researchers who agree to work with China on emerging technologies.[4] The plan has been enormously successful: a CIA official estimated that as many as 10,000 scientists from around the world have participated.

Its potential to grant China a strategic edge over the United States and its allies has also led the U.S. Senate to label the program a threat to American interests. Concerns center around the risk that U.S.-based scientists participating in the plan could transfer research achievements from American laboratories to Chinese ones, thereby accelerating Chinese A.I. development at the United States expense.

Instead of mitigating the issue, Trump era policy responses exacerbated Chinas lead by focusing on increasing A.I. export restrictions. In an attempt to prevent the outflow of sensitive military technologies to China and other hostile states, the U.S. Department of Commerce imposed restrictions on the exports of A.I. technologies. Far from giving the United States a competitive edge, the policy likely stymied A.I. investment by requiring businesses to obtain licenses, a requirement which elongates the export process and imposes high compliance costs on struggling startups. Proof of these policies damaging effects came in 2017 when, for the first time ever, Chinese A.I. startups received a greater share of global venture funding than U.S. startups received.

The Washington Pact

In order to improve U.S. A.I. policy, it is vital that the Biden administration understands two points. First, greater R&D spending is necessary to ensure that the United States can keep up with China on A.I. For the most part, the new administration has embraced this: Bidens campaign reiterated former Google CEO Eric Schmidts assertion that the United States must boost tech R&D because China is on track to surpass the U.S. in R&D. It even went on to claim that Chinas main reason for investing in new technologies was to overtake American technological primacy and dominate future industries.

Second, because American allies are themselves investing heavily into A.I., it is prudent to adopt multilateral solutions which leverage the United States historic alliances as opposed to unilateral America first responses. For instance, Germanys A.I. Made in Germany plan has allocated 3 billion to A.I. research over the next five years, while Frances A.I. for Humanity initiative has injected 1.5 billion into the sector. To balance against Chinas advancements, the United States should take advantage of these alliances and ensure that global investments go into developing A.I. capabilities across the broader liberal democratic sphere.

This second necessity does not appear to have received as much attention from the Biden administration so far. Despite its general recommitment to multilateralism through rejoining the Paris Climate Accord, reprioritizing NATO, and calling for a Summit for Democracy, the Biden administration has largely overlooked the idea of multilateral cooperation on A.I. research.

To match the Chinese technological challenge, the United States must establish research initiatives alongside its historic allies which will benefit U.S. A.I. development. This will have the effect of protecting U.S. national security long into the future by guaranteeing that the United States retains the edge over China in crucial A.I. innovations.

At the center of this policy should be an upgraded equivalent of Chinas thousand talents scheme that would be run as a joint initiative between America and its allies. The European Union, United Kingdom, Australia, and Japans determination to invest heavily into A.I., paired with their historic ties to the United States, suggests potential for large-scale multilateral research collaboration led by the United States.

The Biden administration should therefore suggest the foundation of a multilateral research programcall it One Thousand and One Talentswith the aim of attracting the best A.I. specialists from around the globe. Participating governments would funnel their annual A.I. budgets into the scheme in order to fund research projects with important military and commercial applications. The program would ensure that salaries would be directly competitive with Chinas thousand talents program and that incentives would be put in place to make the Western alternative more attractive than the Chinese one. Like NATO, U.S. leadership would be justified by its status as the main benefactor of the scheme.

The emphasis on multilateralism as a response to U.S.-Chinese competition should come as no surprise. As Princeton professor John Ikenberry writes, the key thing for U.S. leaders to remember when dealing with China is that it may be possible for China to overtake the United States alone, but it is much less likely that China will ever manage to overtake the Western order. It is no different with A.I.

Conclusion

The new technological challenges facing America call for a far-sighted and judicious foreign policy worthy of the worlds greatest superpower. While China may have the advantages of unrestricted State investment and well-planned incentive programs, it lacks alliances that run as deep as the NATO friendships the United States has long depended on. To overcome current Chinese advancements in A.I., the United States must unite with its partners around the world in order to increase the talent, funding, and skill available to it.

The proposed Thousand And One Talents research scheme would boost the United States competitiveness vis-a-vis China by pooling the resources of some of the wealthiest and most technologically advanced nations into U.S.-led A.I. development. Given the inevitability of Chinas rise, multilateral cooperation with like-minded democracies is the only way of ensuring that the U.S. does not face an existential security threat in the future.

The Biden administration must rise to the challenge by uniting with U.S. allies to compete with China on A.I. It is too risky to go it alone.

Editors Note:An earlier version of this essay received an honorable mention in New AmericasReshaping U.S. Security Policy for the COVID Eraessay competition.

[1] Quoted in Strittmatter, Kai, We Have Been Harmonized: Life in Chinas Surveillance State, p.165

[2] Ibid, p.166-167

[3] Quoted in Ibid, p.166-167

[4] Strittmatter, Kai, We Have Been Harmonized: Life in Chinas Surveillance State, p.171

Follow this link:
One Thousand and One Talents: The Race for AI Dominance - Just Security

Get inside Wall Street with StreetInsider Premium. Claim your 1-week free trial here.

OXFORD, United Kingdom, March 22, 2021 (GLOBE NEWSWIRE) -- Diffblue, creators of the worlds first AI for code solution that automates writing unit tests for Java, today announced that its free IntelliJ plugin, Diffblue Cover: Community Edition, is now available to use to create unit tests for all of an organizations Java code both open source and commercial.

Free for any individual user, the IntelliJ plugin is availablehere for immediate download. It supports both IntelliJ versions 2020.02 and 2020.03. The Diffblue Cover: Community Edition to date has already automatically created nearly 150,000 Java unit tests!

Diffblue also offers a professional version for commercial customers who require premium support as well as indemnification and the ability to write tests for packages. In addition, Diffblue offers a CLI version of Diffblue Cover, perfect for teams to collaborate using.

Diffblues pioneering technology, developed by researchers from the University of Oxford, is based on reinforcement learning, the same machine learning strategy that powered AlphaGo, Alphabet subsidiary DeepMinds software program that beat the world champion player of Go.

Diffblue Cover automates the burdensome task of writing Java unit tests, a task that takes up as much as 20 percent of Java developers time. Diffblue Cover creates Java tests at speeds 10X-100X faster than humans that are also easy for developers to understand, and automatically maintains the tests as the code evolves even on applications with tens of millions of lines of code. Most unit test generators create boilerplate code for tests, rather than tests that compile and run. These tools guess the inputs that can be used as a starting point, but developers have to finish them to get functioning tests. Diffblue Cover is uniquely able to create complete human-readable unit tests that are ready to run immediately.

Diffblue Cover today supports Java, the most popular enterprise programming language in the Global 2000. The technology behind Diffblue Cover can also be extended to support other popular programming languages such as Python, Javascript and C#.

About DiffblueDiffblue is leading the automation of software creation through the power of AI. Founded by researchers from the University of Oxford, Diffblue Cover uses AI for code to write unit tests that help software teams and organizations efficiently improve their code coverage and quality and to ship software faster, more frequently and with fewer defects. With customers including AWS and Goldman Sachs, Diffblue is venture-backed by Goldman Sachs and Oxford Sciences Innovation. Follow us on Twitter:@diffblueHQ

Editorial contact DiffblueLonn Johnston, Flak42lonn@flak42.com+1.650.219.7764

Original post:
Diffblue's First AI-Powered Automated Java Unit Testing Solution Is Now Free for Commercial and Open Source Software Developers - StreetInsider.com