The Apache Software Foundation (ASF) recently announced that SINGA, a framework for distributed deep-learning, has graduated to top-level project (TLP) status, signifying the project’s maturity and stability. SINGA has already been adopted by companies in several sectors, including banking and healthcare.

Originally developed at the National University of Singapore, SINGA joined ASF’s incubator in March 2015. SINGA provides a framework for distributing the work of training deep-learning models across a cluster of machines, in order to reduce the time needed to train the model. In addition to its use as a platform for academic research, SINGA has been used in commercial applications by Citigroup and CBRE, as well as in several health-care applications, including an app to aid patients with pre-diabetes.

The success of deep-learning models has been driven by the use of very large datasets, such as ImageNet with hundreds of thousands of images, and complex models with millions of parameters. Google’s BERT natural-language model contains 300 million parameters and is trained on nearly 3 billion words. However, this training often requires hours, if not days, to complete. To speed up this process, researchers have turned to parallel computing, which distributes the work across a cluster of machines. According to Professor Beng Chin Ooi, leader of the research group that developed SINGA:

It is essential to scale deep learning via distributed computing as…deep learning models are typically large and trained over big datasets, which may take hundreds of days using a single GPU.

There are two broad parallelism strategies for distributed deep-learning: data parallelism, where multiple machines work on different subsets of the input data, and model parallelism, where multiple machines train different sections of the neural-network model. SINGA supports both of these strategies, as well as a combination of the two. These strategies do introduce some communication and synchronization overhead, required to coordinate the work among the machines in the cluster. SINGA implements several optimizations to minimize this overhead.

Acceptance as a top-level project means that SINGA has passed several milestones related to software quality and community, which in theory makes the software more attractive as a solution. However, one possible barrier to adoption is that instead of building upon an existing API for modeling neural networks, such as Keras, SINGA’s designers chose to implement their own. By contrast, the Horovod framework open-sourced by Uber allows developers to port existing models written for the two most popular deep-learning frameworks, TensorFlow and PyTorch. PyTorch in particular is the framework used in a majority of recent research papers.

ASF has several other top-level distributed-data processing projects that support machine-learning, including Spark and Ignite. Unlike these, SINGA is designed specifically for deep-learning’s large models. ASF is also home to MXNet, a deep-learning framework similar to TensorFlow and PyTorch, which is still in incubator status. AWS touted MXNet as its framework of choice in late 2016, but MXNet still hasn’t achieved widespread popularity, hovering at just under 2% in KDNugget’s polls.

Apache SINGA version 2.0 was released in April, 2019. The source code is available on GitHub, and a list of open issues can be tracked in SINGA’s Jira project. According to ASF, upcoming features include “SINGA-lite for deep learning on edge devices with 5G, and SINGA-easy for making AI usable by domain experts (without deep AI background).

The post Deep-Learning Framework SINGA Graduates to Top-Level Apache Project appeared first on Artificial Intelligence.

Organisations are looking to AI models to bring out digital transformation in business. But, understanding which kind of models are most suited to the business needs is crucial

As enterprises are discovering the benefits of artificial intelligence (AI), they realise the journey to AI is long and bumpy. Many CIOs (chief information officers) want AI to quickly transform their business without identifying which processes will perform better with AI. In an ideal world, one can pick any process, infuse it with AI and then discover the pros and cons during the journey. The best way to learn is on projects rather than researching through theoretical case studies.

But there needs to be some order to the madness. Can we generalise some patterns that could make it easy for business owners to apply AI? Let’s discuss some scalable, enterprise-relevant AI patterns.

Discover new processes: This is about finding new opportunities afforded by AI. Consider the example of a defect in a machine, degrading over time. An experienced mechanical engineer can deduct the condition of the machine from the sound generated. What if there was a process wherein the mechanical engineer documents what he ‘hears’ and how he maintains the machine? This is where an acoustic AI model can be created, which can analyse sound samples of the machine to predict failures. It’s common sense for an engineer that a noisy machine is the first sign of mechanical failure; shouldn’t this important data can be put to value?

Imagine driving down a highway and an alert pops up on the dashboard saying, “Possible less lubricant”. It confirms the driver’s gut feeling that there’s something wrong in the car.

Most of the acoustic models today use humans to classify the data fed into the model; over time, the model learns to classify on its own. For example, we need to gather at least 10,000 sound clips of failed and normal ball bearings to classify the anomalous sounds and detect the issue. Not an easy task, but it can tremendously help in predicting failures in mines, underground subways, nuclear plants and highly critical sites unapproachable by humans.

Reinvigorate old processes: This pattern improves existing processes by introducing AI. For instance, almost every organisation collects data from the employees’ badges, which provides information on access control and employee movement. Reinvigorating this process by adding occupancy sensors and then adding AI will help derive deeper insights such as the number of people per floor. This data can be fed into an AI model to predict the occupancy rate and help organisations reduce the cost associated with each desk and decide how much office space should be leased or vacated. The ability to predict churn of clients, machine failure, energy usage etc are all examples of how old processes can be reinvigorated with new AI models.

Unlock data: Organisations can derive value from their data by applying AI. For example, machine learning algorithms can be used to detect fraud in financial transactions or even an asset defect, which would otherwise go unnoticed by humans. One Machine learning model can be fed time-series data to discover patterns of anomalies, while another can be fed asset manuals to find contextual text of the faults. One of the widely used examples of applying AI to businesses is handling unstructured data in the form of texts, videos, and tweets. Several organisations across industries have benefitted from this pattern, including telcos with millions of call records, banks with loan records, manufacturing units with work orders etc.

Opening new channels: This is another area where we have seen several businesses apply AI successfully. This essentially means starting a new channel of interaction with customers or employees using AI-based virtual assistants with natural language processing technologies. Unlike the dated IVR system, this new channel is helping organisations reach their clients and service them in unique ways.

We can pick any AI process and it’s sure to fall in one of the four patterns mentioned above. What is needed is the right understanding of which process to choose and then applying the right AI methodology to solve the business problem. Once the process is chosen, along with the the right algorithm and data quality, one also needs to check for bias in the model. Explanation of why a certain recommendation is the most right one needs to be included too.

The post How to choose the right AI model for your business appeared first on Artificial Intelligence.

With the demand for workers with advanced tech skills skyrocketing, many companies are putting more resources into recruiting, hiring and nurturing the right talent to remain in the global competition.

That means employees willing to put time into developing tech skills may have the upper-hand in landing some of the most in-demand roles.

But as technology continues to evolve at a rapid pace, it can be difficult to know exactly what skill sets are necessary to thrive in different fields. One way to find out is to understand people are spending their time learning, which is exactly what Udemy’s 2020 Workplace Learning Trends Report aims to do.

After analyzing data from more than 40 million users, the online learning platform found that the most popular tech skill people are learning is Python, a programming language. Overall, the report notes there’s a huge interest in learning about artificial intelligence (AI) and data science.

Shelley Osborne, vice president of learning at Udemy, told CNBC Make It that while tech courses tend to be more robust, some of the briefer introductory courses can be helpful to those who don’t even work in the tech field. “We sometimes see these topics trending with executive-level leaders who want to better understand their business’ approach using data science,” she said.

As it becomes easier for companies to parse out data, the need for workers to interpret those data sets is becoming more crucial. Jobs involving data science skills have been named as some of the most promising jobs in the U.S., according to LinkedIn.

“Organizations are becoming more data-driven, and that’s partly because they’re harnessing the power of AI, and there’s a need to analyze and process data across all kinds of roles,” Jennifer Juo, who leads the content marketing team at Udemy, told CNBC Make It.

Hiring managers are also having an especially hard time trying to fill roles that require skills in software development (e.g., Python, JavaScript), AI and cloud computing (e.g., Amazon Web Services, Google Cloud) and business intelligence (e.g., Microsoft Business Intelligence).

According to a 2019 report from iCIMS, a recruitment software provider, it took companies an average of 55 days to fill a tech role in 2016. In 2019, that number jumped to 66 days. These unfilled roles can cost about $680 in lost revenue per day per vacancy, according to iCIMS.

“We’re seeing a shift in skills development that requires us to think differently about how we approach talent,“said Osborne. She suggests companies encourage current employees to take classes and develop the essential skills that they lack.

Here are the top 10 most popular tech skills of 2020 — and where workers are leveling up the most:

1. Python

A programming language used in software development, infrastructure management and data analysis.

2. React (web)

A JavaScript library for building user interfaces.

3. Angular

A JavaScript-based open-source front-end web framework.

4. Machine learning

The scientific study of algorithms and statistical models.

5. Docker

An open-source platform used to create software packages called containers.

6. Django

A Python-based free and open-source web framework

7. CompTIA

A professional tech organization that has four IT certification series ranging from entry-level to expert.

8. Amazon AWS

A certification that validates cloud expertise.

9. Deep learning

A class of machine learning based on artificial neural networks.

10. React Native (mobile)

An open-source mobile application framework created by Facebook to develop apps for Android, iOS, Web and Universal Windows Platform.

The post These will be the top 10 most popular tech skills of 2020 appeared first on Artificial Intelligence.

Machine learning is one of the most promising technologies today that makes it possible for machines to learn how the human brain works and replicate this learning to analyze varied data types and deduce meaningful insights.

Machine learning models

At the core of machine learning are three models that help machines unearth insights and patterns. These are:

• Supervised models: These are used with historical data where the output is pre-defined. For instance, when you speak, Alexa can recognize the words and sentences she has been trained on and respond appropriately.
• Unsupervised models: These are used on transactional data to identify patterns. Based on your interaction with Alexa, she can identify the patterns to suggest topics you may be interested in.
• Reinforcement learning: It is a technique where machines learn to respond to situations on their own, without instructions. For every mistake (a negative outcome) that Alexa makes, she ‘learns’ from it to become smarter and refine the response next time.

FIs can benefit the most from machine learning

Businesses are increasingly leaning on machine learning, as volumes of data are exploding and they need actionable insights to fuel business growth. Given the benefits it promises, numerous industries—manufacturing, energy, healthcare, cyber defense, financial institutions—are making significant investments in machine learning. In fact, financial institutions (FIs) stand to benefit the most from machine learning, according to a PwC report.

Money-rich FIs, especially banks, have always been a favorite target for criminals. And, today’s technological advancements have provided cyber criminals with sophisticated techniques—data breach, phishing, malware, sweatshops, and so forth—to break into business systems and cause losses.

Machine learning, with its innate ability to monitor millions of online transactions in real-time, can help financial institutions in a myriad of ways.

• Document interpretation: Machine learning helps financial institutions interpret financial and legal documents—bank statements, tax statements, contracts, etc—across a wide range of parameters that help gain in-depth insights into customers’ financial health.
• Risk management: Financial institutions can accurately assess the credit-worthiness of a customer—whether an individual or a company—and make informed lending decisions for improved risk management.
• Additional revenue: Using analytics to understand customer preferences and inclination to spend, financial institutions can harness these insights to pitch other products and services to increase their revenue.
• Customer service: Applying behavioral analytics, banks and financial institutions can better understand the financial needs of their customers and offer more relevant services. This enables financial institutions to strengthen customer relationships and earn their trust.
• Channel-agnostic access: Leveraging customer data to anticipate customers’ channel preferences, financial institutions can provide seamless user experience to their customers across devices and locations.
• Process automation: Machine learning helps financial institutions make automated decisions in real-time that reduces the response time. According to Accenture, FIs can reduce costs incurred on middle and back offices across infrastructure, maintenance, and operations by 20-25%.
• Security: With fraud on the rise, financial institutions are obliged to ensure online security of their customers. Customer security is the most important area where machine learning has proved immensely helpful in fighting fraud by accurately identifying fraudsters from a group of authentic customers. Real-time analysis of digital intelligence enables financial institutions to prevent fraud from poisoning their business ecosystem, thereby providing customers with a safe and secure online journey.

The post 7 ways machine learning helps financial institutions appeared first on Artificial Intelligence.

Despite huge advances in science over the past century, our understanding of nature is still far from complete. Not only have scientists failed to find the Holy Grail of physics – unifying the very large (general relativity) with the very small (quantum mechanics) – they still don’t know what the vast majority of the universe is made up of. The sought after Theory of Everything continues to elude us. And there are other outstanding puzzles, too, such as how consciousness arises from mere matter.

Will science ever be able to provide all the answers? Human brains are the product of blind and unguided evolution. They were designed to solve practical problems impinging on our survival and reproduction, not to unravel the fabric of the universe. This realization has led some philosophers to embrace a curious form of pessimism, arguing there are bound to be things we will never understand. Human science will therefore one day hit a hard limit – and may already have done so.Rented shoes are grossBut bowling is fun! Join us for Bowlr, Amsterdam’s best networking eventYEAH!

Some questions may be doomed to remain what the American linguist and philosopher Noam Chomsky called “mysteries”. If you think that humans alone have unlimited cognitive powers – setting us apart from all other animals – you have not fully digested Darwin’s insight that Homo Sapiens is very much part of the natural world.

But does this argument really hold up? Consider that human brains did not evolve to discover their own origins either. And yet somehow we managed to do just that. Perhaps the pessimists are missing something.

Mysterian arguments

“Mysterian” thinkers give a prominent role to biological arguments and analogies. In his 1983 landmark book The Modularity of Mind, the late philosopher Jerry Fodor claimed that there are bound to be “thoughts that we are unequipped to think”.

Similarly, the philosopher Colin McGinn has argued in a series of books and articles that all minds suffer from “cognitive closure” with respect to certain problems. Just as dogs or cats will never understand prime numbers, human brains must be closed off from some of the world’s wonders. McGinn suspects that the reason why philosophical conundrums such as the mind/body problem – how physical processes in our brain give rise to consciousness – prove to be intractable is that their true solutions are simply inaccessible to the human mind.

If McGinn is right that our brains are simply not equipped to solve certain problems, there is no point in even trying, as they will continue to baffle and bewilder us. McGinn himself is convinced that there is, in fact, a perfectly natural solution to the mind–body problem, but that human brains will never find it.

Even the psychologist Steven Pinker, someone who is often accused of scientific hubris himself, is sympathetic to the argument of the mysterians. If our ancestors had no need to understand the wider cosmos in order to spread their genes, he argues, why would natural selection have given us the brainpower to do so?

Mind-boggling theories

Mysterians typically present the question of cognitive limits in stark, black-or-white terms: either we can solve a problem, or it will forever defy us. Either we have cognitive access or we suffer from closure. At some point, human inquiry will suddenly slam into a metaphorical brick wall, after which we will be forever condemned to stare in blank incomprehension.

Another possibility, however, which mysterians often overlook, is one of slowly diminishing returns. Reaching the limits of inquiry might feel less like hitting a wall than getting bogged down in a quagmire. We keep slowing down, even as we exert more and more effort, and yet there is no discrete point beyond which any further progress at all becomes impossible.

There is another ambiguity in the thesis of the mysterians, which my colleague Michael Vlerick and I have pointed out in an academic paper. Are the mysterians claiming that we will never find the true scientific theory of some aspect of reality, or alternatively, that we may well find this theory but will never truly comprehend it?

In the science fiction series The Hitchhiker’s Guide to The Galaxy, an alien civilization builds a massive supercomputer to calculate the Answer to the Ultimate Question of Life, the Universe and Everything. When the computer finally announces that the answer is “42”, no one has a clue what this means (in fact, they go on to construct an even bigger supercomputer to figure out precisely this).

Is a question still a “mystery” if you have arrived at the correct answer, but you have no idea what it means or cannot wrap your head around it? Mysterians often conflate those two possibilities.

In some places, McGinn suggests that the mind–body problem is inaccessible to human science, presumably meaning that we will never find the true scientific theory describing the mind–body nexus. At other moments, however, he writes that the problem will always remain “numbingly difficult to make sense of” for human beings, and that “the head spins in theoretical disarray” when we try to think about it.

This suggests that we may well arrive at the true scientific theory, but it will have a 42-like quality to it. But then again, some people would argue that this is already true of a theory like quantum mechanics. Even the quantum physicist Richard Feynman admitted, “I think I can safely say that nobody understands quantum mechanics.”

Would the mysterians say that we humans are “cognitively closed” to the quantum world? According to quantum mechanics, particles can be in two places at once, or randomly pop out of empty space. While this is extremely hard to make sense of, quantum theory leads to incredibly accurate predictions. The phenomena of “quantum weirdness” have been confirmed by several experimental tests, and scientists are now also creating applications based on the theory.

Mysterians also tend to forget how mindboggling some earlier scientific theories and concepts were when initially proposed. Nothing in our cognitive make-up prepared us for relativity theory, evolutionary biology or heliocentrism.

As the philosopher Robert McCauley writes: “When first advanced, the suggestions that the Earth moves, that microscopic organisms can kill human beings, and that solid objects are mostly empty space were no less contrary to intuition and common sense than the most counterintuitive consequences of quantum mechanics have proved for us in the twentieth century.” McCauley’s astute observation provides reason for optimism, not pessimism.

Mind extensions

But can our puny brains really answer all conceivable questions and understand all problems? This depends on whether we are talking about bare, unaided brains or not. There’s a lot of things you can’t do with your naked brain. But Homo Sapiens is a tool-making species, and this includes a range of cognitive tools.

For example, our unaided sense organs cannot detect UV-light, ultrasound waves, X-rays or gravitational waves. But if you’re equipped with some fancy technology you can detect all those things. To overcome our perceptual limitations, scientists have developed a suite of tools and techniques: microscopes, X-ray film, Geiger counters, radio satellites detectors and so forth.

All these devices extend the reach of our minds by “translating” physical processes into some format that our sense organs can digest. So are we perceptually “closed” to UV light? In one sense, yes. But not if you take into account all our technological equipment and measuring devices.

In a similar way, we use physical objects (such as paper and pencil) to vastly increase the memory capacity of our naked brains. According to the British philosopher Andy Clark, our minds quite literally extend beyond our skins and skulls, in the form of notebooks, computers screens, maps and file drawers.

Mathematics is another fantastic mind-extension technology, which enables us to represent concepts that we couldn’t think of with our bare brains. For instance, no scientist could hope to form a mental representation of all the complex interlocking processes that make up our climate system. That’s exactly why we have constructed mathematical models and computers to do the heavy lifting for us.

Cumulative knowledge

Most importantly, we can extend our own minds to those of our fellow human beings. What makes our species unique is that we are capable of culture, in particular cumulative cultural knowledge. A population of human brains is much smarter than any individual brain in isolation.

And the collaborative enterprise par excellence is science. It goes without saying that no single scientist would be capable of unravelling the mysteries of the cosmos on her own. But collectively, they do. As Isaac Newton wrote, he could see further by “standing on the shoulders of giants”. By collaborating with their peers, scientists can extend the scope of their understanding, achieving much more than any of them would be capable of individually.

Today, fewer and fewer people understand what is going on at the cutting edge of theoretical physics – even physicists. The unification of quantum mechanics and relativity theory will undoubtedly be exceptionally daunting, or else scientists would have nailed it long ago already.

The same is true for our understanding of how the human brain gives rise to consciousness, meaning and intentionality. But is there any good reason to suppose that these problems will forever remain out of reach? Or that our sense of bafflement when thinking of them will never diminish?

In a public debate I moderated a few years ago, the philosopher Daniel Dennett pointed out a very simple objection to the mysterians’ analogies with the minds of other animals: other animals cannot even understand the questions. Not only will a dog never figure out if there’s a largest prime, but it will never even understand the question. By contrast, human beings can pose questions to each other and to themselves, reflect on these questions, and in doing so come up with ever better and more refined versions.

Mysterians are inviting us to imagine the existence of a class of questions that are themselves perfectly comprehensible to humans, but the answers to which will forever remain out of reach. Is this notion really plausible (or even coherent)?

Alien anthropologists

To see how these arguments come together, let’s do a thought experiment. Imagine that some extraterrestrial “anthropologists” had visited our planet around 40,000 years ago to prepare a scientific report about the cognitive potential of our species. Would this strange, naked ape ever find out about the structure of its solar system, the curvature of space-time or even its own evolutionary origins?

At that moment in time, when our ancestors were living in small bands of hunter-gatherers, such an outcome may have seemed quite unlikely. Although humans possessed quite extensive knowledge about the animals and plants in their immediate environment, and knew enough about the physics of everyday objects to know their way around and come up with some clever tools, there was nothing resembling scientific activity.

There was no writing, no mathematics, no artificial devices for extending the range of our sense organs. As a consequence, almost all of the beliefs held by these people about the broader structure of the world were completely wrong. Human beings didn’t have a clue about the true causes of natural disaster, disease, heavenly bodies, the turn of the seasons or almost any other natural phenomenon.

Our extraterrestrial anthropologist might have reported the following:

Evolution has equipped this upright, walking ape with primitive sense organs to pick up some information that is locally relevant to them, such as vibrations in the air (caused by nearby objects and persons) and electromagnetic waves within the 400-700 nanometer range, as well as certain larger molecules dispersed in their atmosphere.

However, these creatures are completely oblivious to anything that falls outside their narrow perceptual range. Moreover, they can’t even see most of the single-cell life forms in their own environment, because these are simply too small for their eyes to detect. Likewise, their brains have evolved to think about the behavior of medium-sized objects (mostly solid) under conditions of low gravity.

None of these earthlings has ever escaped the gravitational field of their planet to experience weightlessness, or been artificially accelerated so as to experience stronger gravitational forces. They can’t even conceive of space-time curvature, since evolution has hard-wired zero-curvature geometry of space into their puny brains.

In conclusion, we’re sorry to report that most of the cosmos is simply beyond their ken.

But those extraterrestrials would have been dead wrong. Biologically, we are no different than we were 40,000 years ago, but now we know about bacteria and viruses, DNA and molecules, supernovas and black holes, the full range of the electromagnetic spectrum and a wide array of other strange things.

We also know about non-Euclidean geometry and space-time curvature, courtesy of Einstein’s general theory of relativity. Our minds have “reached out” to objects millions of light years away from our planet, and also to extremely tiny objects far below the perceptual limits of our sense organs. By using various tricks and tools, humans have vastly extended their grasp on the world.

The verdict: biology is not destiny

The thought experiment above should be a counsel against pessimism about human knowledge. Who knows what other mind-extending devices we will hit upon to overcome our biological limitations? Biology is not destiny. If you look at what we have already accomplished in the span of a few centuries, any rash pronouncements about cognitive closure seem highly premature.

Mysterians often pay lip service to the values of “humility” and “modesty”, but on closer examination, their position is far less restrained than it appears. Take McGinn’s confident pronouncement that the mind–body problem is “an ultimate mystery” that we will “never unravel”. In making such a claim, McGinn assumes knowledge of three things: the nature of the mind–body problem itself, the structure of the human mind, and the reason why never the twain shall meet. But McGinn offers only a superficial overview of the science of human cognition, and pays little or no attention to the various devices for mind extension.

I think it’s time to turn the tables on the mysterians. If you claim that some problem will forever elude human understanding, you have to show in some detail why no possible combination of mind extension devices will bring us any closer to a solution. That is a taller order than most mysterians have acknowledged.

Moreover, by spelling out exactly why some problems will remain mysterious, mysterians risk being hoisted by their own petard. As Dennett wrote in his latest book: “As soon as you frame a question that you claim we will never be able to answer, you set in motion the very process that might well prove you wrong: you raise a topic of investigation.”

In one of his infamous memorandum notes on Iraq, former US secretary of defense, Donald Rumsfeld, makes a distinction between two forms of ignorance: the “known unknowns” and “unknown unknowns”. In the first category belong the things that we know we don’t know. We can frame the right questions, but we haven’t found the answers yet. And then there are the things that “we don’t know we don’t know”. For these unknown unknowns, we can’t even frame the questions yet.

It is quite true that we can never rule out the possibility that there are such unknown unknowns, and that some of them will forever remain unknown, because for some (unknown) reason human intelligence is not up to the task.

But the important thing to note about these unknown unknowns is that nothing can be said about them. To presume from the outset that some unknown unknowns will always remain unknown, as mysterians do, is not modesty – it’s arrogance.

The post Why philosophers believe we’ve reached peak human intelligence appeared first on Artificial Intelligence.