Putting Computer Vision into Focus

Just a few short years ago, computer vision was seen as something that held great potential but which wasn’t quite there yet. Today, thanks to advances in AI, more affordable GPU capacity and an accumulation of data and the means to train it, it’s becoming a strategic technology asset for companies in a variety of industries. In fact, according to a Forrester blog post , 58% of senior business purchase influencers said that their firms are implementing, planning to implement, or interested in implementing computer vision in the coming year.

This thinking was recently outlined in a Forrester research report entitled, All Enterprises Need (Computer) Vision, June 14, 2019 (access requires subscription or payment). In the report, Forrester describes the four most established use cases for computer vision, while also sharing the key questions to ask when planning your computer vision strategy. We at Wovenware were honored that we were cited as an example of a provider that creates labeled training datasets, which we consider the secret sauce to effective computer vision solutions.

While the common strategy of computer vision is to capture, process and analyze real world images and videos to uncover meaningful information, there are currently different ways to get there. Automated machine learning tools for creating computer vision apps are available that provide plug-and-play capabilities that make it possible for a basic programmer to pretty easily build a basic model. We’ve found, however, that these types of machine learning solutions might be a good way to begin, yet companies serious about leveraging computer vision for competitive advantage need custom solutions built from the ground up, around real-world images that are identified and labeled from scratch.

Custom deep learning-based computer vision solutions provide the deep-dive analysis that comes from understanding and honing in on unique business needs and specific customer behaviors. They leverage a deep understanding of customer behaviors and challenges – within specific industries – as well as experience into what resonates in certain markets, and how to transfer that knowledge to machines.

What’s important as well, for companies looking into implementing computer vision, is to consider that in order to be effective, it requires constant care and feeding – in the form of new data, images, video and other content. Computer vision can never be a one and done proposition.

Yet, as Forrester eloquently describes in the report, computer vision is allowing companies to collect unprecedented intelligence about the most important aspects of their businesses. It is enabling a whole new level of awareness, understanding and insights that can improve lives, making people safer, cities more efficient and health diagnoses more accurate. Consider the following examples of computer vision at work today.

  • Transforming advertising. Companies, such as Gannett, are turning to deep learning and computer vision to design better online ads, determining which colors, images and fonts work best. The company says that this has boosted click-through rates across different news sites.
  • Improving patient outcomes. Computer vision can help physicians diagnose diseases, among other applications. For example, a physician or radiologist can use it to review brain scans and determine healthy or not so healthy areas of the brain.
  • Enabling safer autonomous driving. Deep learning-enabled computer vision is being applied in autonomous driving to navigate roads and make quick decisions in real time, such as identifying an oncoming vehicle or slowing down on icy pavement.
  • Making shopping easier. In one example, cameras are being placed in the ceiling above aisles and on shelves in a brick & mortar retail location, and using computer vision technology these cameras can determine when an object is taken from a shelf and who has taken it. If an item is returned to the shelf, the system is also able to remove that item from a customer’s virtual basket. The network of cameras allows the app to track people in the store at all times, ensuring it bills the right items to the right shopper when they walk out, without having to use facial recognition.

Satellites Bring Computer Vision to a Whole New Level

When computer vision is deployed in satellites, its possibilities are boundless. Consider the following:

  • Tackling deforestation. Computer vision and deep learning can help detect the number or specific species of trees in certain forests and parks to determine their growth or risk, and if deforestation is occurring, it can help to address the specific factors that could be causing it.
  • Tracking economic growth. By monitoring the numbers of cars, electric lights in the night sky, construction, we can track the development and economic growth of countries around the world.
  • Responding to world crises. In situations such as a refugee crisis or war, satellite imagery can help provide valuable information that can be used to plan for the supply of life-sustaining resources like food and shelter materials.

Computer vision is gaining major traction in a variety of industries, providing an extra set of really smart eyes that can identify vulnerabilities to safety, identify anomalies in medical images and improve customer experience. Yet, companies need to take a strategic approach to its implementation, mapping out the most direct and efficient route to making the promise of computer vision a reality.

Using Artificial Intelligence (AI) to Gain Insight in the Financial Services Industry

Artificial Intelligence (AI) has been taking the business world by storm. It’s been used in applications from determining good locations for solar panels and increasing crop yield to predicting which medical devices might fail and which type 1 diabetes patients may be at particular risk. And financial services is no exception. The importance of providing an outstanding customer experience and predicting customer behavior are some of the key reasons that AI is gaining popularity in financial services. In fact, banks are investing heavily in AI technology, with anticipated spending of $5.6 billion in 2019 according to IDC.

One visible way that AI is impacting financial services is through chatbots. These programs, based on an area of AI called Natural Language Processing (NLP), can mimic human communication in speech or text. They are frequently used by contact centers or for customer service support on websites to answer basic questions, such as – “When is my mortgage payment due?” or “What is the balance on my account?” – or to engage with customers for upsell opportunities. They provide great value to financial services organizations by allowing them to staff customer support lines 24/7 – which is critical in today’s global marketplace, yet near impossible to accomplish logistically and economically by humans alone.

Improving the customer experience

Chatbots significantly help financial services organizations boost the customer experience by providing answers to customer questions quickly rather than making them wait 10, 20 minutes or more to speak to a person while listening to annoying Muzak or being shunted around to different people. Chatbots also can be used in combination with humans to provide better customer service. For example, when a chatbot recognizes a phone number or uses voice biometrics to identify a caller, it can direct a valued customer to a more senior representative for quicker, VIP treatment.

Taking it a step further, AI programs can also increase the value customers receive when they speak with service representatives. Based on an individual customer’s history and current situation as well as aggregate customer information from the AI algorithm, agents can make real-time recommendations on financial service products that might be most appropriate for the customer.

This is similar to the way Amazon, Netflix and others make recommendations on products or films based on past behavior and the collective data they aggregate from a large dataset of users. Providing these customized, relevant recommendations results in a better customer experience and greater satisfaction.

Making better customer decisions with Predictive AI

In addition to enhancing the customer experience, following are four more ways AI is being used to add value in the financial services arena:

  • Detecting fraud. Along with growth of online banking and investing comes new opportunities for fraud. Financial services firms are fighting back with AI programs that are designed to look for suspicious patterns of behavior that differ from the way specific customers typically transact. It also helps them determine which credit cards may likely be impacted if a breach has occurred, so firms can avoid the expense of re-issuing credit cards to all of their customers.
  • Predicting customer churn. Based on patterns in specific customer behavior and historic data, AI programs enable organizations to predict which customers might be likely to leave, so they can take actions or provide specific promotions to entice them to stay.
  • Developing risk models. There may be more accurate ways to determine credit worthiness than the traditional credit scores which are based on generic criteria. Now companies can look at specific customer information, evaluate different criteria and use AI algorithms to predict customers’ credit worthiness. For example, loan candidates who may not have a credit history accumulated because they are just entering the workforce or typically pay with cash, may be good credit risk because of other factors, such as debt-to-income ratio or the amount of years they have been employed.
  • Ensuring regulatory compliance. With data on a large number of regulations, AI programs can flag potential issues and help ensure that organizations are in compliance with them. This is critical for avoiding potential problems and hefty fines that could come with non-compliance. For example, AI programs can analyze call center conversations to ensure that representatives comply with privacy regulations, and they can flag suspicious financial behavior, such as cyberattacks or money laundering to ensure that they are safeguarding customer accounts and not inadvertently breaking any laws.

Challenges

According to a 2019 PWC survey, financial services firms expect AI to help them grow revenue and profits, improve the customer experience and develop innovative products. But there are challenges that need to be overcome. It’s difficult for financial services firms, often steeped in traditional processes, to implement new technologies which can involve retraining employees at multiple levels of the organization, establishing a new mindset and creating new ways of doing things. And, on the technical side, it can involve capturing, cleaning and aggregating data that is often poorly structured and may be walled off in silos.

It also requires the talents of data scientists to develop algorithms and continually fine-tune them so they are accurate and up-to-date, and data engineers to collect and prepare the critical data that is needed to run these programs. Financial services firms may lack employees with this expertise, and a current shortage of these skilled professionals makes it difficult to bring these services in house.
To address this, many companies are turning to outsourcers, and nearshorers in particular, to gain the strategic and technical expertise that is needed to develop custom AI programs that address their business needs.

The opportunities, business benefits and competitive edge that AI offers to financial services is enormous. And, in a fast-changing industry where insights can rapidly result in financial gains – or a lack thereof can lead to losses – innovators embracing this technology can achieve a critical advantage.

What the H-1B Visa Issue Means for Tech Advancement

New changes to the H-1B visa program this year may be spurring renewed concern about exactly how the U.S. can fill a critical skills gap.

On one hand, given the president’s Executive Order, “Buy American, Hire American,” the government is reforming the H-1B program to, among other things, give priority to only the most highly skilled and highest paid H-1B candidates. While bringing in qualified and experienced professionals may seem like a good thing, it’s still not enough to fill critical positions in the tech industry to meet the growing demand for advanced software. And, the government has added restrictions to the H-1B application process that make it harder for all skilled workers from other countries to obtain visas to work in the U.S.

Unfortunately, the tech industry is headed for a perfect storm when it comes to recruiting skilled tech professionals. Many foreign workers are not able to work for U.S. firms, despite the fact that the country is already facing a critical shortage of qualified workers.

One key area where demand for skilled workers just can’t be met is data science. Highly-skilled data scientists are needed in droves to develop and continually maintain algorithms that teach AI programs to recognize patterns and predict behaviors and other outcomes.

Part of the problem stems from the fact that there are just not enough programs to educate data scientists. According to the University of California, Riverside, only one-third of U.S. universities offer data scientist programs and most of them are offered to graduate students; only six offer these programs to undergraduates. It is estimated that by 2020 there will be 490,000 data scientist positions but only 200,000 skilled professionals to fill the openings.

Recent H-1B visa developments

The H-1B visa could be one way to help fill the skilled tech worker gap, yet recent regulations have added restrictions to the application process. The government is taking longer to approve the visas and denying more workers. The application acceptance rate has also been impacted, dropping 13 percent from the previous year (74 percent vs 87 percent), making it the lowest percentage in 10 years.

Other changes include a new Labor Condition Application (LCA) form that employers of H-1B applicants need to submit. Additionally, earlier last fall, the government had suspended premium processing for certain H-1B petitions which has been used to expedite processing when employees move to a different location in the same company or to a new company, but premium processing has since been reinstated.
All of these changes are creating an air of uncertainty and, for the most part, are making it harder for companies to get the specialized tech expertise they need.

The answer is close by: nearshoring

To address the worker shortage caused by a lack of skilled IT professionals – from the U.S. and overseas – many companies are turning to nearshore providers. Nearshorers offer the next best option to having employees on staff – close proximity to clients which facilitates strong communication and collaboration. The ability to work closely with their clients, and share a common language enables these providers to become immersed in and fully understand a client’s business environment – which is critical to developing and maintaining advanced IT solutions.

Most nearshore firms share the same work ethic and high industry standards as their mainland U.S. counterparts, and hire very skilled data scientists, many of whom have attended U.S. universities.

A great combination: H-1B visa program and nearshoring

Experienced IT professionals will continue to be in demand by all types of companies in all industries, as, for example, advances in AI enables them to gain critical insight, improve customer experience and outcomes and make better decisions. Yet, as the demand for data scientists (and other tech experts) continues to grow and greatly outpace supply, companies need to find innovative ways to meet their staffing needs. While the H-1B visas offer a good solution by bringing diverse perspectives to software development, companies need to supplement in-house talent with other approaches as well, given the limited numbers of visas that are being issued.

Nearshoring provides an ideal way to fill this need; it offers the best of both worlds – providing the geographic and cultural closeness that supports the in-depth communication and collaboration needed for high-level software development, while also offering broader perspectives from professionals from other geographical areas. By combining the H-1B visa program with services from a nearshore provider, companies will have a winning talent recruitment strategy that puts them on the fast track to success.

Caribbean American Heritage Month Recognizes Value of Diversity

This month marks the 13 Anniversary of National Caribbean American Heritage month and this year’s theme, Contributing to the American Landscape, couldn’t be more fitting.

Today, Caribbean Americans, as well as U.S. territories located in the Caribbean, are contributing greatly to the U.S. in many ways, through tourism, cultural diversity and increasingly through business expertise.  And, despite set-backs from the hurricane, Puerto Rico is no exception.

While tourism continues to be a major part of our economy, what is happening in the business offices, manufacturing plants, and board rooms across the island is a silent surge in business as the mainland increasingly realizes the expertise, efficiency and cost-savings that can be gained by working in and with Puerto Rico.

As mainland firms increasingly nearshore work – from IT development to manufacturing – they are greatly contributing to the growth of Puerto Rican firms, while filling much-needed roles caused by a growing shortage of tech talent on the mainland.

And, more than five million Puerto Ricans live on the mainland and continue to be proud of their heritage. Many successful state-side Puerto Ricans are committed to serving as mentors to Puerto Rico-based businesses and willing to invest in the island, as well as in its start-ups and firmly established businesses.

As we mark Caribbean American Heritage month, it’s important to note how far Puerto Rico has come in becoming  a prime destination for entrepreneurial businesses, providing talent, creativity and diversity to businesses – both on the island and the mainland. Through continued investment by mainland businesses, collaborative initiatives like nearshoring, and support of government in the form of tax incentives and other programs that spur growth, we can ensure that this Caribbean territory becomes a key contributor to industrial growth in the U.S.

Teaching Chatbots to Do the Right Thing

Along with the great opportunity that AI provides comes great responsibility. As we develop and train algorithms, Natural Language Processing and other forms of AI that chatbots and other applications use to augment – or in some instances, replace – the role of humans, we have a responsibility to make sure that we are creating ethical human stand-ins and that we are not inadvertently imbuing them with our own biases.

This was a topic I wrote about in a recent article on Forbes.com, “Expanding on Asimov’s Laws to Create Responsible Chatbots.” Creating ethical, responsible chatbots and other smart apps requires taking a very thoughtful approach to AI development. Part of that includes hiring a diverse team of experts involved in the development, training and maintenance of these programs, enabling these programs to be more sensitive to – and reflect and be respectful of – different cultures and ways of thinking. In addition to data scientists and data engineers, the team should include natural language experts, who can provide chatbot responses that are polite, correct and appropriate, and reflect the normal flow of conversation. In order for chatbots to respond accurately, programs must be built from the ground up in the language that they will be used. This will avoid any inadvertent errors or insensitivities that could result from poor translations. Additionally, companies need to continually monitor these programs to make sure they are not learning inappropriate language and other wrong things in the course of conversing with humans.

It’s critical that organizations are transparent when it comes to AI programs. While it may not always be obvious if someone is chatting with a bot online, it’s the company’s responsibility to inform users that they are conversing with software and not a human. And, companies should recognize that it’s not always appropriate to use chatbots. There are sensitive conversations, for example, such as discussing medical diagnoses, when people should speak with humans instead of chatbots.

Chatbots enable companies to the increase customer experience by being more attentive and responsive to their customers’ needs. By being respectful of these customers, treating them ethically and creating chatbots that will add to rather than detract from their experience, companies will be able to effectively harness the power and value that these AI programs offer.

Artificial Intelligence (AI) Glossary Of Terms

Artificial Intelligence (AI) is increasingly being used in a variety of industries to solve critical business challenges, but despite its growth, it’s still a nascent technology that continues to evolve. As such, there can be some confusion about AI terminology. To provide greater clarity, the following glossary, while not scientific or authoritative, provides a basic understanding of widely used terms and their definitions.

One term absent in this glossary is Robotic Process Automation (RPA) because it is not a type of AI, even though some people mistakenly categorize it as AI. RPA is a software robot that automates rote, repeatable tasks, such as insurance claims processing. The software is not learning, but rather repeating the same process over and over.

The AI Litmus Test

In order to be classified as AI, the software needs to incorporate some type of “intelligence” that is not explicitly programmed and have the capability to continually learn by recognizing patterns in the data.

The Glossary

AI Algorithm – A set of mathematical instructions that are able to learn from data. Data is run through the AI algorithm to provide the AI solution with the intelligence to look for and identify patterns to answer business questions. Data scientists train AI algorithms to predict specific outcomes.

Chatbot or Bot – A type of AI program that can mimic human communication in speech or text and is based on an area of AI called Natural Language Processing (NLP). Chatbots are often used in call centers or for customer service support on websites to answer basic questions or engage with customers.

Computer Vision – A field of AI that strives to gather insight from images or videos. It can be used for image classification, object detection and object tracking, among other applications.

Crowdsourcing Services – These services distribute, among many people, the work involved in collecting and preparing data, including image recognition, data normalization, and algorithm training for machine learning, among other tasks.

Data Engineers – Professionals who are trained to collect, prepare and manage the big data that will be used by AI programs.

Data Scientists – These highly skilled professionals design and continually train mathematical algorithms to be able to answer business questions or help predict outcomes. They follow the scientific method, formulating hypotheses and carrying out experiments to prove their assumptions, leading to new discoveries and insights.

Deep Learning – This highly cognitive field of AI encompasses the next generation of machine learning, where machines can teach themselves by making connections or comparisons in the data that are often overlooked or unknown. Deep learning software requires a huge amount of data and advanced processors to get accurate outcomes.

Image Recognition – The process of detecting and identifying a variety of objects in images and videos using computer vision.

Insights as a Service – A cloud-delivered, pre-built approach to AI, which allows a company to purchase previously sourced data — without owning the data or the AI solution – to help support a business case it is trying to solve. This approach is helpful if a company is looking for insights into a universal question that does not require personalized data.

Machine Learning – This is a form of AI that is capable of learning by recognizing patterns in identified variables in the training data.

Model as a Service – In this model companies collect their own data, but turn to a cloud-based provider to develop, re-train, maintain and operate the AI algorithm. This approach particularly makes sense if a company has very customized or proprietary information, but not the in-house expertise or resources to build or manage an AI solution. A churn model predicting which customers might leave a company is an example of this.

Natural Language Processing – An AI discipline that enables computers to process and understand human language.

Artificial Neural Networks – This is a component of an AI solution that is similar to biological neural networks. These networks use stages of learning to give AI the ability to solve complex problems by breaking them down into smaller problems.

Pragmatic AI – This is the type of AI that is most prevalent today. It applies AI principles to help answer business questions and predict outcomes for tasks, such as predicting the stock market or customer churn.

Pure AI – This type of AI is futuristic, with the ambitious goal of building human-like AI programs that have the same – or even superior – intelligence and capabilities as humans. For instance, this could refer to human-like robots that can make better decisions than humans and pass the Turing test.

Turing Test – A test developed by Alan Turing, to determine a machine’s ability to exhibit intelligent behavior equivalent or indistinguishable from a human.

Predictive Analytics – Based on a large amount of historical data, these AI programs are designed to predict future outcomes and behaviors based on past data.

Pre-packaged Data – Pre-packaged data sets that can be purchased. They offer a quick way to gather training data for AI algorithms.

Private Crowd – A private group of data specialists, usually under NDAs and personally known to the employer, who help with data collection, identification, labeling and preparation of training data or images. Private crowds provide greater accuracy and business-specific knowledge than more generic public crowds (or crowd sourcing).

Public Crowd – A group of non-professionals who can help collect, identify and label large data sets for AI training, but don’t have any knowledge of the company’s specific business. These people are usually not under NDA and not personally known to the employer.