AI applications in healthcare can literally change patients lives, improving diagnostics and treatment and helping patients and healthcare providers make informed decisions quickly.

AI in the global healthcare market (the total value of products and services sold) was valued at $2.4 billion in 2019 and is projected to reach $31.02 billion in 2025.

Now in the COVID-19 pandemic, AI is being leveraged to identify virus-related misinformation on social media and remove it. AI is also helping scientists expedite vaccine development, track the virusand understand individual and population risk, among other applications.

Companies such as Microsoft, which recently stated it will dedicate $20 million to advance the use of artificial intelligence in COVID-19 research, recognize the need for and extraordinary potential of AI in healthcare.

The ultimate goal of AI in healthcare is to improve patient outcomes by revolutionizing treatment techniques. By analyzing complex medical data and drawing conclusions without direct human input, AI technology can help researchers make new discoveries.

Various subtypes of AI are used in healthcare. Natural language processing algorithms give machines the ability to understand and interpret human language. Machine learning algorithms teach computers to find patterns and make predictions based on massive amounts of complex data.

AI is already playing a huge role in healthcare, and its potential future applications are game-changing. Weve outlined four distinct ways that AI is transforming the healthcare industry.

This transformative technology has the ability to improve diagnostics, advance treatment options, boost patient adherence and engagement, and support administrative and operational efficiency.

AI can help healthcare professionals diagnose patients by analyzing symptoms, suggesting personalized treatments and predicting risk. It can also detect abnormal results.

Analyzing symptoms, suggesting personalized treatments and predicting risk

Many healthcare providers and organizations are already using intelligent symptom checkers. This machine learning technology asks patients a series of questions about their symptoms and, based on their answers, informs them of appropriate next steps for seeking care.

Buoy Health offers a web-based, AI-powered health assistant that healthcare organizations are using to triage patients who have symptoms of COVID-19. It offers personalized information and recommendations based on the latest guidance from the Centers for Disease Control and Prevention.

Additionally, AI can take precision medicine healthcare tailored to the individual to the next level by synthesizing information and drawing conclusions, allowing for more informed and personalized treatment. Deep learning models have the ability to analyze massive amounts of data, including information about a patients genetic content, other molecular/cellular analysis and lifestyle factors and find relevant research that can help doctors select treatments.

AI can also be used to develop algorithms that make individual and population health risk predictions in order to help improve outcomes. At the University of Pennsylvania, doctors used a machine learning algorithm that can monitor hundreds of key variables in real time to anticipate sepsis or septic shock in patients 12 hours before onset.

Detecting disease

Imaging tools can advance the diagnostic process for clinicians. The San Francisco-based company Enlitic develops deep learning medical tools to improve radiology diagnoses by analyzing medical data. These tools allow clinicians to better understand and define the aggressiveness of cancers. In some cases, these tools can replace the need for tissue samples with virtual biopsies, which would aid clinicians in identifying the phenotypes and genetic properties of tumors.

These imaging tools have also been shown to make more accurate conclusions than clinicians. A 2017 study published in JAMA found that of 32 deep learning algorithms, seven were able to diagnose lymph node metastases in women with breast cancer more accurately than a panel of 11 pathologists.

Smartphones and other portable devices may also become powerful diagnostic tools that could benefit the areas of dermatology and ophthalmology. The use of AI in dermatology focuses on analyzing and classifying images and the ability to differentiate between benign and malignant skin lesions.

Using smartphones to collect and share images could widen the capabilities of telehealth. In ophthalmology, the medical device company Remidio has been able to detect diabetic retinopathy using a smartphone-based fundus camera, a low-power microscope with an attached camera.

AI is becoming a valuable tool for treating patients. Brain-computer interfaces could help restore the ability to speak and move in patients who have lost these abilities. This technology could also improve the quality of life for patients with ALS, strokes, or spinal cord injuries.

There is potential for machine learning algorithms to advance the use of immunotherapy, to which currently only 20% of patients respond. New technology may be able to determine new options for targeting therapies to an individuals unique genetic makeup. Companies like BioXcel Therapeutics are working to develop new therapies using AI and machine learning.

Additionally, clinical decision support systems can help assist healthcare professionals make better decisions by analyzing past, current and new patient data. IBM offers clinical support tools to help healthcare providers make more informed and evidence-based decisions.

Finally, AI has the potential to expedite drug development by reducing the time and cost for discovery. AI supports data-driven decision making, helping researchers understand what compounds should be further explored.

Wearables and personalized medical devices, such as smartwatches and activity trackers, can help patients and clinicians monitor health. They can also contribute to research on population health factors by collecting and analyzing data about individuals.

These devices can also be useful in helping patients adhere to treatment recommendations. Patient adherence to treatment plans can be a factor in determining outcome. When patients are noncompliant and fail to adjust their behaviors or take prescribed drugs as recommended, the care plan can fail.

The ability of AI to personalize treatment could help patients stay more involved and engaged in their care. AI tools can be used to send patients alerts or content intended to provoke action. Companies like Livongo are working to give users personalized health nudges through notifications that promote decisions supporting both mental and physical health.

AI can be used to create a patient self-service model an online portal accessible by portable devices that is more convenient and offers more choice. A self-service model helps providers reduce costs and helps consumers access the care they need in an efficient way.

AI can improve administrative and operational workflow in the healthcare system by automating some of the process. Recording notes and reviewing medical records in electronic health records takes up 34% to 55% of physicians time, making it one of the leading causes of lost productivity for physicians.

Clinical documentation tools that use natural language processing can help reduce the time providers spend on documentation time for clinicians and give them more time to focus on delivering top-quality care.

Health insurance companies can also benefit from AI technology. The current process of evaluating claims is quite time-consuming, since 80% of healthcare claims are flagged by insurers as incorrect or fraudulent. Natural language processing tools can help insurers detect issues in seconds, rather than days or months.

More here:

How AI is revolutionizing healthcare - Nurse.com

As healthcare businesses transform for a post-COVID-19 era, they are embracing digital technologies as essential for outmaneuvering the uncertainty faced by businesses and as building blocks for driving more innovation. Maturing digital technologies such as social, mobile, analytics and cloud (SMAC); emerging technologies such as distributed ledger, artificial intelligence, extended reality and quantum computing (DARQ);and scientific advancements (e.g., CRISPR, materials science) are helping to make innovative breakthroughs a reality.

These technologies are also proving essential in supporting COVID-19 triage efforts. For example, hospitals in China are using artificial intelligence (AI) to scan lungs, which is reducing the burden on healthcare providers and enabling earlier intervention. Hospitals in the United States are also using AI to intercept individuals with COVID-19 symptoms from visiting patients in the hospital.

Because AI and machine learning (ML) definitions can often be confused, it may be best to start by defining our terms.

AI can be defined as a collection of different technologies that can be brought together to enable machines to act with what appears to be human-like levels of intelligence. AI provides the ability for technology to sense, comprehend, act and learn in a way that mimics human intelligence.

ML can be viewed as a subset of AI that provides software, machines and robots the ability to learn without static program instructions.

ML is currently being used across the health industry to generate personalized product recommendations to consumers, identify the root cause of quality problems and fix them, detect healthcare claims fraud, and discover and recommend treatment options to physicians. ML-enabled processes rely on software, systems, robots or other machines which use ML algorithms.

For the healthcare industry, AI and ML represent a set of inter-related technologiesthat allow machines to perform and help with both administrative and clinical healthcare functions. Unlike legacy technologies that are algorithm-based tools that complement a human, health-focused AI and ML today can truly augment human activity.

The full potential of AI is moving beyond mere automation of simple tasks into a powerful tool enabling collaboration between humans and machines. AI is presenting an opportunity to revolutionize healthcare jobs for the better.

Recent research indicates that in order to maximize the potential of AI and to be digital leaders, healthcare organizations must re-imagine and re-invent their processes and create self-adapting, self-optimizing living processes that use ML algorithms and real-time data to continuously improve.

In fact, theres consensus among healthcare organizations hat ML-enabled processes help achieve previously hidden or unobtainable value, and that these processes are finding solutions to previously unsolved business problems.

Despite these key findings, additional research surprisingly finds that only 39% of healthcare organizations report that they have inclusive design or human-centric design principles in place to support human-machine collaboration. Machines themselves will become agents of process change, unlocking new roles and new ways for humans and machines to work together.

In order to tap into the unique strengths of AI, healthcare businesses will need to rely on their peoples talent and ability to steward, direct, and refine the technology. Advances in natural language processing and computer vision can help machines and people collaborate and understand one another and their surroundings more effectively. It will be vital to prioritize explainability to help organizations ensure that people understand AI.

Powerful AI capabilities are already delivering profound results across other industries such as retail and automotive. Healthcare organizations now have an opportunity to integrate the new skills needed to enable fluid interactions between human and machines and adapt to the workforce models needed to support these new forms of collaboration.

By embracing the growing adoption of AI, healthcare organizations will soon see the potential benefits and value of AI such as organizational and workflow improvements that can unleash improvements in cost, quality and access. Growth in the AI health market is expected to reach $6.6 billion by 2021 thats a compound annual growth rate of 40%. In just the next couple of years,the health AI market will grow more than 10 times.

AI generally, and ML specifically, gives us technology that can finally perform specialized nonroutine tasks as it learns for itself without explicit human programing shifting nonclinical judgment tasks away from healthcare enterprise workers.

What will be key to the success of healthcare organizations leveraging AI and ML across every process, piece of data and worker? When AI and ML are effectively added to the operational picture, we will see healthcare systems where machines will take on simple, repetitive tasks so that humans can collaborate on a larger scale and work at a higher cognitive level. AI and ML can foster a powerful combination of strategy, technology and the future of work that will improve both labor productivity and patient care.

Brian Kalis is a managing director of digital health and innovation for Accenture's health business.

See the rest here:

Now More Than Ever We Should Take Advantage of the Transformational Benefits of AI and ML in Healthcare - Managed Healthcare Executive

If you want a glimpse of the future, check out how developers are using gpt-3.

This natural language processor was trained on parameters ten times greater than its most sophisticated rival and can be used to answer questions and write astoundingly well. Creative professionals everywhere, from top coders to professional writers marvel at what gpt-3 can produce even now in its relative infancy.

Yesterday, New York Times tech columnist Farhad Manjoo wrote that the short glimpse the general public has taken of gpt-3 is at once amazing, spooky, humbling, and more than a little terrifying. GPT-3 is capable of generating entirely original, coherent, and sometimes even factual prose. And not just prose it can write poetry, dialogue, memes, computer code, and who knows what else. Manjoo speculated on whether a similar but more advanced AI might replace him someday.

On the other hand, a recent Technology Review article describes the AI as Shockingly good and completely mindless. After describing some of the gpt-3 highlights the public has seen so far, it concedes, For one thing, the AI still makes ridiculous howlers that reveal a total lack of common sense. But even its successes have a lack of depth to them, reading more like cut-and-paste jobs than original compositions.

Wired noted in a story last week, GPT-3 was built by directing machine-learning algorithms to study the statistical patterns in almost a trillion words collected from the web and digitized books. The system memorized the forms of countless genres and situations, from C++ tutorials to sports writing. It uses its digest of that immense corpus to respond to a text prompt by generating new text with similar statistical patterns. The results can be technically impressive, and also fun or thought-provoking, as the poems, code, and other experiments attest. But the article also stated that gpt-3, often spews contradictions or nonsense, because its statistical word-stringing is not guided by any intent or a coherent understanding of reality.

Gpt-3 is the latest iteration of language-processing machine learning program from Open AI, an enterprise funded in part by Elon Musk, and its training is orders of magnitude more complex than either its previous offering or the closest competitor. The program is currently in a controlled beta test where whitelisted programmers can make requests and run projects on the AI. According to Technology Review, For now, OpenAI wants outside developers to help it explore what GPT-3 can do, but it plans to turn the tool into a commercial product later this year, offering businesses a paid-for subscription to the AI via the cloud.

Gpt-3 provides a staggering glimpse of what the future can be. Simple computer tasks can be built and then confirmed in the AI, so it will know how to create custom buttons on your webpage. Developer Sharif Shameen built a layout generator with gpt-3 so he could simply ask for a button that looks like a watermelon and the AI would give him one.

This outcome shouldnt surprise everyone as a good natural language processor develops capabilities to translate from natural English to action or to another language, and computer code is little more than an expression of intent in a language that the computer can read. So translating simple English instructions into Python should not be impossible for a sophisticated AI that has read multiple Python drafting manuals.

Of course, some of the coding community is freaking out at the prospect of being replaced by this AI. Even legendary coder John Carmack, who pioneered 3D computer graphics in early video games like Doom and is now consulting CTO at Oculus VR, was unnerved: The recent, almost accidental, discovery that GPT-3 can sort of write code does generate a slight shiver.

OK, so gpt-3 has been trained on countless coding manuals and instruction sets. But freaketh not while gpt-3 can sometimes generate usable code, it still has no application of common sense, and therefore non-technical types cant rely on it to produce machine-readable language that can perform sophisticated tasks.

For any of you who have taken a coding course, you know that coaxing the right things out of a computer requires coders to be literal and precise in ways that are difficult for an AI to approximate. So a non-coder is likely to be frustrated with AI-generated code at this point in the process. If anything, gpt-3 is a step in the process toward easier coding, requiring a practiced software engineer to develop the right sets of questions for the AI to produce usable code quickly.

I talked about the hype cycle in one of last weeks posts, and while gpt-3 is worth the hype as an advance in AI training, where more the model has 175 billion parameters is clearly better, but it is only an impressive step in the larger process. OpenAI and its competitors will find useful applications for all of this power and continue to work toward a more general intelligence.

There are many reasons to be wary. Like others, before it, this AI picks up biases in its training, and it was trained on the internet, so expect some whoppers. Wired observed, Facebooks head of AI accused the service of being unsafe and tweeted screenshots from a website that generates tweets using GPT-3 that suggested the system associates Jews with a love of money and women with a poor sense of direction. Gpt-3 has not been trained to avoid offensive assumptions.

But the AI still has the power to astonish and may permit some incredible applications. It hasnt even been officially released as a product yet. Watch this space. As developers, writers, business executives, and artists learn to do more amazing tasks with gpt-3 (and gpt-4 and 5), we will continue to report on it.

[View source.]

See the original post:

New AI Tool GPT-3 Ascends to New Peaks, But Proves How Far We Still Need to Travel - JD Supra

When it comes to artificial intelligence, and technology in general, we as a society are often guilty of thinking of it as separate from humanity. However, AI and humanity are of course intricately entwined, as AI is built by humans. Just as the saying goes, no man is an island, the same can be said of what we build.

Artificial intelligence (AI) has now permeated all sectors of business - and for good reason.

Across multiple industries, difficult or expensive tasks can be automated by AI/ML and, as a result, catapult even failing businesses into success. AI boasts a seemingly infinite list of applications - from improving customer experience to curing sleep disorders.

However, as the use of artificial intelligence rises, so does the need for cross-industry communication on the topic.

Ai4 is unique and worth high creates a bridge between industries, leaders, and technologists. The company provides a common framework for what AI means to both enterprise and the future of our globe as we transition into a new era of responsible human-machine collaboration.

One attendee, senior manager at The Aldo Group, commented on the benefits of this approach stating, I gained great insights into how my peers and competitors are leveraging AI & ML.

Ai4 Live Event

Ai4 was started by co-founders Marcus Jecklin and Michael Weiss as a small 350-person AI for financial services conference at a hotel in Brooklyn, NY. Since then, it has grown to be the top community for industry professionals seeking to learn about artificial intelligence.

Ai4 convenes thousands of people each year and reaches tens of thousands more through offline and online events, AI enterprise trainings, AI blogs and newsletters, AI matchmaking programs, and an AI jobs board.

Ai4 2020 (originally scheduled to take place at the MGM Grand in Las Vegas, now taking place digitally) promises to be an incredibly impactful event.

By gathering leaders of enterprise from across industry, government organizations, disruptive startups, investors, research labs, academia, associations, open source projects, media and analysts, Ai4 is creating the largest and most influential venue for AI-related idea-sharing, commerce, and technological progress.

Speakers for this years event include Salahuddin Khawaja, Managing Director - Automation / Global Risk, Bank of America Merrill Lynch; Stephen Wong, Chief Informatics Officer, Houston Methodist; Ameen Kazerouni, Head of ML/AI Research and Platforms, Zappos; Barret Zoph, Staff Research Scientist, Google Brain; Meltem Ballan, Data Science Lead, General Motors.

Ai4 Live Event

The speakers were amazing, commented an Assortment & Space Analyst, BJs Wholesale regarding past Ai4 live events. They covered a wide range of topics that will certainly help push our AI initiatives forward.

Success in business can often be attributed to networking - and this is no different when it comes to technology. Networks foster the exchange of ideas, as well as mutual confidence and understanding.

They also enable best practices to be created and distributed. Herein lies the genius of Ai4s AI Matchmaking system. Through this system, Ai4 arranges digital 1-1 meetings between industry leaders and vetted AI companies from the Ai4 community.

The results of this model seem to speak for themselves, according to participants. As far as recommendations go, I dont believe there is anything currently on the market, that competes or provides as much value, as Ai4s 1:1 virtual meetings, said Founder & CEO at Medlytics.

For any technology, a lack of open channels for communication will not only stall progress, but in the case of AI, it could also mean profound impacts for society. Simply put, more perspectives we encourage in this field translates into more comprehensive discussions of ethical implications and inclusive development.

Ai4 has been able to effectively address this need in the AI community by not only facilitating (virtual) space such as their AI Slack Community, but also conversations. Webinars led by AIs industry leaders on pressing topics are frequently hosted by Ai4.

Ai4 event

Additionally, Ai4 provides AI training in the form of open enrollment courses for Data Scientists & Execs as well as enterprise AI training advisory services ensuring that the Ai4 community members remain on the cutting edge. With the explosion of AI education providers in recent years, Ai4 is using their expertise to help enterprises navigate the AI education landscape to find the optimal curriculum at a fair price.

The individual presentations and moderated panels had a great combination of thoughtful commentary and technical details, commented Founder & CEO at RCM Brain, satisfying a diverse audience of technologists and business leaders.

Perhaps now more than ever, it is crucial that we remain connected - especially when it comes to the innovations that will shape our future. Ai4 is demonstrating the right way to build an advanced technology community with global, virtual conversations made up of diverse, cross-cultural, cross-industry perspectives and led by the worlds preeminent experts.

Link:

Building The Worlds Top AI Industry Community - Lessons From Ai4 - Forbes

This week, as the heads of four of the largest and most powerful tech companies in the world were called before a virtual congressional antitrust hearing to answer inquiries into how they built and run their respective behemoths, you could see that the bloom on the rose of Big Tech has faded.

Facebooks Mark Zuckerberg, once the rascally college dropout boy genius you loved to hate, still doesnt seem to grasp the magnitude of the problem of globally destructive misinformation and hate speech on his platform. Tim Cook struggles to defend how Apple takes a 30% cut from some of its app store developers revenue a policy he didnt even establish that is a vestige of Apples mid-2000s vise grip on the mobile app market. The plucky young upstarts who founded Google are both middle-aged and have stepped down from executive roles, quietly fading away while Alphabet and Google CEO Sundar Pichai runs the show. And Jeff Bezos wears the untroubled visage of the worlds richest man.

Amazon, Apple, Facebook, and Google all created tech products and services that have undeniably changed the world, some in ways that are undeniably good. But as these tech titans moved fast and broke things, they also largely excused themselves from asking difficult ethical questions, from how they built their business empires to the impacts their products and services have on the people who use them.

As AI continues to lead the next wave of transformative technology, skating over these difficult questions is a mistake the world cant afford to repeat. Whats more, AI technologies wont actually work properly unless companies address the issues at their heart.

Smart and ruthless was the tradition of Big Tech, but AI requires people to be smart and wise. Those working in AI have to not only ensure the efficacy of what they make, but holistically understand the potential harms for people AI tech impacts. Thats a more mature and just way of building world-changing technologies, products, and services. Fortunately, many prominent voices in AI are leading the field down that path.

This weeks best example was the widespread reaction to a service called Genderify, which promised to use natural language processing (NLP) to help companies identify customers gender using only their name, username, or email address. The entire premise is absurd and problematic, and when AI folks got ahold of it to put it through its paces, they predictably found it to be terribly biased (which is to say, broken).

Genderify was such a bad joke that it almost seemed like some kind of performance art. In any case, it was laughed off the internet. Just a day or so after it was launched, the Genderify site, Twitter account, and LinkedIn page were gone.

Its frustrating to many in the field that such ill-conceived and poorly executed AI offerings keep popping up. But the swift and wholesale deletion of Genderify illustrates the power and strength of this new generation of principled AI researchers and practitioners.

The burgeoning AI sector is already experiencing the kind of reckoning Big Tech is only facing after decades. Other recent examples include an outcry over a paper that promised to use AI to identify criminality from peoples faces (really just AI phrenology), which led to the paper being withdrawn from publication. Landmark studies on bias in facial recognition have led to bans and moratoriums on the technologys use in several U.S. cities, as well as a raft of legislation to eliminate or combat its potential abuses. Fresh research is finding intractable problems with bias in well-established data sets like 80 Million Tiny Images and the legendary ImageNet and leading to immediate (if overdue) change. And theres more.

Although advocacy groups play a role in pushing for changes and posing tough questions, the authority for such inquiry and the research-based proof is coming from people inside the field of AI ethicists, researchers looking for ways to improve AI techniques, and actual practitioners.

There is, of course, an immense amount of work to be done and many more battles ahead as AI fuels the next dominant set of technologies. Look no further than problematic AI in surveillance, military, the courts, employment, policing, and more.

But seeing tech giants like IBM, Microsoft, and Amazon pull back on massive investments in facial recognition is a sign of progress. It doesnt actually matter whether their actions are narrative cover for a capitulation to other companies market dominance, a calculated move to avoid potential legislative punishment, or just a PR stunt. For whatever reason, these companies acknowledged the value of slowing down and reducing damage rather than continuing to move fast and break things.

More here:

AI Weekly: Big Techs antitrust reckoning is a cautionary tale for the AI industry - VentureBeat

As expected, artificial intelligence (AI) and machine learning (ML) applications are already having an impact on society. Many industries that we tap into dailysuch as banking, financial services and insurance (BFSI), and digitized health carecan benefit from AI and ML applications to help them optimize mission-critical operations and execute functions in real time.

The BFSI sector is an early adopter of AI and ML capabilities. Natural language processing (NLP) is being implemented for personal identifiable information (PII) privacy compliance, chatbots and sentiment analysis; for example, mining social media data for underwriting and credit scoring, as well as investment research. Predictive analytics assess which assets will yield the highest returns. Other AI and ML applications include digitizing paper documents and searching through massive document databases. Additionally, anomaly detection and prescriptive analytics are becoming critical tools for the cybersecurity sector of BFSI for fraud detection and anti-money laundering (AML).1

Scientists searching for solutions to the COVID-19 pandemic rely heavily on data acquisition, processing and management in health care applications. They are turning to AI, ML and NLP to track and contain the coronavirus, as well as to gain a more comprehensive understanding of the disease. Among the applications for AI and ML include medical research for developing a vaccine, tracking the spread of the disease, evaluating the effects of COVID-19 intervention, using natural processing of language in social media to understand the impact on society, and more.2

Processing a Data Avalanche

The fuel for BFSI applications like fraud detection, AML applications and chatbots, or health applications such as tracking the COVID-19 pandemic, are decision support systems (DSSs) containing vast amounts of structured and unstructured data. Overall, experts predict that by 2025, 79 trillion GB of data will have been generated globally.3 This avalanche of data is making data mining (DM) difficult for scalar-based high-performance computers to effectively and efficiently run a DSS for its intended applications. More powerful accelerator cards, such as vector processing engines supported by optimized middleware, are proving to efficiently process enterprise data lakes to populate and update data warehouses, from which meaningful insights can be presented to the intended decision makers.

Resurgence of Vector Processors

There is currently a resurgence in vector processing, which, due to the cost, was previously reserved for the most powerful supercomputers in the world. Vector processing architectures are evolving to provide supercomputer performance in a smaller, less expensive form factor using less power, and they are beginning to outpace scalar processing for mainstream AI and ML applications. This is leading to their implementation as the primary compute engine in high performance computing applications, freeing up scalar processors for other mission critical processing roles.

Vector processing has unique advantages over scalar processing when operating on certain types of large datasets. In fact, a vector processor can be more than 100 times faster than a scalar processor, especially when operating on the large amounts of statistical data and attribute values typical for ML applications, such as sparse matrix operations.

While both scalar and vector processors rely on instruction pipelining, a vector processor pipelines not only the instructions but also the data, which reduces the number of fetch then decode steps, in turn reducing the number of cycles for decoding. To illustrate this, consider the simple operation shown in Figure 1, in which two groups of 10 numbers are added together. Using a standard programming language, this is performed by writing a loop that sequentially takes each pair of numbers and adds them together (Figure 1a).

Figure 1: Executing the task defined above, the scalar processor (a) must perform more steps than the vector processor (b).

When performed by a vector processor, this task requires only two address translations, and fetch and decode is performed only once (Figure 1b) , rather than the 10 times required by a scalar processor (Figure 1a). And because the vector processors code is smaller, memory is used more efficiently. Modern vector processors also allow different types of operations to be performed simultaneously, further increasing efficiency.

To bring vector processing capabilities into applications less esoteric than scientific ones, it is possible to combine vector processors with scalar CPUs to produce a vector parallel computer. This system comprises a scalar host processor, a vector host running LINUX, and one or more vector processor accelerator cards (or vector engines), creating a heterogeneous compute server that is ideal for broad AI and ML workloads and data analytics applications. In this scenario, the primary computational components are the vector engines, rather than the host processor. These vector engines also have self-contained memory subsystems for increased system efficiency, rather than relying on the host processors direct memory access (DMA) to route packets of data through the accelerator cards I/O pins.

Software Matters

Processors perform only as well as the compilers and software instructions that are delivered to them. Ideally, they should be based on industry-standard programming languages such as C/C++. For AI and ML application development, there are several frameworks available with more emerging. A well designed vector engine compiler should utilize both industry-standard programming languages and open source AI and ML frameworks such as TensorFlow and PyTorch. A similar approach should be taken for database management and data analytics, using proven frameworks such as Apache Spark and Scikit-Learn. This software strategy allows for seamless migration of legacy code to vector engine accelerator cards. Additionally, by using the message passing interface (MPI) to implement distributed processing, the configuration and initialization become transparent to the user.

Conclusion

AI and ML are driving the future of computing and will continue to permeate more applications and services in the future. Many of these application deployments will be implemented in smaller server clusters, perhaps even a single chassis. Accomplishing such a feat requires revisiting the entire spectrum of AI technologies and heterogeneous computing. The vector processor, with advanced pipelining, is a technology that proved itself long ago. Vector processing paired with middleware optimized for parallel pipelining is lowering the entry barriers for new AI and ML applications, and is set to solve the challenges both today and in the future that were once only attainable by the hyperscale cloud providers.

References

About the Author

Robbert Emery is responsible for commercializing NEC Corporations advanced technologies in HPC and AI/ML platform solutions. His role includes discovering and lowering the entry point and initial investment for enterprises to realize the benefits of big data analytics in their operations. Robbert has developed a career of over 20 years in the ICT industrys emerging technologies, including mobile network communications, embedded technologies and high-volume manufacturing. Prior to joining NECs technology commercialization accelerator, NEC X Inc., in Palo Alto California, Robbert led the product and business plan for an embedded solutions company that resulted in a leadership position, in terms of both volume and revenue. He has an MBA from SJSUs Lucas College and Graduate School of Business, as well as a bachelors degree in electrical engineering from California Polytechnic State University.

Related

Read more here:

How AI and ML Applications Will Benefit from Vector Processing - EnterpriseAI

In the 18th century, Hungarian inventor Wolfgang von Kempelen created a never-before-seen chess-playing machine. The automaton, called the Mechanical Turk, could handle a game of chess against a human player, and pretty well with that: it even defeated Napoleon Bonaparte in 1809, during a campaign in Vienna.

It was eventually revealed that von Kempelen's invention was an elaborate hoax. The machine, in reality, secretly hid a human chess master who directed every move. The Mechanical Turk was destroyed in the mid-19th century; but hundreds of years later, the story provides a telling metaphor for artificial intelligence.

A common narrative that surrounds AI is that the technology has agency. We hear that AI can solve climate change, build smart cities and find new drugs, and less often that in fact, it is a human programmer who is using an AI system to achieve all of those feats. Just like the human chess master hid behind von Kempelen's ingenious mechanism, so do engineers, programmers, and software developers disappear behind the algorithm.

SEE: Managing AI and ML in the enterprise 2020: Tech leaders increase project development and implementation (TechRepublic Premium)

The relevance of the Hungarian machine is such that Amazon borrowed the name for one of its units one that is often less known than Prime or Fresh. Amazon Mechanical Turk is a division of the company that crowdsources the tedious job of labeling the huge datasets that feed AI systems to millions of remote "Turkers".

"I'm amazed that every four months or so, I catch a Tweet from someone who realizes what Amazon's Mechanical Turk is," Daniel Leufer, Mozilla fellow and technologist, tells ZDNet. "I find it fascinating that Amazon calls a platform designed to mask the human agency behind AI Mechanical Turk. We're not even hiding what we're trying to do here."

Leufer has just put the final touches to a new project to debunk common AI myths, which he has been working on since he received his Mozilla fellowship an award designed for web activists and technology policy experts. And one of the most pervasive of those myths is that AI systems can and act of their own accord, without supervision from humans.

It certainly doesn't help that artificial intelligence is often associated with humanoid robots, suggesting that the technology can match human brains. An AI system deployed, say, to automate insurance claims, is very unlikely to come in the form of a human-looking robot, and yet that is often the portrayal that is made of the technology, regardless of its application.

Leufer calls those "inappropriate robots", often shown carrying out human tasks that would never be necessary for an automaton. Among the most common offenders feature robots typing on keyboards and robots wearing headphones or using laptops.

The powers we ascribe to AI as a result even have legal ramifications: there is an ongoing debate about whether an AI system should own intellectual property (a proposal refuted by the European Patent Office and the UK Intellectual Property Office), or whether automatons should be granted citizenship. In 2017, for instance, Shibuya Mirai became the first chatbot to be granted residency in Tokyo by the Japanese government.

The current representation of AI feeds into the perception that the technology comes in one form, and one form only: a super-powerful system capable of general intelligence that is, of performing intelligently across a range of complex tasks, and eventually completing anything that a human can do.

Although achieving such a sophisticated form of artificial intelligence is not a prospect envisaged by many scientists, it seems to be the narrative that dominates even the highest level of geo-politics. "There is an entire narrative around the race for AI supremacy going on between the US, China and Europe," says Leufer. "That just doesn't make sense."

"If you believe we're headed towards an end-point, where a super-intelligence will grant you technological supremacy, then maybe it makes sense, but that's not the case. This is not a zero-sum game," he continues.

In reality, AI as we know it is still narrow. It can only solve a range of single tasks, and the step up to general intelligence is still far away in the future. But even if the anticipation of super-intelligence is currently unfounded, the consequences of misrepresenting the technology are very real.

Leufer takes the example of facial recognition, which he believes needs to be banned across the EU. The response he got from regulators, he argues, shows a lack of understanding of the technology.

"The idea is that this is a part of AI, and AI is inevitable, so we'll have to adopt it eventually and we better develop it ourselves so it is imbued with European values," says Leufer. "But AI is not just one technology. There are many ways you can use it."

SEE: CIO Jury: 58% of tech leaders say robotics will play a significant role in their industry within the next two years

Becoming a leader in industry robotics doesn't have to go hand-in-hand with developing facial recognition, just because both tap AI-enabled capabilities. It might be less exciting than the prospect of a super-intelligence, but AI is not one huge technology waiting to be cracked. In other words, artificial intelligence is not an all or nothing.

And so, as countries around the world race to develop all potential AI applications, regulation is crucial to make sure that the development of what Leufer calls "creepy stuff" is limited.

He is currently working with German NGO AlgorithmWatch to push for the creation of public registers for AI systems, in which public authorities and governments would have to provide basic information about the ways that they are using the technology, together with risk assessments, and even a way for citizens to contest the application.

"At the moment we're working in the dark, we don't know what's being used," says Leufer. Super-intelligent humanoid robots might still be a long way off, but narrow AI isn't short of issues that need fixing right now.

The rest is here:

We've forgotten the most important thing about AI. It's time to remember it again - ZDNet

When Deep Blue defeated world chess champion Garry Kasparov in 1997, it may have seemed artificial intelligence had finally arrived. A computer had just taken down one of the top chess players of all time. But it wasnt to be.

Though Deep Blue was meticulously programmed top-to-bottom to play chess, the approach was too labor-intensive, too dependent on clear rules and bounded possibilities to succeed at more complex games, let alone in the real world. The next revolution would take a decade and a half, when vastly more computing power and data revived machine learning, an old idea in artificial intelligence just waiting for the world to catch up.

Today, machine learning dominates, mostly by way of a family of algorithms called deep learning, while symbolic AI, the dominant approach in Deep Blues day, has faded into the background.

Key to deep learnings success is the fact the algorithms basically write themselves. Given some high-level programming and a dataset, they learn from experience. No engineer anticipates every possibility in code. The algorithms just figure it.

Now, Alphabets DeepMind is taking this automation further by developing deep learning algorithms that can handle programming tasks which have been, to date, the sole domain of the worlds top computer scientists (and take them years to write).

In a paper recently published on the pre-print server arXiv, a database for research papers that havent been peer reviewed yet, the DeepMind team described a new deep reinforcement learning algorithm that was able to discover its own value functiona critical programming rule in deep reinforcement learningfrom scratch.

Surprisingly, the algorithm was also effective beyond the simple environments it trained in, going on to play Atari gamesa different, more complicated taskat a level that was, at times, competitive with human-designed algorithms and achieving superhuman levels of play in 14 games.

DeepMind says the approach could accelerate the development of reinforcement learning algorithms and even lead to a shift in focus, where instead of spending years writing the algorithms themselves, researchers work to perfect the environments in which they train.

First, a little background.

Three main deep learning approaches are supervised, unsupervised, and reinforcement learning.

The first two consume huge amounts of data (like images or articles), look for patterns in the data, and use those patterns to inform actions (like identifying an image of a cat). To us, this is a pretty alien way to learn about the world. Not only would it be mind-numbingly dull to review millions of cat images, itd take us years or more to do what these programs do in hours or days. And of course, we can learn what a cat looks like from just a few examples. So why bother?

While supervised and unsupervised deep learning emphasize the machine in machine learning, reinforcement learning is a bit more biological. It actually is the way we learn. Confronted with several possible actions, we predict which will be most rewarding based on experienceweighing the pleasure of eating a chocolate chip cookie against avoiding a cavity and trip to the dentist.

In deep reinforcement learning, algorithms go through a similar process as they take action. In the Atari game Breakout, for instance, a player guides a paddle to bounce a ball at a ceiling of bricks, trying to break as many as possible. When playing Breakout, should an algorithm move the paddle left or right? To decide, it runs a projectionthis is the value functionof which direction will maximize the total points, or rewards, it can earn.

Move by move, game by game, an algorithm combines experience and value function to learn which actions bring greater rewards and improves its play, until eventually, it becomes an uncanny Breakout player.

So, a key to deep reinforcement learning is developing a good value function. And thats difficult. According to the DeepMind team, it takes years of manual research to write the rules guiding algorithmic actionswhich is why automating the process is so alluring. Their new Learned Policy Gradient (LPG) algorithm makes solid progress in that direction.

LPG trained in a number of toy environments. Most of these were gridworldsliterally two-dimensional grids with objects in some squares. The AI moves square to square and earns points or punishments as it encounters objects. The grids vary in size, and the distribution of objects is either set or random. The training environments offer opportunities to learn fundamental lessons for reinforcement learning algorithms.

Only in LPGs case, it had no value function to guide that learning.

Instead, LPG has what DeepMind calls a meta-learner. You might think of this as an algorithm within an algorithm that, by interacting with its environment, discovers both what to predict, thereby forming its version of a value function, and how to learn from it, applying its newly discovered value function to each decision it makes in the future.

LPG builds on prior work in the area.

Recently, researchers at the Dalle Molle Institute for Artificial Intelligence Research (IDSIA) showed their MetaGenRL algorithm used meta-learning to learn an algorithm that generalizes beyond its training environments. DeepMind says LPG takes this a step further by discovering its own value function from scratch and generalizing to more complex environments.

The latter is particularly impressive because Atari games are so different from the simple worlds LPG trained inthat is, it had never seen anything like an Atari game.

LPG is still behind advanced human-designed algorithms, the researchers said. But it outperformed a human-designed benchmark in training and even some Atari games, which suggests it isnt strictly worse, just that it specializes in some environments.

This is where theres room for improvement and more research.

The more environments LPG saw, the more it could successfully generalize. Intriguingly, the researchers speculate that with enough well-designed training environments, the approach might yield a general-purpose reinforcement learning algorithm.

At the least, though, they say further automation of algorithm discoverythat is, algorithms learning to learnwill accelerate the field. In the near term, it can help researchers more quickly develop hand-designed algorithms. Further out, as self-discovered algorithms like LPG improve, engineers may shift from manually developing the algorithms themselves to building the environments where they learn.

Deep learning long ago left Deep Blue in the dust at games. Perhaps algorithms learning to learn will be a winning strategy in the real world too.

Update (6/27/20): Clarified description of preceding meta-learning research to include prior generalization of meta-learning in RL algorithms (MetaGenRL).

Image credit: Mike Szczepanski /Unsplash

Follow this link:

DeepMind's Newest AI Programs Itself to Make All the Right Decisions - Singularity Hub