Daily Archives: January 13, 2022

At GE, Artificial Intelligence (AI) development is primarily focused on connecting minds and industrial machines to enable intelligent and user-friendly products and services that move, cure and power the world. GE Research spearheads this charter via the invention and deployment of AI solutions that can execute on industrial devices, at the edge or in the cloud.

AIresearch is practiced as a multidisciplinary exercise at GE, where insights from data-driven machine learning is fused with domain-specific knowledge drawn from areas such as materials, physics, biology and design engineering, to amplify the quality as well as causal-veracity of the predictions derivedwhat we call hybrid AI. We are creating state-of-the-art perception and reasoning capabilities for our AI technology to observe and understand contextual meaning, to improve the performance and life of our assets, industrial systems and human health. We are developing continuous learning systems that teach or learn from other assets or agents and learn from real and virtual experiences to understand and improve behavior.

Some key challenges we tackle include a lack of sufficient labels needed for traditional supervised learning approaches, the need to ingest and link multiple data modalities, and the need to build AI solutions that are interpretable due to safety-related regulatory requirements.

State-of-the-art capabilities in computer vision, machine learning, knowledge representation, reasoning and human system interactions are used to robustly monitor, assess and predict the performance and health of assetsinformation that, when coupled with uncertainty quantification and assurance, provides the information needed to multi-objectively optimize customer-specific metrics.

Examples of customer outcomes enhanced by AI products include reduced downtime on assets through AI-driven proactive intervention (for e.g., airline delays and cancellation), increased throughput (for e.g., optimal control of wind turbine settings to maximize farm output), or reduced costs (for e.g., optimal power plant operation to minimize fuel costs). GE Research is developing and integrating artificial intelligence in healthcare by working to incorporate the technologyinto every aspect of the patient journey (for e.g., improved disease diagnosis, augmenting doctors and clinicians by increasing workflow efficiencies to save precious time).In addition to asset-awareness and management, active AI research areas include Computer Vision, automation, autonomy, User Experience, Augmented Reality and Robotics.

Read more from the original source:

Artificial Intelligence | GE Research

Scientists, students, and community members came together last month to discuss the promise and pitfalls of artificial intelligence at MITs Computer Science and Artificial Intelligence Laboratory (CSAIL) for the fourth TEDxMIT event held at MIT.

Attendees were entertained and challenged as they explored the good and bad of computing, explained CSAIL Director Professor Daniela Rus, who organized the event with John Werner, an MIT fellow and managing director of Link Ventures; MIT sophomore Lucy Zhao; and grad student Jessica Karaguesian. As you listen to the talks today, Rus told the audience, consider how our world is made better by AI, and also our intrinsic responsibilities for ensuring that the technology is deployed for the greater good.

Rus mentioned some new capabilities that could be enabled by AI: an automated personal assistant that could monitor your sleep phases and wake you at the optimal time, as well as on-body sensors that monitor everything from your posture to your digestive system. Intelligent assistance can help empower and augment our lives. But these intriguing possibilities should only be pursued if we can simultaneously resolve the challenges that these technologies bring, said Rus.

The next speaker, CSAIL principal investigator and professor of electrical engineering and computer science Manolis Kellis, started off by suggesting what sounded like an unattainable goal using AI to put an end to evolution as we know it. Looking at it from a computer science perspective, he said, what we call evolution is basically a brute force search. Youre just exploring all of the search space, creating billions of copies of every one of your programs, and just letting them fight against each other. This is just brutal. And its also completely slow. It took us billions of years to get here. Might it be possible, he asked, to speed up evolution and make it less messy?

The answer, Kellis said, is that we can do better, and that were already doing better: Were not killing people like Sparta used to, throwing the weaklings off the mountain. We are truly saving diversity.

Knowledge, moreover, is now being widely shared, passed on horizontally through accessible information sources, he noted, rather than vertically, from parent to offspring. I would like to argue that competition in the human species has been replaced by collaboration. Despite having a fixed cognitive hardware, we have software upgrades that are enabled by culture, by the 20 years that our children spend in school to fill their brains with everything that humanity has learned, regardless of which family came up with it. This is the secret of our great acceleration the fact that human advancement in recent centuries has vastly out-clipped evolutions sluggish pace.

The next step, Kellis said, is to harness insights about evolution in order to combat an individuals genetic susceptibility to disease. Our current approach is simply insufficient, he added. Were treating manifestations of disease, not the causes of disease. A key element in his labs ambitious strategy to transform medicine is to identify the causal pathways through which genetic predisposition manifests. Its only by understanding these pathways that we can truly manipulate disease causation and reverse the disease circuitry.

Kellis was followed by Aleksander Madry, MIT professor of electrical engineering and computer science and CSAIL principal investigator, who told the crowd, progress in AI is happening, and its happening fast. Computer programs can routinely beat humans in games like chess, poker, and Go. So should we be worried about AI surpassing humans?

Madry, for one, is not afraid or at least not yet. And some of that reassurance stems from research that has led him to the following conclusion: Despite its considerable success, AI, especially in the form of machine learning, is lazy. Think about being lazy as this kind of smart student who doesnt really want to study for an exam. Instead, what he does is just study all the past years exams and just look for patterns. Instead of trying to actually learn, he just tries to pass the test. And this is exactly the same way in which current AI is lazy.

A machine-learning model might recognize grazing sheep, for instance, simply by picking out pictures that have green grass in them. If a model is trained to identify fish from photos of anglers proudly displaying their catches, Madry explained, the model figures out that if theres a human holding something in the picture, I will just classify it as a fish. The consequences can be more serious for an AI model intended to pick out malignant tumors. If the model is trained on images containing rulers that indicate the size of tumors, the model may end up selecting only those photos that have rulers in them.

This leads to Madrys biggest concerns about AI in its present form. AI is beating us now, he noted. But the way it does it [involves] a little bit of cheating. He fears that we will apply AI in some way in which this mismatch between what the model actually does versus what we think it does will have some catastrophic consequences. People relying on AI, especially in potentially life-or-death situations, need to be much more mindful of its current limitations, Madry cautioned.

There were 10 speakers altogether, and the last to take the stage was MIT associate professor of electrical engineering and computer science and CSAIL principal investigator Marzyeh Ghassemi, who laid out her vision for how AI could best contribute to general health and well-being. But in order for that to happen, its models must be trained on accurate, diverse, and unbiased medical data.

Its important to focus on the data, Ghassemi stressed, because these models are learning from us. Since our data is human-generated a neural network is learning how to practice from a doctor. But doctors are human, and humans make mistakes. And if a human makes a mistake, and we train an AI from that, the AI will, too. Garbage in, garbage out. But its not like the garbage is distributed equally.

She pointed out that many subgroups receive worse care from medical practitioners, and members of these subgroups die from certain conditions at disproportionately high rates. This is an area, Ghassemi said, where AI can actually help. This is something we can fix. Her group is developing machine-learning models that are robust, private, and fair. Whats holding them back is neither algorithms nor GPUs. Its data. Once we collect reliable data from diverse sources, Ghassemi added, we might start reaping the benefits that AI can bring to the realm of health care.

In addition to CSAIL speakers, there were talks from members across MITs Institute for Data,Systems, and Society; the MIT Mobility Initiative; the MIT Media Lab; and the SENSEableCity Lab.

The proceedings concluded on that hopeful note. Rus and Werner then thanked everyone for coming. Please continue to reflect about the good and bad of computing, Rus urged. And we look forward to seeing you back here in May for the next TEDxMIT event.

The exact theme of the spring 2022 gathering will have something to do with superpowers. But if Decembers mind-bending presentations were any indication the May offering is almost certain to give its attendees plenty to think about. And maybe provide the inspiration for a startup or two.

Here is the original post:

The promise and pitfalls of artificial intelligence explored at TEDxMIT event - MIT News

Using finely tuned hardware, a specialized network, and large data storage, supercomputers have long been used for computationally intense projects that require large amounts of data processing. With the rise of artificial intelligence and machine learning, there is an increasing demand for these powerful computers and, as a result, processing power is rapidly increasing. As such, the growth of AI is inextricably linked to the growth in processing power of these high-performing devices.

Supercomputers arent new. The term appeared in the late 1920s and the CDC 6600 (released in 1964) is generally considered to be the first true supercomputer. Early supercomputers used only a few extremely powerful processors but, in the late 1990s, computer experts realized that stringing together thousands of off-the-shelf processors would yield the greatest processing power. Current state-of-the-art supercomputers have over 60,000 massively parallel processors to approach petaflop performance levels.

Mindful of the threats to security that are posed by supercomputers, a consortium of countries, including the United States, Germany, and South Korea, developed the Wassenaar Arrangement, which restricts the sale of, among other things, supercomputers that can be used for military purposes. Nonetheless, supercomputers can be found in most countries pursuing AI research.

As such, much of the development of AI is predicated on two pillars: technologies and human capital availability. Our prior reports for Brookings, How different countries view artificial intelligence and Analyzing artificial intelligence plans in 34 countries, detailed how countries are approaching national AI plans, and how to interpret those plans. In a follow-up piece, Winners and losers in the fulfillment of national artificial intelligence aspirations, we discussed how different countries were fulfilling their aspirations along technology-oriented and people-oriented dimensions. In our most recent post, The people dilemma: How human capital is driving or constraining the achievement of national AI strategies, we discussed the people dimension and so, in this piece, we will examine how each country is prepared to meet their AI objectives in the second pillarthe technology dimension.

In order to analyze each countrys technology preparedness, we assembled a country-level dataset containing: the number and size of supercomputers in each country, the amount of public and private spending on AI initiatives in each country, the number of AI startups in each country, and the number of AI patents and conference papers each countrys scholars produced. This resulted in ten distinct data elements.[1]

As with our previous analyses, we conducted a factor analysis to determine if any of the data elements were closely related. Closely related items can be mathematically combined into a composite factor, which aids in interpretation. In this factor analysis, two clear factors emerged. The first factor contained country ranks by theoretical peak computer performance, number of processing cores, number of supercomputers, and maximal LINPACK performance achieved; country ranks for the number of conference papers and journal papers; and the country rank for the number of patents. The second factor contained private and public investments in AI. One field, AI startups, was not closely associated with either factor and was dropped from further analysis.

It is clear that all of the fields in the first factor are either directly related to technology or its use in research. As a result, we name this factor Technology and Research. The second factor is solely focused on investments, and so we name this field Investments.

Figure 1 shows where a select group of countries sit along these sub-dimensions.

We interpret and name the quadrants as follows. The countries that are in the upper right-hand corner we dub Leaders; these have both a robust technology and research platform (factor one) and substantial public/private investments (factor two). Countries in the lower right quadrant we dub Technology Skilled. These countries have a strong current technology and research platform but are lacking strong public and private investments. Countries in the upper left quadrant we dub Funding Positioned, and are countries that have a strong funding stream but are behind in terms of technology and research. Finally, we dub the lower left quadrant Unprepared, which reflects countries that are both lacking in technology and research and are also lacking from a funding perspective.

The race for technology dominance is clearly a two-horse race between the U.S. (94th percentile for technology and research and 96th percentage for investment) and China (94th percentile for technology and research and 91st percentage for investments). While the U.S. holds a very slight lead overall, both countries are in the top three positions for every single one of our data elements. This is not surprising, as the size of the U.S. and Chinese economies (largest and second-largest respectively at $20 trillion and $15 trillion respectively) dwarf Japan, which is the third-largest economy ($4.9 trillion). As a result, we see no technology-centric hindrances for either country to continue to excel.

The United Kingdom (75th percentile in technology and research and 88th percentile in investments), France (75th percentile in technology and research and 81st percentile in investments), Japan (87th percentile in technology and research and 75th percentile in investments), and Germany (83rd percentile in technology and research and 68th percentile in investments) form a strong chase pack to the two leaders. Of the four countries, we view the United Kingdom as being in the strongest position to challenge the U.S. and China and this is based on their stronger investments in technology. We feel that these investments will allow them to close the gap more quickly than the other countries are capable of. However, we cannot ignore the fact that Japans economy is the largest of the four and this suggests that, if they decide to do so, they can quickly accelerate their efforts.

India (57th percentile in technology and research and 78th percentile in investments), Canada (68th percentile in technology and research and 60th percentile in investments), South Korea (71st percentile in technology and research and 60th percentile in investments), and Italy (71st percentile in technology and research and 60th percentile in investments) complete the Leaders quadrant. As with the United Kingdom, India is also well-positioned from a funding standpoint and should quickly separate itself from the other four countries.

Almost without exception, there is a strong relationship between the countrys economic size and its position in our quadrants. The U.S. is certainly leveraging its status as the worlds largest economy and is making dramatically larger technology investments than almost any other country; China is a close second. While we were concerned with the U.S. position from a people perspective, there are no similar concerns from a technology standpoint. America remains a world leader in digital innovation and supercomputers are no exception to that fact.

The uncomfortable reality for the U.S. is that its economic strength is very helpful to make the necessary technology infrastructure investments which are necessary but not sufficient to be successful in the pursuit of AI but this economic strength has little or no bearing on the other necessary element the ability to attract the necessary people to develop and implement its AI strategy. By contrast, China also has the economic strength for the necessary technology infrastructure investments but also has a sizeable population to provide the people power which is also necessary. In other words, China has both conditions necessary for AI success while the U.S. only has one of them. As such, China is currently in far better shape than the U.S. to achieve its AI goals and, without changes on the people front, the U.S. will fall increasingly far behind.

In our next post, we will exclusively focus on what the U.S. needs to do to improve its position and in our subsequent posts, we will examine different teaming strategies that leverage each countrys respective strengths.

[1]: These were: Rpeak (country rank by theoretical peak computer performance), Cores (country rank by number of processing cores), Count (country rank by number of supercomputers), Rmax (country rank by maximal LINPACK floating point calculation performance achieved), AI Startups (country rank for number of AI-based startups), Private Investment (country rate for private investments in AI), Public Investments (country rank for public investments in AI), AI Conference Papers (country rank for number of AI conference papers), AI Journal Papers (country rank for number of AI papers) and AI Patents (country rank for number of AI patents).

See the original post here:

How countries are leveraging computing power to achieve their national artificial intelligence strategies - Brookings Institution

What can AI do for you? A lot, as the recent Intel AI Summit 2021 has shown. Artificial intelligence opens up a broad range of possibilities, from tiny devices to the massive cloud. Suppose you dont know where to start or how to develop and scale up your ideas and innovation further. In that case, the two-day summits contents are now available on-demand to inspire you with the latest from the Intel AI technology stack, as well as successful customer use cases from the Asia Pacific and Japan Territory (APJ-T).

AI is no longer the purview of those in the know. In the executive keynote, Dr. Nash Palaniswamy gave an overview of Intels AI strategy: Taking AI from being specialized, proprietary and for the few to being ubiquitous, open and for all. The Vice President, SMG, General Manager, AI, HPC, and Datacenter Accelerators Solutions and Sales at Intel also shared the companys cutting-edge offerings in the field from hardware and software applications to AI deployment in the cloud and edge ecosystems. His address was just the beginning of a virtual event featuring over 20 industry expert speakers leading more than 25 innovative sessions, which you can view at your leisure until May 2022.

An overview of how users implement AI on the Intel Architecture (IA) platform was presented by Hong Wei Yi, Asia AI Sales Director, DCG Sales Group at Intel. She curated some examples of how AI and the platform led to easier deployment, better performance, and lower total cost of ownership (TCO) in various applications such as drug discovery, neuro-linguistic programming (NLP), and more.

Get a front-row seat to the fireside chats and hear how organisations make wonderful things with Intel AI, such as Tokopedia, the largest e-commerce start-up in Indonesia, and Max Kelson, an AI consultancy based in Australia.

The fireside chat with Tokopedia brought together three speakers representing the collaboration between the start-up, Intel and Google Cloud Platform (GCP), that helped the company scale up and turn AI into ROI (return on investment): Tahir Hashmi, technical fellow, and VP of growth engineering at Tokopedia; Erwin Huizenga, APAC Solution Lead Machine Learning at GCP; and Ayu Ginanti, APJ-T Cloud Lead at Intel. They discussed how they achieved success, from overcoming funding shortfalls, skills, and data challenges to measuring the success of AI projects as they went from lab demo to live, daily.

Tokopedia was launched in 2009 and became a unicorn six years later. It has been using Google Cloud since 2018. In May, it merged with ride-hailing and payments unicorn GoJek to create the GoTo Group. GoTo is currently Indonesias most prominent digital services platform and contributes about 2% of the countrys GDP, with more than 100 million active users. Last July, GoTo announced its collaboration with Google Cloud for its next growth phase.

We look forward to our continued partnership with Intel and Google Cloud as a key technology partner to support GoTos continued expansion across cloud infrastructure, data with cloud artificial intelligence (AI) and machine learning (ML), as well as productivity and collaboration needs with Google Workspace. We hope that this partnership can also empower us to provide the convenience of accessing high availability and scalable services from anywhere at any time for business, especially MSMEs, and consumers, said Herman Widjaja, Chief Technology Officer, Tokopedia, in the announcement.

Google Cloud is supported by Intel architecture, which provides the most demanding enterprise workloads and applications security, compute, and memory requirements.

Meanwhile, the meldCX breakout presentation shed light on how GCP built its Edge AI solutions on Intel technology. Joy Chua, EVP of strategy and development at meldCX, shared how the independent software vendor (ISV) overcame its business challenges and accelerated its IoT (Internet of Things) journey with AI building blocks. She illustrated the process from product ideation to deployment for its customers like Australia Post and Westpac.

Australia Post upgraded its package shipping with meldCX Concept SALi (Smart Automated Lodgement API) self-service kiosks last year. The kiosks use machine learning and computer vision technology to scan and detect each package, thus automating the parcel delivery operations. This led to a 67% reduction in queues and 95% accuracy in recognizing handwritten labels at Australias number one provider of postal services. Concept SALi leverages Intel AI technology such as Intel Core Processors, Intel Movidius, and the Intel OpenVINO Toolkit.

The Intel AI Summit on-demand sessions also featured Demo Showcases with a line-up of specialists from Intel and its partners, namely Databricks, Dell Technologies, Fortanix, Hewlett Packard Enterprise, LAB3, Lenovo Global Technology, and L&T Technology Services. All sessions from Day 2 of the summit are also available such as the keynote address by Pradeep K Dubey, senior fellow and director of parallel computing lab at Intel, and his special guests. They talked about how Intel technology, software, and innovations help clear the path forward in AI by making it more scalable, productive, performant, and intelligent.

The Intel AI Summit 2021 on-demand can be accessed here. Begin something extraordinary with AI today.

Read this article:

Expand your horizon with the Intel AI Summit 2021 on-demand - Tech Wire Asia

When University of Vermont researcher Josh Bongard and his colleagues in Massachusetts began working on a project to build robots using artificial intelligence and frog stem cells, one of their first challenges was communicating in a common vocabulary: Were they building biological robots, engineered organisms or something entirely new?

Normally, scientists know whether they're working with a living creature or a machine. But since their collaboration began in the summer of 2018, Bongard and his fellow researchers including computer scientists, roboticists, biologists, and a philosopher and biomedical ethicist have grown accustomed to the ambiguities inherent in their work.

They've created what they call "reconfigurable organisms" in scientific literature and "xenobots" in mainstream press. The latter is a cheeky portmanteau of "robot"; Xenopus laevis, the frog species they use; and the prefix "xeno," meaning "alien" or "strange." Be they machines or organisms, xenobots move independently, are self-powered and self-replicate before they run out of power or "die."

"Xenobots do a good job of cutting away at those distinctions," Bongard said in an interview. "For me, part of the fun of science is that new discoveries challenge our preconceived notions."

Bongard, 47, is a professor of computer science and director of the Morphology, Evolution & Cognition Laboratory at UVM. A celebrity of sorts in his field, Bongard is collaborating on the xenobots project with researchers at Tufts and Harvard universities. They include Michael Levin, a biology professor and director of Tuft's Allen Discovery Center.

Their goal has been to explore how AI can be used to design and build robots out of cells rather than conventional materials such as metal, plastics and computer components. These novel creations could be the prototypes for an entirely new class of autonomous organic machines. One day, xenobots could be used in cutting-edge medical therapies, as environmental cleanup tools and as tools to decode the mysteries of how cells function.

Xenobots also raise an important philosophical question: What is life?

In January 2020, Bongard and Levin's team published its first paper on how its AI designed a xenobot that could walk. In March 2021, it published another about making a xenobot that could swim.

The team's latest paper, "Kinematic self-replication in reconfigurable organisms," published last month in the Proceedings of the National Academy of Sciences of the United States of America, attracted considerable attention, both within the scientific community and from the mainstream press with good reason. The paper describes using a supercomputer to design self-replicating robots out of living cells, a feat that immediately brings to mind scenarios like those in The Terminator or Jurassic Park. In fact, in the latter movie, it was scientists' use of frog DNA that let the prehistoric genie out of the bottle.

Notwithstanding Hollywood's scary-monster scenarios, xenobots are tiny, simple, innocuous organisms, Bongard said, as unlikely to escape from the laboratory and take over the world as a skin biopsy. Swallowing one or several would cause you no harm whatsoever.

What exactly are xenobots, and how are they made? As Bongard explained, researchers take about 5,000 genetically unmodified frog stem cells, then rearrange them into a new pattern that would never arise in nature.

One of the first surprises for the researchers, Bongard said, was the discovery that, once they rearranged frog cells, not only would the new cell grouping not die or revert to its natural shape, it would also retain its new form.

"We all learned in high school biology that frog DNA encodes frogs and human DNA encodes humans. It seems that's not the case," Bongard said. "Whatever genes are doing, it's more complicated than that."

In these experiments, the frog DNA remained unchanged. What defined the xenobot was not its DNA but how its cells were configured. "In xenobots," Bongard explained, "the shape ... dictates what a xenobot is going to do. It dictates whether it's going to replicate, how it replicates [and] what it's going to look like."

To find a shape that would create more xenobots, the team used a form of AI called an evolutionary algorithm, which was programmed into DeepGreen, UVM's Colchester-based supercomputer. It created a virtual environment, akin to a video game, using virtual cells it assembled at random. It then put that virtual xenobot into a virtual petri dish and watched what happened.

If the virtual xenobot couldn't replicate itself, the AI discarded that design and moved on to a new one. If it created a virtual xenobot that could replicate itself somewhat, the AI continually revised and tested the design to improve it.

After testing numerous designs at lightning speed, the AI finally created a virtual xenobot that was adept at self-replication. This evolutionary process, which in nature would take millennia, took DeepGreen about a month, Bongard said.

Next, Doug Blackiston, a senior scientist at Tuft's Allen Discovery Center and Harvard's Wyss Institute for Biologically Inspired Engineering, meticulously assembled actual xenobots by hand under a microscope using frog cells. The process took about four hours per xenobot.

Blackiston then put them into a petri dish and watched what they did. At slightly less than a millimeter in diameter, each xenobot was visible to the naked eye and looked like a poppy seed that moved around the petri dish using its cilia, hairlike structures on its exterior.

As the xenobot moved, it accumulated loose cells in its environment, like a broom collecting dirt. A video posted on the Proceedings of the National Academy of Sciences website described the process thus: "A swarm of frog-cell parents push frog cells into piles that mature into self-moving 'children.'"

Because frog cells contain a small amount of yolk that powers them, each xenobot lasted about 10 days, after which the yolk was depleted and the xenobot stopped functioning.

But as Bongard pointed out, the second-generation xenobots are neither clones of the original xenobot nor its descendants in the traditional biological sense. While each cell contains a set of chromosomes and DNA, when a xenobot produces offspring, it doesn't impart its genetic material onto the next generation. (This is one reason the team uses the term "replicate" rather than "reproduce.") In fact, the DNA between a "parent" and "child" xenobot may be the same or different.

The project draws ideas from various scientific disciplines. Blackiston spent much of his career researching how life-forms morph from one form to another, such as how tadpoles become frogs and how certain animals regenerate limbs and organs. A butterfly researcher, he spent years studying how memory is carried over from caterpillar to butterfly.

Blackiston, who had no prior experience in robotics or computer science before joining the team, said that the overwhelming response to the work has been "positive and huge." Why?

"It's exciting. It's scary. It's a little bit creepy but also cool," he said, "so everyone sees something in this project that they like."

Perhaps not everyone. Blackiston said his work has "enraged" some fellow developmental biologists, in part because of the terminology the team uses, such as describing xenobots as "biological robots," or "biobots."

"I've never heard a roboticist bat an eye when you call it that," he said. "They'll say, 'Who cares what the material is? I don't care if it's made out of wood or metal or cells. If it's something you design, it's programmable and it moves, it's a robot.'"

Even the word "organism" is problematic, he said, because biologists themselves can't agree on a definition. Some argue that an organism must have certain signatures of life, such as growth, metabolism and the ability to reproduce. But that definition immediately raises red flags, he said. A mule, which is a cross between a horse and a donkey, is sterile but still considered an organism.

Xenobot research also raises metaphysical questions, such as whether the team has created a new life-form and, if so, what ethical norms it should follow.

Jeantine Lunshof is a philosopher and ethicist at Harvard's Wyss Institute and a member of the research team. Though Lunshof is neither a biologist nor a roboticist, "I would use the term 'new life-forms,'" she said.

Lunshof poses ethical reality checks for the researchers, asking philosophical questions that other team members might not necessarily pose, such as: What are the larger implications of this research? What are the potential risks and harms compared to the potential benefits?

As a bioethicist, Lunshof said, she was acutely sensitive to the fact that this research became public in the midst of a pandemic, especially given speculation that the SARS-CoV-2 virus had escaped from a government laboratory. Early on, she inquired about biosafety and potential hazards should this material leave the lab. The team reassured her that it works with cells gathered at no harm to the frogs, which regularly shed these kinds of cells into the environment.

Nevertheless, some of Lunshof's colleagues believe that her involvement in the project puts her personal and professional reputation at risk. Social scientists have been particularly critical, she said, accusing her of "being in the wrong camp" and giving the research her ethical seal of approval.

"There is a very common misunderstanding that ethics is a justification process, that I give research legitimacy by not condemning it, which is completely wrong," she said. "As scientists and ethicists, we need to earn the trust of the public ... and we need to be good stewards of that trust."

To that end, if Lunshof were to see scientific decisions or practices that she found ethically unacceptable, she could bring them to the attention of the universities' institutional review boards, which must approve and oversee all research involving living creatures.

Lunshof also considers what society might forgo by not pursuing this kind of research.

Possible applications for xenobots are diverse, Bongard said. They could include cleaning up radioactive and other toxic contamination and removing microplastics from the oceans. His colleagues at Tufts and Harvard cited the long-term potential for developing cancer treatments, regenerative therapies for regrowing damaged organs and diseased tissues, and even antiaging and life-extending technologies. There's also interest in using xenobot technology to grow meat in laboratories rather than on factory farms, which could offer both environmental and ethical benefits.

Blackiston is intrigued by the potential to design xenobots for environmental use. Currently, if conventional robots break down or lose function, they pollute their environment with batteries, heavy metals and other debris.

By contrast, xenobots could detect pollutants in waterways and study the root systems in hydroponic growing operations. They might coat the exterior of a decaying bridge and reinforce its structure, then biodegrade once the task is complete. Unlike drones or conventional robots, xenobots could perform their work without direct human or computer intervention.

For now, scientists are using xenobots solely for basic research. Bongard likened them to microscopes that eventually may help scientists better understand how cells communicate and what causes them to malfunction, as in cancer cells.

"As scientists," he said, "we don't have a good handle on the language of cells, what they say to one another and the conditions under which they change their tune."

Regarding the philosophical question of whether xenobots have crossed the threshold into a new life-form, "I'm not a biologist, so I'm not going to throw my hat into that ring," Bongard said. "Whatever xenobots are, they're putting more pressure onto our assumptions of what life-forms are."

View post:

A UVM Researcher Uses Artificial Intelligence and Frog Cells to Create Self-Replicating 'Xenobots' - Seven Days

When you work in the digital sphere, it is easy to become disconnected. A year ago, clinical professor and former Wall Street data analyst Frank Parisi, alongside other Pace faculty, conceptualized a space where individuals with an interest in data science and machine learning could connect. We wanted to make a central repository for all kinds of data, where we have the computational power to do interesting things, work together and collaborate across the University and, in the long-term, with outside partners for research, said Parisi.

Now, the space has been set up, the machines moved in, and Paces Computational Intelligence Lab is open for business.

Computational intelligence refers to the machine learning and data analysis abilities of a computerits what allows us to collect data, speak to Siri, and play the newest video game. Jon Lee, a clinical professor, was one of the architects of the lab and he believes it will be unique in what it offers.

There are other Pace hubs that exist for design, digital forensics, and cybersecurity, he says. This will be a proper space for computational intelligence specifically, from Artificial Intelligence, pattern recognition and machine learning.

This will be a proper space for computational intelligence specifically, from Artificial Intelligence, pattern recognition and machine learning.

Parisi notes that not only can the lab be a tutoring resource for those learning programming languages like Python and R, but it can also elevate the quality of our data scientist professionals.

My particular favorite aspect is the development of conceptual workshops, where we cover things like probability theory, how to build models, and statistical computing, he explains.

Having a physical lab with quality equipment also means that students and faculty engaged in deeper analysis will not have to rely on remote Google servers. Furthermore, as Lee notes, the lab will serve as a way to get Seidenberg students engaged, active, and doing amazing things on campus, especially those who are feeling disconnected after COVID-19.

All in all, the Computational Intelligence Lab will empower Pace faculty and students to do what they do bestconnect, innovate, and build great things for both today and tomorrow.

Want to see the Computational Intelligence Lab or learn how you can get involved? Reach out to co-directors Frank Parisi atfparisi@pace.edu or Yegin Genc at ygenc@pace.edu.

See the article here:

The Science of Machine Learning - Pace News

Our Startups series looks at companies and founders who are innovating in the fields of athlete performance, fan engagement, team/league operations and other high-impact areas in sports. If youd like to be considered for this series, tell us about your mission.

* * * * *

Worlds shortest elevator pitch: We're the only platform that allows users to use artificial intelligence to create their very own custom predictive models, as well as come up with player production and fantasy and prop stats.

Company: Pine Sports

Location: Ridgewood, New Jersey

Year founded: 2020

Website/App: https://www.pine-sports.com/Funding round to date: Were self-funded.

Who are your investors? Were self-funded.

Are you looking for more investment? I would say were potentially looking for smart money, sort of a strategic partner. We've already had a lot of great conversations with VCs who already love what we're doing and want to be in the space. They see the product as a differentiator, with the artificial intelligence and the custom modeling. If it makes sense, that's definitely a path well consider.

Tell us about yourself, co-founder Mike Yam: My co-founder, Vijay Dewan, and I are high school friends. Both of us

Co-founder and NFL Network broadcaster Mike Yam

Who are your co-founders/partners? Vijay was commuting to work two hours a day and decided to go back to his roots in computer programming. COVID was one particular thing he was interested in and used programming tools that are in Python to try to understand whether this was going to be nothing or something important. As the tools said, it was something major. During COVID, over time, he learned to translate the programming skills he learned analyzing COVID data to other data. He called me and said: Is there AI in sports? Are people using AI? Is there AI modeling that any user can do? The answer was really no. We couldn't find anything. He built the programming. I know the sports. And we teamed up, and it's been great since.

How does your product/service work? From the product and service standpoint, there are three different verticals for the website. There's what I think is sort of the gold standard in terms of predictive modeling, which we call Predict. Its super easy to use. You pick a sport, select how many seasons you want to go back with our artificial intelligence, you pick a rolling average and then you get the opportunity to select from over 200 stats to determine what decides the outcome of a game. The artificial intelligence essentially does the rest. It gives you a competence percentage based on actual games that happened, and that competence percentage is for upcoming games. The other vertical would be Explore, which is so insanely easy to use. Basically has a drop-down menu after you type in a player's name. Within three clicks, our machine learning tool basically does the rest. You type in the categorywhether it's points, rebounds or assists or any other stats and how many games you want it to go backand it'll give you a prop line if you want to use that for prop bets or a fantasy projection really is the outcome. What's really a cool differentiator on that particular product is the normalized number, which takes into consideration an opponents defense. That's really cool. Weve got a robot that basically tells you it likes the over, the under, and then gives you that projection. Having tools like that is insanely helpful for fantasy players. The third vertical is the social side of it. A lot of our users right now are able to take a lot of the learning insights they're using from playing around with the data, and then write articles surrounding the things they're learning. It's really cool to see the community sort of exchange a lot of different ideas.

What problem is your company solving? The problem is that users right now don't have the ability to create custom, predictive modeling using artificial intelligence anywhere. The ones who do know how to do it have to have a background in coding. Pine is no-code AI, and anyone essentially within a few clicks can figure out how to use these predictive models or basically create the predictive models. Other sites are kind of using Excel spreadsheets, but there's no real math and there's certainly no real analysis on actual games that have happened. The way we like to look at it is we're letting the artificial intelligence do what the brain can't do. I worked with so many coaches and players in my career and they can reference specific plays like Sean McVay. What AI does, though, is take every single play from every single game and analyze it. There's literally not a human being that is going to be able to do what AI can. As a sports fan, you have the ability to tell the perfect memory of AI what's important, and that leads you to an outcome or potentially an outcome.

What does your product cost and who is your target customer? Free right now for users. We were invite-only, but the site is now open for anyone to use the tools for free. For our user base, we've been in beta and they made more than 13,500 custom model projections. Doing a lot of different iterations and getting feedback from a lot of users.

How are you marketing your product? The marketing comes already on the platform with our writers who are using the tools to write about their learned insights. Anyone who goes to pine-sports.com has the ability to read the articles our users are generating. There's Twitter amplification, as well. So, anyone who's posting articles, that immediately goes to our Twitter feed, not to mention our Discord feed where there's a few hundred users that are on there, just swapping different ideas, watching the games together, and having some fun. Our marketing is our users. Hopefully, our goal is that they love the platform and tell their friends. Thats our current marketing, and it's been going well.

You pick a sport, select how many seasons you want to go back with our artificial intelligence, you pick a rolling average and then you get the opportunity to select from over 200 stats to determine what decides the outcome of a game. The artificial intelligence essentially does the rest.

How do you scale, and what is your targeted level of growth? From a scalability standpoint, its once we open up and have the ability to kind of hammer home our user base and let them use the tools. The sky's the limit, to be honest with you, because these tools don't exist. You have other sites that are charging an ungodly amount of money for inferior tools. To me, the scalability is just going to come as soon as we open it up. For us, it's about community. We have a very strong community, both on the platform and on Discord. Our targeted level of growth is to scale that community to a point where it's large. The great thing about Pine is that it's language agnostic and data agnostic. What I mean by that is people from all different countries can write in whatever language they want. We're getting more data on the platform month after month. Weve got soccer and cricket, we've got all sorts of data on the platform. From our perspective, it's getting users from across the globe involved in the community. We don't have specific numbers. It's about scaling smart, which means making sure the community's sort of ethos is the same as it has been since the beginning. Which is that it's people who are helping each other, people who are trying to make better bets and people who are trying to grow together. We have a large target market. In terms of growth, if you think of other people in the space, or just generally the social space, I looked at Redditwhich grew sort of linearly for a long time because what they did was they made a product that was good and people found it. I don't think you need that extreme exponential growth; I think you need to focus on community first.

Who are your competitors, and what makes you different? I actually don't think there's competitors for what we're doing because no one out there is currently putting artificial intelligence in the hands of normal sports fans. The closest thing would be Action Labs. But once again, no AI, no machine learning, no math, no analysis of actual games.

Whats the unfair advantage that separates your company? The easiest way to describe this is imagine having a friend who has a perfect memory. That's artificial intelligence. You get to tell it what's important about sports, and then the AI does the rest of the work. The unfair advantage for us is having a friend who literally has a perfect memory and never forgets anything.

What milestone have you recently hit or will soon hit? One of the biggest milestones is opening up the site for everyone and growing the community and seeing people engage in the data, messing around with building models and having fun with it. That's the biggest next milestone, opening this thing up and going outside of beta. What is really important is giving people the ability to use AI for themselves. Right now, AI is being used against you every day of the week, whether it's Amazon predicting what you're going to buy next and shoving that in your face or whether it's Netflix telling you what show you're going to want to watch next or whether it's Facebook telling you what ads you're going to want to watch. Its literally the same AI model. They're being used to take your attention away. What we're trying to do is show you what AI is and give you the ability to use it to help yourself. Some recent notable trends: We had more than 8,000 prop projections in December, and Pine users are averaging more than five minutes on the site. For context, according to similarweb.com, thats more than Yahoo Sports, CBS Sports, NBA.com and NBCSports.

In what ways have you adapted to the COVID-19 pandemic? Pine Sports actually came because of COVID. Vijay was aggregating data in local jurisdictions in New Jersey, realizing there's a hole in the use of AI for sports fans and empowering those users. In a lot of ways, Pine doesn't exist without COVID.

Beyond the pandemic, what obstacles has your company had to overcome? Time is one of them. It feels like there's never enough time in the day for the two of us. Relationships, just sort of leaning on the people we've worked with in the past. A lot of times that's been an advantage for us, but getting it in front of the right people, the decision makers. The people we have talked to have really loved the product. Overall, we're in a really good place. The track line of where we're trending is certainly an upward trajectory. Were taking something that's really complicated and trying to simplify it. In the sports space, unfortunately, a lot of people take really simple data analytics and dress it up to make it sound complicated. When we first opened up to beta users, it was a really complicated product and was really hard to use. It was chiseling and simplifying until we got something that still held that core of being really good at what it does, but also easy to use. That was our largest challenge to date. Some people might call that stickiness, but it's getting people who are in the sports base to understand the value of what we're providing.

What are the values that are core to your brand? The best way to describe that is looking at Pine Sports as a way to empower sports fans by fulfilling two different goals. One would be to put the power of artificial intelligence in their hands. The other side of it is to create a community. Vijay and I talk about this all the time and just how cool it is to see people using the tools, writing about the tools and then writing about how much they love the tools and how it's helping them just become a smarter fan. Making every sports fan smarter is one of the main objectives of Pine.

What does success ultimately look like for your company? We want every sports fan to use our tools. We live in sort of a day and age where people are scared of the math, scared of the data. Vijay and I have been able to pare down this product so that it is, at its core, super easy to use but also still really powerful. Once again, to see people engage with Pine and to see their reactions when we run a demo for them and that light bulb goes off, is a really cool thing and in a lot of ways gratifying for the two of us.

What should investors or customers know about youthe person, your life experiencesthat shows they can believe in you? First and foremost, Vijay and I were high school friends who were really fortunate to have a lot of advantages because of the sacrifices of our parents and our families. I would classify us both as grinders who really work hard on our craft. For me, specifically, as a storyteller and someone who's been around sports my entire career, I want to see a lot of our users use the data and have fun with it, but I also want people to get a sense of control that just didn't exist before Pine. Vijay spent a good portion of his career as an attorney, but also as a federal prosecutor putting bad people away. At the end of the day, at our core, both of us want to do the right thing. Right now, from a sports perspective, the right thing is putting the power of artificial intelligence into the hands of every sports fan and making it easy to use.

How has the rise of legalized gambling affected Pine Sports? Artificial Intelligence is about making predictions about the future. Betting is about making predictions about the future. Fantasy is too, but the more you are giving people the opportunity to make predictions about the future, the more useful our tool is. Betting is clearly very important. Everything we do is making predictions about the future, which is why AI is so important, because it allows us to do that in a way that's really smart. Some people might see a line for Tom Brady that's 310 yards and they might have no idea how many yards he's passed for previously. They may not even really think about his opponents, but the AI will do all of that for you and give you a projection about the future and do really complicated math to do that. So, yes, betting is important. Betting is about making predictions about the future and putting your money down. AI, at its core, helps you make really smart, informed decisions about the future. Thats where we shine.

Do you have a favorite quote about leadership? Vijay and I have similar backgrounds and a similar thought process. Both of us feel it's important to listen first. Thats something both of us do when we're doing demos or even bouncing back and forth ideas. But the biggest thing we try to ask ourselves on a regular basis is: What's your impact? It goes back to the predictive nature and the power of artificial intelligence. The impact we're trying to have on sports fans and those communities of users is the ability to give you something that is very difficult to master, but making it really simple so that anyone can use it. We're trying to build a community of people who want to be smarter and give them those tools.

Question? Comment? Story idea? Let us know at [emailprotected]

See more here:

Pine Sports Is the Intersection Between Artificial Intelligence and Game Prediction, as in the Photographic Memory You Always Wished You Had -...

13.01.2022From Tobias Gundermann

The initial question in the title of this article might seem a bit odd at a first glance as it is probably rarely the case that the terminologies "game theory" and "metal casting" are used together in one sentence. So, how can both be brought together so that one serves as a baseline to optimize the other?

Related Companies

The answer to that question lies in data science, machine learning and the increasingly arising field of explainable artificial intelligence (XAI). If you are curious how it works, then take a few minutes and read on!

Let's first clear the dust a bit and have a short look at what "game theory" actually is.

"The branch of mathematics concerned with the analysis of strategies for dealing with competitive situations where the outcome of a participant's choice of action depends critically on the actions of other participants." [1]

Well, the definition from Oxford Dictionaries doesn't seem to help much for our understanding, so let's visualize it with the following example:

Consider you have a football game with 11 players in each team. After thrilling 90 minutes of high-class football, both teams split up 2:1. The principles of game theory could now be used to find out how much each of the players contributed to the end-result. (Basically, how valuable were the individual players for their team.)

There are various approaches to calculate the contributions of the individual players. The specific approach we will have a look at in the following lines of this article are Shapley Values (invented by Lloyd Shapley in 1951). Shapley values are used to calculate the average marginal contribution of each individual player - basically the average contribution of each player across all possible orders in which they can be brought into the match. [2]

Further, let's take the use-case of quality prediction in the casting of aluminum wheels with the low-pressure die casting process. In this, molten and degassed aluminum is stored in the holding furnace of a low-pressure casting machine. The casting process takes place in 3 steps:

a) The pressurization in which pressure is applied to the holding furnace which causes the molten aluminum to rise through the riser tube into the mold

b) Filling up the mold during which the pressure is increased to fill the mold in a controlled and uniform way

c) The solidification in which a high pressure is applied to prevent shrinkage in the casted wheel.

The problem faced by our customer in this case were microporosities, blow holes and shrinkage which lead to an increased cost & remelting, excessive emissions and reduced OEE.

To enable operators, shift supervisors, process engineers and foundry managers to proactively take corrective actions in order to avoid scrap, a machine learning (ML) based model can be developed to predict the quality of the casting during the LPDC process. This model then takes real-time data collected within the production process (e.g. temperatures, air cooling rates, pressures etc.) to continuously monitor the casting process in near real-time.

The predictive quality model helps to detect quality deviations as early as possible and enable the engineers to make adjustments and to eliminate the root-causes of the quality deviations. But what if the root-causes and measures to be taken are unknown?

That's exactly where both of the terminologies "game theory" and "process optimization" come together and the connection of these is explainable artificial intelligence (XAI).

Explainable Artificial Intelligence (XAI) describes a field of research for the development, advancement and improvement of methods to make predictions or classifications of ML-based models interpretable and/or functionally comprehensible.

Given a ML-based model which predicts the product quality based on the collected process parameters, technologies such as SHAP (SHapley Additive exPlanations; based on the above-mentioned Shapley-Values) can be used to determine the most influential process parameters (players) with regard to their effect (contribution) on the product quality (result of the match). This is achieved in the form of so-called feature importance scores which assign a value to each of the input parameters of the model depending on their effect of the output of the model. [4] A visualization of SHAP-values can be seen in the following extract from the TVARIT Industrial AI Software (TiA) for the quality prediction in the aluminum casting process (please note that the concrete parameter names and values have been changed due to data privacy reasons):

Extract from TiA (TVARIT Industrial AI Software) for quality prediction in the aluminum casting process; Feature Importance.

(Source: TVARIT GmbH)

Extract from TiA (TVARIT Industrial AI Software) for quality prediction in the aluminum casting process; SHAP Values.

(Source: TVARIT GmbH)

With the help of SHAP-Values in the form of feature-importance scores, manufacturing engineers receive information on the most influential parameters for achieving the target product quality based on the collected data from their production processes.

Now given that quality deviations of castings are known in real-time and game theory helps to understand the root-causes of casting errors, the remaining question is still: How do the casting set-points need to be adjusted dynamically to avoid casting errors?

In this case, we have to go one step further than conventional game theory but modern AI technology also provides a solution here: So-called prescriptive dynamic recipes. These give dynamic recommendations for optimal casting set-points.

The methodologies used here are advanced clustering methods that determine how the various influencing factors (data such as set-points, pressure and temperature curves in the casting machine and ambient conditions in the foundry etc.) play together to create a good (or rejected) casting.

If you have difficulties in following the last sentence, dont worry let's break it down step-by-step:

Clustering: "Cluster analysis is the name given to a set of techniques which ask whether data can be grouped into categories on the basis of their similarities or differences [4].

This time, the definition of Andrew M. McIntosh gives us pretty concrete hints on how this might work put into manufacturing practice: The clustering is applied to group castings by their similarity. The metrics used to measure the similarity here are the influencing factors (data such as set-points, sensor values etc.) for that particular casting (or that particular batch).

Prescriptive analytics then identify which of these groups (called clusters) have the best quality results which then can be used to identify the optimal values for the influencing factors. This can be seen below in the Principal Component Analysis (basically a 2-dimensional representation of the influencing factors for the sake of visualization). Here, the green group has been identified as the optimal group (cluster) of influencing factors as the castings (the red crosses) that lie in that area have the best quality results. The gradient of the crosses indicates the timing (the darker crosses are the most recent castings).

Principle component analysis.

(Source: Tvarit GmbH)

Okay now that we got pretty technical and understand that prescriptive analytics define the optimal values of influence factors: How can this be used to reduce scrap?

Put in practice, the knowledge of the optimal values of influence factors can be used to define how the set-points need to be adjusted so that the casting process lies within these optimal cluster. These recommendations are then called prescriptive dynamic recipes (shown in the screenshots below).

Prescriptive dynamic recipes.

(Source: Tvarit GmbH)

To get back to the initial question of this article: Artificial intelligence and game theory help to optimize casting processes in the following way:

[1] Curtis, S. (2013). The Law of Shipbuilding Contracts (4th ed.). Informa Law from Routledge. Definition of game theory. (2018). (Oxford University Press) Retrieved May 2018, from Oxford Dictionaries: https://en.oxforddictionaries.com/definition/us/game_theory [2] Shapley, Lloyd S., und Alvin E. Roth, Hrsg. The Shapley Value: Essays in Honor of Lloyd S. Shapley. Cambridge [Cambridgeshire] ; New York: Cambridge University Press, 1988. [3] Lundberg, Scott M., Gabriel Erion, Hugh Chen, Alex DeGrave, Jordan M. Prutkin, Bala Nair, Ronit Katz, Jonathan Himmelfarb, Nisha Bansal, und Su-In Lee. Explainable AI for Trees: From Local Explanations to Global Understanding. arXiv:1905.04610 [cs, stat], 11. Mai 2019. http://arxiv.org/abs/1905.04610.%5B4%5D Andrew M. McIntosh, ... Stephen M. Lawrie, in Companion to Psychiatric Studies (Eighth Edition), 2010

(ID:47915437)

Read the original here:

How Artificial Intelligence and Game Theory Can Help to Reduce Scrap in Metal Casting - SPOTLIGHTMETAL