10 Big Data Trends to Watch in 2019

By TheWAY - 1월 26, 2019

10 Big Data Trends to Watch in 2019

We seek ever more data for a good reason: it’s the commodity that fuels digital innovation. However, turning those huge data collections into actionable insight remains a difficult proposition. Organizations that find solutions to formidable data challenges will be better positioned to economically benefit from the fruits of digital innovation.

With that basic premise in mind, here are 10 trends in big data that forward-looking organizations should look out for in 2019:

1. Data Management Is Still Hard

The big idea behind big data analytics is fairly clear-cut: Find interesting patterns hidden in large amounts of data, train machine learning models to spot those patterns, and implement those models into production to automatically act upon them. Rinse and repeat as necessary.

(kuruneko/Shutterstock)

However, the reality of putting that basic recipe into production is a lot harder than it looks. For starters, amassing data from different silos (see prediction #1) is difficult and requires ETL and database skills. Cleaning and labeling the data for the machine learning training also takes a lot of time and money, particularly when deep learning techniques are used. And finally, putting such a system into production at scale in a secure and reliable fashion requires another set of skills entirely.

For these reasons, data management remains a big challenge, and data engineers will continue to be among the most sought-after personas on the big data team.

2. Data Silos Continue Proliferating

This is not a difficult prediction to make. During the Hadoop boom five years ago, we were entranced with the idea that we could consolidate all of our data – for both analytical and transactional workloads – onto a single platform.

Data is stored in silos. Get used to it. (Aleutie/Shutterstock)

That idea never really panned out, for a variety of reasons. The biggest challenges is that different data types have different storage requirements. Relational database, graph databases, time-series databases, HDFS, and object stores all have their respective strengths and weakness. Developers can’t maximize strengths if they’ve crammed all their data into a one-size-fits-all data lake.

In some cases, amassing lots of data into a single place does make sense. Cloud data stores like S3, for instance, are providing companies with flexible and cost-effective storage, and Hadoop continues to be a cost-effective store for unstructured data storage and analytics. But for most companies, these are simply additional silos that must be managed. They’re big and important silos, of course, but they’re not the only ones.

In the absence of a strong centralizing force, data silos will continue to proliferate. Get used to it.

3. Streaming Analytics Has Breakout Year

The quicker you can act on a new piece of data, the better off your organization will be. That’s the driving force behind real-time, or streaming, analytics. The challenge has always been that it’s rather difficult to actually pull off and expensive too, but that’s changing as organizations’ analytic teams mature and the technology gets better.

NewSQL databases, in-memory data grids, and dedicated streaming analytic platforms are converging around a common capability, which is ultra-fast processing of incoming data, often using machine learning models to automate decision-making.

Combine that with the SQL capabilities in open source streaming frameworks like Kafka, Spark, and Flink, and you have the recipe for real progress in 2019.

4. Data Governance Builds Steam

Some people call data the “new oil.” It’s also been called the “new currency.” Whichever analogy you want to use, we all agree that data has value, and that treating it carelessly carries a risk.

The European Union spelled out the financial consequences for poor data governance with last year’s enactment of the GDPR. While there’s no similar law in the United States yet, American companies still must abide by 80-some different data mandates created by various states, countries, and unions.

Data breaches are bringing the issue to a head. According to an online survey by The Harris Poll, nearly 60 million Americans were affected by identity theft in 2018. That’s an increase of 300% from 2017, when just 15 million say they were affected.

Most organizations have realized that the Wild West days of big data are coming to an end. While the US Government won’t (yet) fine you for being reckless with data or abusing the privacy of American citizens, the writing is on the wall that this behavior is no longer tolerated.

5. Skills Shift as Tech Evolves

Human resources are typically the biggest costs in a big data project, because people ultimately are the ones that build it and run it and make it all work. Finding the right person with the right skills is absolutely critical to turning data into insight, no matter what technologies or techniques you’re using.

But as technology advances, the skills mix does too. In 2019, you can expect to see continued huge demand for anybody who can put a neural network into production. Among mere data scientists (as opposed to legit AI experts), Python continues to dominate among languages, although there’s plenty of work for folks who know R, SAS, Matlab, Scala, Java, and C.

(Image courtesy Qubole)

As data governance programs kick into gear, demand for data stewards will go up. Data engineers who can work with the core tools (databases, Spark, Airflow, etc.) will continue to see their opportunities grow. You can also expect to see demand for machine learning engineers accelerate.

However, thanks to the advance of automated data science platforms, organizations will be able to accomplish quite a bit with mere data analysts, or “citizen data scientists,” as they’re commonly known. Knowledge of the data and the business – as opposed to expertise in statistics and coding – may get you further down the big data road than you imagined.

6. Deep Learning Gets Deeper

The “Cambrian explosion” of deep learning, which has powered the current AI summer that we currently find ourselves in, shows no signs of letting up in 2019. Organizations will continue to experiment with deep learning frameworks like TensorFlow, Caffe, Keras, PyTorch, and MXnet as they seek to monetize vast data sets.

(Connect world/Shutterstock)

Organizations will expand deep learning beyond its initial use cases, like computer vision and natural language processing (NLP), and find new and creative ways of implementing the powerful technology. Large financial institutions have already found that neural network algorithms are better at spotting fraud than “traditional” machine learning approaches, and the exploration into new use cases will continue in 2019.

This will also prop up demand for GPUs, which are the favored processors for training deep learning models. It’s unclear if new processor types, including ASICs, TPUs, and FPGAs, will become available. But there’s clearly demand for faster training and inference too.

However, the deep learning ecosystem will remain relatively young, and a lack of generalized platforms will keep this the realm of true experts.

7. ‘Special K’ Expands Footprint

Software needs something to run on. Operating systems used to provide that common substrate, but now developers are targeting something a bit lower: Kubernetes.

Developed by Google to manage and orchestrate virtualized Linux containers in the cloud, Kubernetes has become one of the hottest technologies in the big data ecosystem, if not the IT industry as a whole. As multi-cloud and hybrid deployments become more common, Kubernetes is the glue that holds it all together.

Big data software vendors that used to write their software to run on Hadoop are now writing it to run on Kubernetes, which at least gets them in the front door (if not an invite to dinner). Supporting Kubernetes software has become the number one requirement for software vendors — including the Hadoop vendors too.

8. Clouds Hard to Ignore

The cloud is big, and getting bigger. In 2018, the three biggest public cloud vendors grew at a rate approaching 50%. With an array of big data tools and technology – not to mention cheap storage for housing all that data – it will be hard to resist the allure of the cloud.

(Valery Brozhinsky/Shutterstock)

In 2019, small businesses and startups will gravitate to the major public cloud providers, which are investing majors sums in building ready-to-run big data platforms, replete with automated machine learning, analytical databases, and real-time streaming analytics.

Bigger companies will also find the cloud hard to resist in 2019, even if the economics aren’t nearly so attractive. However, the looming threat of lock-in will keep bigger companies wary of putting all their eggs in a single cloud basket.

9. New Tech Will Emerge

Many of the major big data frameworks and databases that are driving innovation today were created by the Web giants in Silicon Valley, and released as open source. The good news is there’s no sign the well is drying up. If anything, innovation may be accelerating.

(Kentoh/Shutterstock)

In 2019, big data practitioners would do well to retain as much flexibility as possible in their creations. While it may be tempting to cement your application to a certain technology for performance reasons, that could come back to haunt you when something better and faster comes along.

As much as you can, seek to keep your applications “loosely coupled but tightly integrated,” because you’ll eventually have to tear it apart and rebuild it.

10. Smart Things Everywhere

It’s tempting to dismiss the smart toaster as a cute gizmo that has no practical purpose in our lives. But perhaps it’s something less sinister: a prelude to an always-on world where smart devices are constantly collecting data and adapting to our conditions.

(elenabsl/Shutterstock)

Driven by consumer demand, smart devices are proliferating at an astounding rate. Smart device ecosystems are springing up around the two leading platforms, Amazon Alexa and Google Assistant, providing consumers with the opportunity to infuse remote access and AI smarts into everything from lighting and HVAC systems to locks and home appliances.

Buoyed by the rollout of super-fast 5G wireless networks, what’s happening in the home will soon happen in the world at large. Consumers will be able to interact with a multitude of devices, providing new levels of personalization everywhere we go.

In 2019, progress will be made across a multitude of fronts. Yes, there are substantial technical, legal, and ethical hurdles presented by big data and AI, but the potential benefits are too great to ignore.

SOURCE : https://www.datanami.com/2019/01/21/10-big-data-trends-to-watch-in-2019/

0 Comments

Can Artificial Intelligence Be Biased?

By TheWAY - 1월 25, 2019

Can Artificial Intelligence Be Biased?

Bias?SHUTTERSTOCK ENHANCED BY COGWORLD

Introduction

In pursuit of automation-driven efficiencies, the rapidly evolving artificial intelligence (AI) tools and techniques (such as neural networks, machine-learning, predictive analytics, speech recognition, natural-language processing and more) are now routinely used across nations: its governments, industries, organizations and academia (NGIOA) for navigation, translation, behavior modeling, robotic control, risk management, security, decision making and many other applications.

As AI is becoming democratized, these evolving intelligent algorithms are now rapidly becoming prevalent in most, if not all, aspects of human and machine decision-making. While Decision Utilities like intelligent algorithms have been in use for many years, there are rising concerns about the general lack of algorithmic understanding, usage practices, the rapidly penetrating bias in automated decisions, and the lack of transparency and accountability. As a result, ensuring integrity, transparency and trust in algorithmic decision-making is becoming a complex challenge for the creators of algorithms with huge implications for the future of society.

Human versus Machine Decision-Making Processes

Irrespective of cyberspace, geospace or space (CGS), since technology revolutions are driven not just by accidental discovery but also by societal needs, the question we all individually and collectively need to first and foremost evaluate is whether there really is a need for decision-making algorithms—and if yes, where and why. Furthermore, what gaps in the human decision-making process can be filled by decision-making algorithms?

Artificial intelligence tools and techniques are increasingly expanding and enriching decision support not only by coordinating diverse data sources delivery in a timely and efficient manner but also by analyzing evolving data sources, trends, providing defined forecasts, developing data consistency, quantifying uncertainty of all data variables, anticipating the human or machine user’s data needs, providing information to the human or machine user in the most appropriate forms, and suggesting decisive courses of all possible action based on the intelligence gathered. Understandably, this is being welcomed—since in a fast-changing digital age environment, it is becoming difficult for human decision-makers to keep up, analyze the mountains of growing data in front of them, and make informed and intelligent decisions.

However, even for an algorithmic decision-making process, there are complex challenges to reach an informed decision. For example, it is difficult to know whether decision-making algorithms will be able to make effective decisions with the current computing and data analytics infrastructure and processing capability. While it seems very likely that artificial intelligence will become universal in most, if not all, aspects of decision-making in the near future, it will be interesting to see how the emerging competition between human decision-making versus AI decision-making will play out.

Algorithmic Engineering Process and Penetration of Bias

While there are growing concerns about machine learning decision-making models, it seems AI is being woven into the very fabric of human society and everything individuals and entities do across nations: its government, industries, organizations and academia (NGIOA) in cyberspace, geospace and space (CGS).

Moreover, it needs to be understood that the rapidly evolving machine-learning model is not a static piece of code since we are constantly feeding it data from diverse sources and are constantly training, re-training and fine-tuning it to how predictions can be made. In each of these data journey steps, humans at the moment play a significant and influential role. As a result, while machine-learning models are becoming almost like a living, breathing thing with a growing dynamic data ecosystem from CGS around it, the very involvement of humans brings with it the same complex human bias. Since, we are trying to re-define and re-design systems that brings us more trust and transparency, there is a clear need to promote equality, transparency and accountability in algorithm design and development for decision-making—and to ensure that data transparency, training, review and remediation are being considered throughout the entire algorithmic engineering process.

According to ProPublica, an investigative journalism organization, a computer program used by US courts across the nation has been reported to be biased against black prisoners. The program, named the Correctional Offender Management Profiling for Alternative Sanctions, mistakenly flagged black defendants as likely to reoffend at almost twice the rate as white defendants (45% to 24%). The program likely factored in the higher rates of arrest for black people into its predictions but was not able to escape the same racial biases that contributed to those higher levels of arrests. Bias has also been reported in granting credit to home buyers, even going as far as to potentially violate the Fair Housing Act. Rates of defaulting may be higher in some neighborhoods, but an algorithm using this information to make black and white calls runs the risk of heading towards “red-lining” territory. Examples abound, with plenty of cases to show AI and technology to be both sexist and racist. Let’s not forget Google’s search algorithm including black people in the results of a search on “gorilla.”

While decision-making algorithms are inherently not biased and algorithmic decision-making depends on a number of variables -- including how the software is designed, developed, deployed and the quality, integrity and representativeness of the underlying data sources -- there is a need for a new approach to define and design decision-making algorithms. We perhaps need adaptive computing that integrates intelligence gathering into its very fabric and which does not rely on humans training the algorithms in how to make decisions.

Since it is important to evaluate what the implications will be if bias penetrates decision-making algorithms—it brings us to evaluating further whether data protection safeguards can be built into the algorithms from the earliest stages of development to prevent bias from penetrating them. It is important to address this as the very foundations of the systems that are being re-defined and re-designed depend on it.

Now, since it is not possible to interrogate algorithms, and there are no effective rules or regulations around decision-making algorithms that focus on the algorithmic accountability, how to remove bias remains a complex challenge facing society. Acknowledging this emerging reality, Risk Group initiated the much-needed discussion on Algorithmic Decision-Making with Prof. (Dr.) Steve Omohundro, President at Possibility Research.

Disclosure: Risk Group LLC is my company.

Risk Group discusses Algorithmic Decision- Making with Prof. (Dr.) Steve Omohundro, President at Possibility Research, based in CA, United States.

Perhaps the key to making decision-making algorithms work for everyone on a fair and balanced playing field is to build in accountability, responsibility, neutrality and outcomes from the very beginning—right in the code. If not, without question, all the efforts that are being put in re-defining and re-designing the systems in cyberspace, geospace and space will bring no real value for society overall.

Data: Nature, Sources, Efficiency

The growing availability, volume and accumulation of diverse sources of data means it can be overwhelming for any human decision maker to effectively make decisions—irrespective of whether these are strategic decisions or tactical. Therefore, it is important to evaluate what role dynamic growing data plays in how the algorithms are structured to take into consideration the growing data input. Further, it's important to evaluate, irrespective of the nature or source of data, whether decision-making algorithms give consistent decisions using different scenarios, models, tools and techniques? And how is this tested?

So, while efficiency seems to be at the core of many emerging automation applications, the transparency and integrity of the data on which the algorithmic decisions are being made will be critical to ensure its accountability. That brings us to another question--is there a way algorithms can detect data sources, credibility, authenticity, transparency and rate the integrity of the algorithmic decision itself? An algorithmic disclaimer?

The democratization of computing infrastructure allows anyone to build any algorithm any way they want. However, when it comes to its decision-making applications for systems at all levels: global, national or local (may it be government agencies, banks, credit agencies, courts, prisons, education institutions etc.,) there is a need for a global standard on the best practices to define and determine whose algorithm can be used for equality, fairness and objectivity. That brings us to another question of who will be testing the different versions of algorithms and rating them for public use?

The question now is not just whether humans or machines will make decisions—rather it's about whether intelligent algorithms replacing humans in decision-making will bring with them the same biases of race, religion, class, gender and ideology that are harmful to society. So, the question remains, what do we do about it? And will we ever be able to build truly objective and transparent machines?

What Next?

The question today for each one of us to individually and collectively evaluate is whether intelligent algorithms are and will remain an aid to the human decision-making process or whether they will become the ultimate decision-makers. And if artificial intelligence becomes the decision-maker, what will be the implications of relinquishing decision-making control to intelligent machines--as the very use of automated AI based decision-making techniques raises challenges for humanity as a whole.

SOURCE : https://www.forbes.com/sites/cognitiveworld/2019/01/20/can-artificial-intelligence-be-biased/#6f59f1557e7c

0 Comments

Yes, artificial intelligence can be racist

By TheWAY - 1월 24, 2019

Yes, artificial intelligence can be racist

Alexandria Ocasio-Cortez says AI can be biased. She’s right.

Getty Images/EyeEm

Open up the photo app on your phone and search “dog,” and all the pictures you have of dogs will come up. This was no easy feat. Your phone knows what a dog “looks” like.

This modern-day marvel is the result of machine learning, a form of artificial intelligence. Programs like this comb through millions of pieces of data and make correlations and predictions about the world. Their appeal is immense: Machines can use cold, hard data to make decisions that are sometimes more accurate than a human’s.

But machine learning has a dark side. If not used properly, it can make decisions that perpetuate the racial biases that exist in society. It’s not because the computers are racist. It’s because they learn by looking at the world as the way it is, not as it ought to be.

Recently, newly elected Rep. Alexandria Ocasio-Cortez (D-NY) made this point in a discussionat a Martin Luther King Jr. Day event in New York City.

“Algorithms are still made by human beings, and those algorithms are still pegged to basic human assumptions,” she told writer Ta-Nehisi Coates at the annual MLK Now event. “They’re just automated assumptions. And if you don’t fix the bias, then you are just automating the bias.”

The next day, the conservative website the Daily Wire derided the comments.

But Ocasio-Cortez is right, and it’s worth reflecting on why.

If we’re not careful, AI will perpetuate bias in our world. Computers learn how to be racist, sexist, and prejudiced in a similar way that a child does, as computer scientist Aylin Caliskan, now at George Washington University, told me in a 2017 interview. The computers learn from their creators — us.

“Many people think machines are not biased,” Caliskan, who was at Princeton at the time, said. “But machines are trained on human data. And humans are biased.”

We think artificial intelligence is impartial. Often, it’s not.

Nearly all new consumer technologies use machine learning in some way. Take Google Translate: No person instructed the software to learn how to translate Greek to French and then to English. It combed through countless reams of text and learned on its own. In other cases, machine learning programs make predictions about which résumés are likely to yield successful job candidates, or how a patient will respond to a particular drug.

Machine learning is a program that sifts through billions of data points to solve problems (such as “can you identify the animal in the photo”), but it doesn’t always make clear how it has solved the problem. And it’s increasingly clear these programs can develop biases and stereotypes without us noticing.

In 2016, ProPublica published an investigation on a machine learning program that courts use to predict who is likely to commit another crime after being booked. The reporters found that the software rated black people at a higher risk than whites.

“Scores like this — known as risk assessments — are increasingly common in courtrooms across the nation,” ProPublica explained. “They are used to inform decisions about who can be set free at every stage of the criminal justice system, from assigning bond amounts … to even more fundamental decisions about defendants’ freedom.”

The program learned about who is most likely to end up in jail from real-world incarceration data. And historically, the real-world criminal justice system has been unfair to black Americans.

This story reveals a deep irony about machine learning. The appeal of these systems is they can make impartial decisions, free of human bias. “If computers could accurately predict which defendants were likely to commit new crimes, the criminal justice system could be fairer and more selective about who is incarcerated and for how long,” ProPublica wrote.

But what happened was that machine learning programs perpetuated our biases on a large scale. So instead of a judge being prejudiced against African Americans, it was a robot.

Other cases are more ambiguous. In China, researchers paired facial recognition technology with machine learning to look at driver’s license photos and predict who is a criminal. It purported to have an accuracy of 89.5 percent.

Many experts were extremely skeptical of the findings. Which facial features were this program picking up on for the analysis? Was it the physical features of certain ethnic groups that are discriminated against in the justice system? Is it picking up on the signs of a low-socioeconomic upbringing that may leave lasting impressions on our faces?

It can be hard to know. (Scarier: There’s one startup called Faception that claims it can detect terrorists or pedophiles just by looking at faces.)

“You got the algorithms which are super powerful, but just as important is what kind of data you feed the algorithms to teach them to discriminate,” Princeton psychologist and facial perception expert Alexander Todorov told me in a 2017 interview, while discussing a controversial paper on using machine learning to predict sexual orientation from faces. “If you feed it with crap, it will spit out crap in the end.”

It’s stories like the ProPublica investigation that led Caliskan to research this problem. As a female computer scientist who was routinely the only woman in her graduate school classes, she’s sensitive to this subject.

She has seen bias creep into machine learning in often subtle ways — for instance, in Google Translate.

Turkish, one of her native languages, has no gender pronouns. But when she uses Google Translate on Turkish phrases, it “always ends up as ‘he’s a doctor’ in a gendered language,” she said. The Turkish sentence didn’t say whether the doctor was male or female. The computer just assumed if you’re talking about a doctor, it’s a man.

How robots learn implicit bias

In 2017, Caliskan and colleagues published a paper in Science that finds as a computer teaches itself English, it becomes prejudiced against black Americans and women.

Basically, they used a common machine learning program to crawl through the internet, look at 840 billion words, and teach itself the definitions of those words. The program accomplishes this by looking for how often certain words appear in the same sentence. Take the word “bottle.” The computer begins to understand what the word means by noticing it occurs more frequently alongside the word “container,” and near words that connote liquids like “water” or “milk.”

This idea to teach robots English actually comes from cognitive science and its understanding of how children learn language. How frequently two words appear together is the first clue we get in deciphering their meaning.

Once the computer amassed its vocabulary, Caliskan ran it through a version of the implicit association test.

In humans, the IAT is meant to undercover subtle biases in the brain by seeing how long it takes people to associate words. A person might quickly connect the words “male” and “engineer.” But if a person lags on associating “woman” and “engineer,” it’s a demonstration that those two terms are not closely associated in the mind, implying bias.

Here, instead of looking at the lag time, Caliskan looked at how closely the computer thought two terms were related. She found that African-American names in the program were less associated with the word “pleasant” than white names. And female names were more associated with words relating to family than male names. (There are some reliability issues with the IAT in humans, which you can read about here. In a weird way, the IAT might be better suited to use on computer programs than for humans, because humans answer its questions inconsistently while a computer will yield the same answer every single time.)

Like a child, a computer builds its vocabulary through how often terms appear together. On the internet, African-American names are more likely to be surrounded by words that connote unpleasantness. That’s not because African Americans are unpleasant. It’s because people on the internet say awful things. And it leaves an impression on our young AI.

This is as big as a problem as you think.

The consequences of racist, sexist AI

Increasingly, Caliskan said, job recruiters are relying on machine learning programs to take a first pass at résumés. And if left unchecked, the programs can learn and act on gender stereotypes in their decision-making.

“Let’s say a man is applying for a nurse position; he might be found less fit for that position if the machine is just making its own decisions,” she said. “And this might be the same for a woman applying for a software developer or programmer position. … Almost all of these programs are not open source, and we’re not able to see what’s exactly going on. So we have a big responsibility about trying to uncover if they are being unfair or biased.”

And that will be a challenge in the future. Already, AI is making its way into the health care system, helping doctors find the right course of treatment for their patients. (There’s early research on whether it can help predict mental health crises.)

But health data, too, is filled with historical bias. It’s long been known that women get surgery at lower rates than men. One reason is that women, as primary caregivers, have fewer people to take care of them post-surgery.

Might AI then recommend surgery at a lower rate for women? It’s something to watch out for.

So are these programs useless?

Inevitably, machine learning programs are going to encounter historical patterns that reflect racial or gender bias. And it can be hard to draw the line between what is bias and what is just a fact about the world.

Machine learning programs will pick up on the fact that most nurses throughout history have been women. They’ll realize most computer programmers are male. “We’re not suggesting you should remove this information,” Caliskan said. It might actually break the software completely.

Instead, Caliskan thinks there need to be more safeguards. Humans using these programs need to constantly ask, “Why am I getting these results?” and check the output of these programs for bias. They need to think hard on whether the data they are combing is reflective of historical prejudices. Caliskan admits the best practices of how to combat bias in AI are still being worked out. “It requires a long-term research agenda for computer scientists, ethicist, sociologists, and psychologists,” she said.

But at the very least, the people who use these programs should be aware of these problems, and not take for granted that a computer can produce a less biased result than a human.

And overall, it’s important to remember: AI learns about how the world has been. It picks up on status quo trends. It doesn’t know how the world ought to be. That’s up to humans to decide.

source : https://www.vox.com/science-and-health/2019/1/23/18194717/alexandria-ocasio-cortez-ai-bias

0 Comments

New AI computer vision system mimics how humans visualize and identify objects

By TheWAY - 1월 19, 2019

New AI computer vision system mimics how humans visualize and identify objects

A 'computer vision' system developed at UCLA can identify objects based on only partial glimpses, like by using these photo snippets of a motorcycle.

Credit: UCLA Samueli

Researchers from UCLA Samueli School of Engineering and Stanford have demonstrated a computer system that can discover and identify the real-world objects it "sees" based on the same method of visual learning that humans use.

The system is an advance in a type of technology called "computer vision," which enables computers to read and identify visual images. It is an important step toward general artificial intelligence systems -- computers that learn on their own, are intuitive, make decisions based on reasoning and interact with humans in a more human-like way. Although current AI computer vision systems are increasingly powerful and capable, they are task-specific, meaning their ability to identify what they see is limited by how much they have been trained and programmed by humans.

Even today's best computer vision systems cannot create a full picture of an object after seeing only certain parts of it -- and the systems can be fooled by viewing the object in an unfamiliar setting. Engineers are aiming to make computer systems with those abilities -- just like humans can understand that they are looking at a dog, even if the animal is hiding behind a chair and only the paws and tail are visible. Humans, of course, can also easily intuit where the dog's head and the rest of its body are, but that ability still eludes most artificial intelligence systems.

Current computer vision systems are not designed to learn on their own. They must be trained on exactly what to learn, usually by reviewing thousands of images in which the objects they are trying to identify are labeled for them.

Computers, of course, also cannot explain their rationale for determining what the object in a photo represents: AI-based systems do not build an internal picture or a common-sense model of learned objects the way humans do.

The engineers' new method, described in the Proceedings of the National Academy of Sciences, shows a way around these shortcomings.

The approach is made up of three broad steps. First, the system breaks up an image into small chunks, which the researchers call "viewlets." Second, the computer learns how these viewlets fit together to form the object in question. And finally, it looks at what other objects are in the surrounding area, and whether or not information about those objects is relevant to describing and identifying the primary object.

To help the new system "learn" more like humans, the engineers decided to immerse it in an internet replica of the environment humans live in.

"Fortunately, the internet provides two things that help a brain-inspired computer vision system learn the same way humans do," said Vwani Roychowdhury, a UCLA professor of electrical and computer engineering and the study's principal investigator. "One is a wealth of images and videos that depict the same types of objects. The second is that these objects are shown from many perspectives -- obscured, bird's eye, up-close -- and they are placed in different kinds of environments."

To develop the framework, the researchers drew insights from cognitive psychology and neuroscience.

"Starting as infants, we learn what something is because we see many examples of it, in many contexts," Roychowdhury said. "That contextual learning is a key feature of our brains, and it helps us build robust models of objects that are part of an integrated worldview where everything is functionally connected."

The researchers tested the system with about 9,000 images, each showing people and other objects. The platform was able to build a detailed model of the human body without external guidance and without the images being labeled.

The engineers ran similar tests using images of motorcycles, cars and airplanes. In all cases, their system performed better or at least as well as traditional computer vision systems that have been developed with many years of training.

source : https://www.sciencedaily.com/releases/2018/12/181220163210.htm

0 Comments