The latest release of Hazelcast’s Jet event stream processing engine for AI and ML deployments of mission-critical applications adds new app development features designed to simplify the integration of an event-driven architecture into brownfield deployments to gain new functionality around real-time and in-memory processing.

“Stream processing” is about processing data in motion–users take action on data at the time it is created. It typically involves multiple tasks performed on an incoming series of data, and it can be performed serially, in parallel, or both. The stream processing pipeline starts with the generation of the data, followed by the processing of the data, and finally the delivery of the data to its destination.

With the release of Hazelcast 4.2, the in-memory computing platform maker
is making it possible to add Jet’s extensibility to existing applications through real-time caching and stateful microservices.

The list of updates and enhancements in this release includes new support for streaming integration with MySQL and PostgreSQL databases using a unified high-level API. Traditional RDBMS-based applications require hundreds or thousands of lines of code to add functionality, as well as significant testing. With the new API in Jet 4.1, the integration becomes more of a declarative task to reduce the custom error-prone code. 

Additionally, Hazelcast Jet makes the database available as a stream. It deals with connectivity, object mapping and unifies the event handling across database vendors. The series of database updates form an event stream on which developers can more easily add microservices without impacting existing applications. This simplification lets developers focus on adding new business logic in high-performance applications rather than managing complex and error-prone integrations. 

The transactional data from MySQL or PostgreSQL can be augmented and enriched with other datasets from Hadoop, Amazon S3, Google Cloud Storage, Azure Data Lake, and others, the company says, and served through thousands of concurrent low-latency queries and fine-grained, key-based access.

“With these enhancements, enterprises can take advantage of in-memory speeds to accelerate analytical queries to scale your architecture by offloading certain workloads from your transactional database into an in-memory store,” the company said in a statement.

Hazelcast provided support in Jet earlier this year for change data capture (CDC) via the open-source Debezium project. In version 4.2 the CDC integration has been optimized to reduce the manual coding required to utilize this capability, the company says.

Also, over the last year, the library of connectors for Hazelcast Jet has been expanded to include Apache Beam, Confluent, MongoDB, JDBC, Apache Cassandra, and others. Version 4.2 includes connectors for Elasticsearch and Apache Pulsar. By connecting Jet to Elasticsearch, the company says, enterprises can rapidly enrich large data sets, including those from relational databases, and transform them into formats suitable for indexing and search-based analysis by Elasticsearch.

The post Hazelcast Simplifies Application Modernization with Event Driven Architectures appeared first on Artificial Intelligence.

Many technology commentators have talked about data as the “new oil”. In the wake of fallout from the current epidemic, data might also come to be viewed as a utility – like electricity, water and broadband; a vital resource essential to shaping, supporting, securing and optimising, all life.

Through the rapid growth of IoT deployments, organisations are capturing more data than ever before. But there are still a number of questions around the use of data that need clarity. What is the value in the data? How can it be made available and used effectively to benefit all stakeholders – councils, citizens and businesses? How can it be monetised, if at all? In light of GDPR and the post-Covid-19 landscape, this is a highly topical area.

Mining the data lakes

The use of Internet of Things adds additional data streams from new devices, environments and processes that organisations have not previously been able to connect to or explore. That data can be treated as a standalone asset in the proof of concept phase but the ultimate objective is to combine it with other data to create “data lakes” that can be dissected and utilised. Yet, while many organisations and entities such as local authorities are collecting this data at pace, they are yet to mine that data effectively, examining it to generate new information and maximise its true value.

As more IoT projects are deployed and mature, vast volumes of data points, potentially in the billions, will be added to existing repositories. Data is amassing at a significant rate and this data will be valuable and useful to organisations and citizens. Some data may be designed to be collected and actively shared with third parties to aid decision-making for communities. An example of this is in smart city infrastructure – town planners, construction companies, utilities, public enquiry, service providers, among others, may want access to particular data for planning, building or to improve service offerings.

On the other hand, there could be conflicts in sharing data widely as IoT rolls out. For example, there may be an increasing number of data points measuring climate and pollution but that could reveal data on a city’s rising CO2 levels, which results in the city being held up to scrutiny and fined. Measuring air quality levels has now become a top priority as the linkage between exposure to pollution and susceptibility to the effects of the new “mega-viruses” has been established. The balance of data insight versus the potential societal and commercial impacts could become a complex issue to manage.

But first, the data has to be mined and structured in a way that it can be used, so there must be a system in place around who has permission to use that data. Is the data available for free or on a commercial basis? Or is the data confidential and strictly for internal use only? Recent developments with the opening up of track and trace data during the Covid-19 epidemic has overridden GDPR concerns – will this become normal practice to ensure health for our city populations?

There is a lot of education going on in the public sector about how to use data, hiring data scientists versus using third-party resources, how to assemble the tools to make that data mining effective – getting the right expertise on board is crucial.

Asking the right questions and defining the queries for the datasets is also important. There is little value in collecting increasing quantities of data if you don’t ask the right questions to gain the most insight from it. Benefits of using an Internet of Things platform to process data include powerful analytics and visualisations that deliver trend analysis and even return on investment (ROI). These tools and visualisations can be customised and personalised to individual departments and stakeholders.

Securing data flows

When it comes to implementing a network infrastructure for IoT, security is a priority. For many IoT deployments, a combination of public and private data sources will need to be used but how will the data flows be securely managed? Some of the data will be related to mission-critical infrastructure and operations (such as traffic flows, energy and water infrastructure). Public domain or ‘open’ data published by central government, local authorities and public bodies includes environment (weather, flooding, air quality), transport (airports, roads, electric vehicles, parking, buses), towns and cities (housing, urban planning, leisure, waste and energy), education, health (hospitals, medicine performance) and others.

This is where middleware is required that can effectively segment the data network and prioritise appropriate traffic, enabling data to be routed correctly and efficiently to the right repositories and analytics engines. The more data that is amassed, the more the challenge will increase. A comprehensive data strategy is required that not only encompasses the variety of sources of data but also the routes and collection methods whereby the data is brought in.

Addressing cultural concerns

Culture continues to be a significant barrier to adoption when it comes to IoT deployments. The idea of 24/7 monitoring or corporations accessing our personal data makes people nervous. This is primarily due to the unknown facts about what happens to that data and the question of who ultimately owns it? Without insight into this, and how data can be used in a positive way, the automatic response to data being recorded and used is an initial degree of scepticism.

The good news is that Internet of Things data is collected and delivered in an anonymised, secure format. This data is decrypted, collated, analysed and integrated with other datasets as part of a tracked process. The data composite is used primarily to provide an overall picture and to track trends and changes, as compared to personal data collected by an eCommerce site. Privacy by design is an inbuilt feature of an IoT system solution, mitigating privacy concerns.

As IoT continues to become a part of our everyday lives, we will likely see an evolution in this setup, perhaps even at a granular level where permission is granted for particular data to be used for specific purposes. Education is therefore crucial to overcoming cultural concerns, to communicate the benefits of IoT data use. For example, within smart buildings, how IoT can have a positive impact on elements such as energy usage, not just in terms of lowering bills but in turn reducing the impact on the environment and many more subsequent benefits. Another example is the use of air quality data to allow personalised views and planning routes into school and work.

There is currently a lack of legislative framework around how data should be shared and it is something the industry has been crying out for. In the meantime, GDPR is the only universal mechanism around data sharing and data processing, but it requires augmentation and localisation. We’re now in the early adopter phase of data sharing, where organisations are looking to mirror best practice and consultancy, replicating what others are doing in terms of data mining and management in order to identify the best strategies, but are also trying new methods and innovations to see what the effects are.

The evolution around the use of IoT data is one that is continuing to develop at pace but can also be challenging for those organisations that are amassing growing volumes of data from initiatives and projects. Combining IoT data with other sources, mining and making it available in ever-more flexible and tailored ways to a variety of stakeholders is a complex task, requiring expertise and teamwork.

Organisations who wish to make effective use of the new data utility should look to create an ecosystem of complementary experts and providers, capable of guiding the data collection, mining and distribution through the complexities, addressing all the necessary hurdles including infrastructure, security and cultural barriers. The potential prize contained in the use of data is great – better, safer, more efficient living for all.

The post IoT data: Maximising its value appeared first on Artificial Intelligence.

The king of awe-inspiring viral videos, Boston Dynamics is continuing its push to commercialize its agile robot creations with a new pilot program at Ford. The car-maker will be leasing two robotic dogs, known as Spot, for its Van Dyke Transmission Plant.

The dogs will be on something of a short leash, guided by handler Paula Wiebelhaus, who uses a controller and will personally monitor the bots in operation. Painted bright yellow, the quadruped robots will each carry five cameras and two hours worth of battery power and will walk the floor capturing plant data and dimensions that will eventually be used to retool the line. In the future, the robots could be used to perform this task autonomously.

“We design and build the plant. After that, over the years, changes are made that rarely get documented,” says Mark Goderis, Ford’s digital engineering manager. “By having the robots scan our facility, we can see what it actually looks like now and build a new engineering model. That digital model is then used when we need to retool the plant for new products.”

Scanning physical spaces is a job well-suited to robots. Other robots — even other dog-shaped robots — are used for similar purposes. Aided by a partnership with Lenovo, UK-based React Robotics has a four-legged robotic helper called DogBot designed specifically for the construction sector, though with potential applications in other industries. Like spot, DogBot is a mobile sensor platform that can autonomously navigate spaces utilizing machine learning algorithms for locomotion, perception, and proprioception. 

According to Ford, the Spot deployments could save the company $300,000 per year, which is the cost of manually scanning the giant facility.

“We used to use a tripod, and we would walk around the facility stopping at different locations, each time standing around for five minutes waiting for the laser to scan,” Goderis recalls. “Scanning one plant could take two weeks. With Fluffy’s help, we are able to do it in half the time.” 

In line with larger sector trends, Boston Dynamics designed Spot to be application agnostic, outfitting it with baseline capabilities to perform multiple duties. As a result, the company is celebrating early deployments in a number of use cases. In one deployment, a construction firm in Canada used a Spot robot to automate the capture of thousands of images weekly on a 500,000 square foot building site, creating an ongoing record of progress and enabling the builders to identify growing problems and inefficiencies early.

Amusingly, Ford’s Spot will, at times, catch a lift atop another robot as it sits on its haunches like a robotic king. That small courier robot, known informally as Scouter, will ferry the dog robots up and down the aisles of the plant to conserve the dog’s batteries. 

The post Ford fetches robot dogs to work in a factory appeared first on Artificial Intelligence.

There’s no doubt that artificial intelligence continues to be swiftly adopted by companies worldwide. In just the last few years, most companies that were evaluating or experimenting with AI are now using it in production deployments. When organizations adopt analytic technologies like AI and machine learning (ML), it naturally prompts them to start asking questions that challenge them to think differently about what they know about their business across departments, from manufacturing, production and logistics, to sales, customer service and IT. An organization’s use of AI and ML tools and techniques – and the various contexts in which it uses them – will change as they gain new knowledge.

O’Reilly’s learning platform is a treasure trove of information about the trends, topics, and issues tech and business leaders need to know to do their jobs and keep their businesses running. We recently analyzed the platform’s user usage to take a closer look at the most popular and most-searched topics in AI and ML. Below are some of the key findings that show where the state of AI and ML is, and where it is headed.

Unrelenting Growth in AI and ML

First and foremost, our analysis found that interest in AI continues to grow. When comparing 2018 to 2019, engagement in AI increased by 58% – far outpacing growth in the much larger machine learning topic, which increased only 5% in 2019. When aggregating all AI and ML topics, this accounts for nearly 5% of all usage activity on the platform. While this is just slightly less than high-level, well-established topics like data engineering (8% of usage activity) and data science (5% of usage activity), interest in these topics grew 50% faster than data science. Data engineering actually decreased about 8% over the same time due to declines in engagement with data management topics.

We also discovered early signs that organizations are experimenting with advanced tools and methods. Of our findings, engagement in unsupervised learning content is probably one of the most interesting. In unsupervised learning, an AI algorithm is trained to look for previously undetected patterns in a data set with no pre-existing labels or classification with minimum human supervision or guidance. In 2018, the usage for unsupervised learning topics grew by 53% and by 172% in 2019.

But what’s driving this growth? While the names of its methods (clustering and association) and its applications (neural networks) are familiar, unsupervised learning isn’t as well understood as its supervised learning counterpart, which serves as the default strategy for ML for most people and most use cases. This surge in unsupervised learning activity is likely driven by a lack of familiarity with its uses, benefits, and requirements by more sophisticated users who are faced with use cases not easily addressed with supervised methods.

Deep Learning Spurs Interest in Other Advanced Techniques

While deep learning cooled slightly in 2019, it still accounted for 22% of all AI and ML usage. We also suspect that its success has helped spur the resurrection of a number of other disused or neglected ideas. The biggest example of this is reinforcement learning. This topic experienced exponential growth, growing over 1,500% since 2017.

Even with engagement rates dropping by 10% in 2019, deep learning itself is one of the most popular ML methods among companies that are evaluating AI, with many companies choosing the technique to support production use cases. It might be that engagement with deep learning topics has plateaued because most people are already actively engaging with the technology, meaning growth could slow down.

Natural language processing is another topic that has showed consistent growth. While its growth rate isn’t huge – it grew by 15% in 2018 and 9% in 2019 – natural language processing accounts for about 12% of all AI and ML usage on our platform. This is around 6x the share of unsupervised learning and 5x the share of reinforcement learning usage, despite the significant growth these two topics have experienced over the last two years.

Not all AI/ML methods are treated equally, however. For example, interest in chatbots seems to be waning, with engagement decreasing by 17% in 2018 and by 34% in 2019. This is likely because chatbots were one of the first application of AI and is probably a reflection of the relative maturity of its application.

The growing engagement in unsupervised learning and reinforcement learning demonstrates that organizations are experimenting with advanced analytics tools and methods. These tools and techniques open up new use cases for businesses to experiment and benefit from, including decision support, interactive games, and real-time retail recommendation engines. We can only imagine that organizations will continue to use AI and ML to solve problems, increase productivity, accelerate processes, and deliver new products and services.

The post Key Trends Framing the State of AI and ML appeared first on Artificial Intelligence.

The study “Connected Everything: Taking the I Out of IoT,” was conducted across 200 enterprise executives in North America and Europe in several vertical industries already deploying IoT. The industries ranged from financial services to health care and hospitality. As per the report, most of the companies are focusing on the concerns and opportunities around the Internet of Things. A significant number of companies are witnessing strong business cases for IoT adoption as part of a broader digital transformation process. The majority of respondents cited improved customer retention, enhanced productivity, and high efficiency as their highest objectives.

Companies are working towards expanding the use of IoT solutions; however, to realize this potential, they need to secure, easy-to-deploy and flexible ways to support IoT integration with their businesses and processes. The report also notes a major growth in the role of expert third-party suppliers that have helped businesses in tapping new opportunities available via a bigger global IoT connectivity.

Half of the executives identified data, network, and device security as the biggest challenges to IoT adoption, the study found. Moreover, half of them also said they prefer putting IoT devices on their own private networks to tackle IoT security concerns. Nearly all respondents said they are either considering or are currently using a private network for their IoT deployments.

Other top challenges for IoT adoption in the enterprise were integration and security concerns, the study found. Nearly 45% of respondents said integration with legacy IT and networks are challenging, while 40% cited integration with business processes as a major challenge. As per the report, if enterprises plan to embrace IoT in its full capabilities, they will face issues including IoT concerns, providing technical solutions, consulting services, and a transparent approach to IoT security.

IoT adoption comes with its challenges for companies; however, experience, network reach, and technology expertise can be instrumental in addressing their concerns. Using private networks to avoid the inherent risks of the public internet is a viable but technically complex approach. Today, enterprises are eager to adopt and realize the economic and productivity benefits but understand they can’t go it alone in their deployments.

The post IoT – Majority of Enterprises See Strong Cases for IoT Adoption appeared first on Artificial Intelligence.

QCon London Sam Newman, author of Building Microservices and Monolith to Microservices, told attendees at the QCon developer conference in London that “microservices should not be the default choice.”

Microservices have become today’s equivalent of “nobody ever got fired for buying IBM”, a common catchphrase back in the ’80s, said Newman. “With half of all the clients I have worked with, I have told them microservices are not for you.”

In case you should doubt him, another session at QCon described how Segment, a data analytics company based in San Francisco, went “to microservices and back again.”

“After years of continuing to add to our microservice architecture we found ourselves in a spot where our developer velocity was quickly declining and we were constantly tripping over our microservice architecture and its complexity,” said the abstract. “Moving to a monolith was the solution that worked for us.”

The implication is not that microservices are no good, but that developers have adopted them too readily, persuaded by the benefits. The idea of microservices is that decomposing an application into small independent services makes it possible to update it more quickly and safely, and to scale more efficiently, whereas applications that are compiled into a single large executable – monoliths – require all-or-nothing deployments that are harder to coordinate.

Explaining the benefits of microservices, Newman said: “There are lots of reasons why we might pick a microservices architecture, but the one I keep coming back to is this property of independent deployability.”

If you want to make a change to some part of an application, such as the part that covers shipping, “I should be able to deploy that shipping service into a production environment without having to change the rest of the system. This helps to ship software more quickly and it helps our teams work in a more autonomous fashion.”

What, then, is the problem? The core issue seems to be that it is hard to do microservices well. Newman described what he called “the worst of the monoliths, the distributed monolith”. This describes a system that has been decomposed into multiple processes, “but for whatever reason, we have to deploy the entire system together as part of a lockstep release. Often this can occur because we’ve got our service barriers wrong. We’re smearing business logic all over these different layers. We’re having to coordinate between multiple teams to get anything done.”

A sign of trouble, said Newman, is that “if you have a full-time release coordination manager, chances are you have a distributed monolith.” Newman sees great value in continuous delivery, where software is automatically built and tested and therefore already ready for release.

The distributed monolith has the deployment complexity of a microservices architecture, but without its benefits. The services lack the “independent deployability” critical to success.

This was a popular session at QCon. Why? A clue came when Newman asked attendees if they have some application or system that is “too big, that you’re trying to make smaller.” Hands went up all over the room.

Modularity

Newman made the point that even applications deployed as a single process, or monoliths, can be modular in their design, with different teams working on each module. It is nothing new – it’s an idea “from the early 1970s around structured programming”, he said. This gives “a degree of independent working” and is “a highly underrated option”. A potential problem with modular applications is that “people tend to be not good at defining module boundaries or having discipline about how module boundaries are formed” – the result being “they descend into a ball of mud”. Overcome this, though, and many organisations “would be better off” with a modular monolith than with a microservices architecture.

It is worth noting too that if you have a well-designed modular application, decomposing it into microservices is relatively easy. You will be able to reuse a lot of the code because the concept of isolated sections of code dealing with different features is already in place. The “units of decomposition,” said Newman, are the “domain concepts, such as invoicing, notifications and order management.”

“Monolith has become a replacement for the term we used to use, which was legacy. It’s a problem because some people are starting to see any monolith as something to be removed. That’s deeply inappropriate. The monolith is not the enemy. This is not the problem that you’ve got.”

Given that microservices are not the right solution in every case, what are the clues that an existing monolithic application might actually be a good candidate for decomposing into microservices? “You’ve tried all the easy stuff,” Newman told The Register. “Have you done some value chain analysis? Have you looked at where the bottlenecks are? Have you tried modularisation? Microservices should be a last resort.”

Even scalability may not be a good enough reason to adopt microservices. “My clients say, my application doesn’t scale. I ask, have you tried running 10 copies of your monolith? And they haven’t. Do all the easy stuff first.” 

The post Microservices guru warns devs that trendy architecture shouldn’t be the default for every app, but ‘a last resort’ appeared first on Artificial Intelligence.

It is surprising how much can the world change in a few years. Today, machines can see (a field popularly known as Computer Vision), they can hear (Hi Alexa!) and understand our language (field called Natural Language Processing) to a good degree. And they are only going to get better from here.

On some specific tasks, machines actually perform better than humans. This includes identifying objects in images or doing video surveillance without getting tired. We have seen drastic improvements in virtual assistants, chatbots or use of voice based systems for customer service.

If these developments make you wonder what is driving these changes then you are not alone. All of these developments have been fueled by the power of Data Science. Data Science is nothing but the science of extracting insights from data. With ongoing digitalization, dropping storage costs and increasing compute power, the barriers to use Data Science have come down significantly in the last decade.

Some of the latest developments in Data Science:

1. Machine Learning – Machine Learning is a field in Data Science where we train machines to find patterns and insights on its own and then use that to solve problems for us.

For example, using data to identify which customers are likely going to churn in next 3 months based on latest insights from data is an example of machine learning.

2. Deep Learning – Deep Learning is a special class of machine learning algorithms, which were inspired by how the human brain works. Deep learning has made a significant impact on several real life problems and we are already using it in a lot of products we use on a day to day basis.

For example, text prediction in smartphones to Gmail, or cameras identifying faces by themselves is driven by deep learning.

3. Natural Language Processing – Algorithms which helps machine understand human language and what we mean are part of techniques commonly known as Natural Language Processing. This helps news companies to create summaries of articles to tracking customer sentiment on social media.

4. Computer Vision – Videos and images are nothing but a different form of data. Algorithms which work to help machines see and interpret images (and videos) are popularly known as computer vision. Detecting workers without proper safety gear in a manufacturing plant is an example of computer vision.

Now, a lot of these techniques have been around for some time. But when these techniques are supplied with large amounts of data and compute power, magic starts to happen. This is why we are seeing a whole new set of applications / software becoming smart and changing user experience.

The impact from these technologies increases significantly when we combine these fields to create complex and automated systems. These systems can perform all those tasks which were historically associated with humans.

The coming decade

In the coming decade, we will see a whole range of new products which will increasingly do work we believed only humans could do for a long time. This is liberating for some of us and also challenging for a lot of us as new applications would improve our lives, but at the same point there will also be job replacements at a large scale.

But whatever happens, one thing is for sure that the coming decade is going to be exciting and Data Science will be fuelling the future for the coming years!

The post Check latest, upcoming technologies in Data Science appeared first on Artificial Intelligence.

The deployment of machine learning models in production is failing to keep pace with the everyday operations of hyper-scalers. Those scaling and deployment gaps are being addressed through collaborations and a new batch of DevOps tools tuned to scaling machine learning deployments.

The latest initiative aimed at expanding the MLOps ecosystem comes via a series of technology partnerships and native integration with tools such as TensorFlow along with expanded cloud infrastructure support. Dotscience, a machine learning operations specialist, announced partnerships with GitLab and Grafana Labs along with platform integrations to TensorFlow, H2O.ai and Scikit-learn.

The London-based MLOps vendor also announced expanded multi-cloud support with Amazon Web Services (NASDAQ: AMZN) and Microsoft Azure (NASDAQ: MSFT).

Collectively, the goal is “setting the bar for MLOps best practices for building production ML pipelines today,” said Luke Marsden, CEO and founder at Dotscience.

The partners also hope to tap into a projected $3.9 trillion market for AI-based opportunities over the next two years. Vendors such as Dotscience and Algorithmia are rolling out new tools aimed at enterprises currently struggling to deploy machine learning models in production. Among the initiatives announced by Dotscience on Wednesday (Dec. 18) is an effort to develop an industry benchmark for enterprise AI deployments.

To that end, platform monitoring specialist Grafana Labs and Dotscience are joining forces to improve visibility into machine learning workloads in production. The partnership would allow statistical monitoring of model behavior, including workloads using unlabeled data.

The partners also said their monitoring framework would help simplify model deployments via the Kubernetes cluster orchestrator.

“By bringing DevOps practices to ML, data science and ML teams can eliminate silos,” said Tom Wilkie, Grafana Labs’ vice president for products.

Separately, Dotscience announced a native integration with GitLab, the open source software repository. The collaboration would allow ML developers to use the Dotscience platform for machine learning data and model management. More than 100,000 developers currently use GitLab as a DevOps platform.

“We are enabling data scientists to deploy on their preferred ML framework.” Dotscience CEO Marsden said.

Those MLOps enhancement will be augmented with expanded cloud support via the AWS Marketplace and Microsoft Azure. The Dotscience platform is available either as a software service or on-premises.

The MLOps specialist also this week said financial API provider TrueLayer will use the Dotscience platform to improve reproducibility, model and data versioning along with provenance tracking. The partnership comes in response to growing demand for improved productivity, collaboration, governance and compliance as AI initiatives ramp up, the partners said.

The post MLOps Initiatives Seek to Boost Stalled Deployments appeared first on Artificial Intelligence.

AI is advancing rapidly within the enterprise — by Gartner’s count, more than half of organizations already have at least one AI deployment in operation, and they’re planning to substantially accelerate their AI adoption within the next few years. At the same time, the organizations building and deploying these tools have yet to really grapple with the flaws and shortcomings of AI– whether the models deployed are fair, ethical, secure or even explainable.  

Before the world is overrun with flawed AI systems, IBM is aiming to rev up the development of open source trusted AI workflows. As part of that effort, the company is joining the Linux Foundation AI (LF AI) as a General Member. 

“AI, as it matures, needs to mature in a way that is something that the general public can put their confidence and trust in,” Todd Moore, IBM’s VP of Open Technology, told ZDNet. “Too often, what we hear is the AI is a black box, they don’t understand how it got to its results, there’s bias in the models, there needs to be more fairness… We’ve heard that loud and clear, and we felt it was time to help the industry move forward.”

AI is advancing rapidly within the enterprise — by Gartner’s count, more than half of organizations already have at least one AI deployment in operation, and they’re planning to substantially accelerate their AI adoption within the next few years. At the same time, the organizations building and deploying these tools have yet to really grapple with the flaws and shortcomings of AI– whether the models deployed are fair, ethical, secure or even explainable.  

Before the world is overrun with flawed AI systems, IBM is aiming to rev up the development of open source trusted AI workflows. As part of that effort, the company is joining the Linux Foundation AI (LF AI) as a General Member. 

“AI, as it matures, needs to mature in a way that is something that the general public can put their confidence and trust in,” Todd Moore, IBM’s VP of Open Technology, told ZDNet. “Too often, what we hear is the AI is a black box, they don’t understand how it got to its results, there’s bias in the models, there needs to be more fairness… We’ve heard that loud and clear, and we felt it was time to help the industry move forward.”

As a Linux Foundation project, the LF AI Foundation provides a vendor-neutral space for the promotion of Artificial Intelligence (AI), Machine Learning (ML) and Deep Learning (DL) open source projects. It’s backed by major organizations like AT&T, Baidu, Ericsson, Nokia and Huawei. 

IBM has a long history of supporting open source, and Moore explained why it’s the right way to quickly raise the bar when it comes to building trustworthy AI. “To get all of us working together, iterating quickly, can cover a lot more ground than any single company can,” he said. 

On top of that, supporting open source projects has the added benefit of expanding the market opportunity for AI vendors like IBM. The goal, Moore said, is to build tools that improve the credibility of AI — and “to do it together, in a way that everybody can inspect and contribute to.” 

By joining LF AI, IBM is aiming to bring trusted AI techniques to all of the foundation’s projects. The company will work with LF AI’s committees to create reference architectures and best practices for using open source tools in production. 

IBM has already spearheaded efforts on this front with a series of open source toolkitsdesigned to help build trusted AI. The AI Fairness 360 Toolkit helps developers and data scientists detect and mitigate unwanted bias in machine learning models and datasets. TheAdversarial Robustness 360 Toolbox is an open source library that helps researchers and developers defend deep neural networks from adversarial attacks. Meanwhile, the AI Explainability 360 Toolkit provides a set of algorithms, code, guides, tutorials and demos to support the interpretability and explainability of machine learning models.

Meanwhile, IBM has already started working on an informal basis with the LF AI Foundation, participating in events and contributing to projects like the Foundation Technical Advisory Committee’s ML Workflow project.

The work of creating trusted AI is in nascent stages, but Moore said, “The good thing is it’s started, the problem has been recognized. It’s up to us to build the de facto standards and create the tools to help people.”

The post IBM joins Linux Foundation AI to promote open source trusted AI workflows appeared first on Artificial Intelligence.

loud computing leaders Amazon.com (AMZN), Microsoft (MSFT) and Google, along with internet giants, have the inside track in monetizing artificial intelligence technology, a Mizuho Securities report says, but enterprise software providers also will grab a piece of the action.

At a basic level, artificial intelligence is the use of computer algorithms to attempt to replicate the human ability to learn, reason and make decisions.

Amazon, Microsoft and Google-parent Alphabet (GOOGL) are pushing into AI-as-a-service. They plan to rent AI tools to customers on a pay-as-you-go basis via their public cloud-computing services.

Enterprise software providers such as Salesforce.com (CRM) and ServiceNow (NOW) are in the early stages of embedding AI tools in existing products to make them more predictive, analysts say.

“In evaluating recent artificial intelligence/machine learning trends and innovation put forth by enterprise vendors, we see incremental benefit for Salesforce, ServiceNow and Splunk (SPLK),” Mizuho analyst Abhey Lamba said. “Across the stack, we think the revenue opportunity lies with application vendors that address specific business problems or analytical tool providers that help enable a solution. The core AI engine layer of the stack will likely be dominated by large vendors such as Amazon, Apple (AAPL), Facebook (FB), Google and Microsoft; smaller players are likely to either get acquired or run into significant headwinds.”

AI, around since the 1960s, now comes in new forms. The two most mentioned are “machine learning” and “deep learning networks.” The basic idea is to make better predictions by analyzing massive amounts of data.

Analysts expect companies spanning banks and finance, health care, energy, retail, agriculture and other sectors to increase spending on AI technology.

Salesforce.com introduced its “Einstein” AI cloud platform in September 2016. Einstein tools identify and prioritize sales leads and recommend products and pricing options, making it easier for sales reps to convert leads to sales.

“Salesforce.com continues to be one of the biggest proponents of machine learning, with early organic investments plus robust M&A,” Lamba added. “Einstein is seeing strong interest from customers but production deployments could take some time. ServiceNow is likely to sustain momentum as it pushes for deeper enterprise automation across functional areas with investments in product.”

The post These Stocks Could Profit From Artificial Intelligence: Analyst appeared first on Artificial Intelligence.