Emerrrrrging Tech in Simple Words: April Edition

How is AI used in mental health, and how close are we to creating emotionally intelligent algorithms? What types of AI are getting the most traction in business? Why do 80% of AI projects fail?

We've got the answers.

What is: Digital Patient Engagement

In the What is section, I help fellow entrepreneurs sift through the technology hype and translate buzzwords from geek English to plain English. This time, I'll focus on digital patient engagement solutions, whose power stretches far beyond telehealth and appointment booking.

In a nutshell, patient engagement can be described as a partnership between a patient and a healthcare provider, where the two parties agree to share information and seek the best treatment options possible in a transparent, mutually beneficial way. The concept is sonically different from the traditional, rigid approach that involves prescribing a treatment to a patient without actually discussing its advantages and pitfalls and leaving it up to the patient to decide whether to follow the plan or not.

Technology — namely, mobile applications, wearable-based remote patient monitoring solutions, telemedicine systems, patient portals, and appointment scheduling software with automatic reminders — digitalize patient engagement and encourage patients to assume a more prominent role in their treatment while partially taking the burden of administrative work off healthcare providers' shoulders.

One of our recent articles takes a closer look at off-the-shelf and custom-built digital patient engagement solutions and assesses the cost of implementing such systems in hospitals.

What's Hot: AI in Grocery Stores

In the What's Hot section, I zoom in on strategic trends and transformative technologies that will define 2021. For the April issue of our tech digest, I've chosen artificial intelligence and its applications in supermarkets.

Even though eCommerce sales grew by 18% in 2020, 85% of the US grocery sales still occurred in physical stores last year — not the least due to rapid digital transformations taking place in most brick-and-mortar locations across the country. And it looks like AI is playing a major role in this metamorphosis.

With the help of object and facial recognition technologies, retailers can minimize human interactions in stores, thus preventing the virus from spreading. Smart algorithms that analyze data from POS systems and mobile-based loyalty programs take cross-selling and upselling techniques to a totally new level. And AI-based forecasting systems can monitor inventory levels in real time to make sure you always have items in stock.

Check out our latest article to learn how artificial intelligence is shaping good old grocery stores into the supermarkets of the future.

What's New: a Rundown of Top Tech News from All over the Web

What's New is a collection of technology articles curated for you by the ITRex team.

What Does It Take to Create AI with Emotional Intelligence?

Mental health conditions affect at least one-fifth of the US population, and the pandemic and subsequent lockdowns have only made things worse. As psychologists and psychiatrists are struggling to meet the growing demand, many mental health providers are looking into artificial intelligence to expand access to professional care.

Ranging from intelligent bots that facilitate self-assessment and remote therapy sessions to expert systems analyzing data from EHRs, questionnaires, and even social media to assess the likelihood of developing a mental illness, AI-driven systems do have the potential to make mental therapy more accessible and affordable.

However, the researchers from Microsoft's Human Understanding and Empathy group doubt AI's future in the mental health realm, emphasizing the lack of emotional intelligence as a major obstacle data scientists still need to overcome.

The quest for algorithms that can sense, decipher, and mimic human emotions began in 1997 when the "affective computing" term was coined. Over the years, engineers have made significant progress in this field, completely eliminating the need for expensive hardware and applications with cumbersome interfaces. With the help of tiny wearable sensors, facial recognition technology, and natural language processing algorithms, researchers can now scavenge information about a person's physical health and mental well-being from various sources and feed this data to next-gen machine learning models. The question is, does this make artificial intelligence genuinely empathetic? Here's what the Microsoft team thinks.

How Has COVID-19 Changed Businesses' Approach to Disaster Recovery Planning?

When COVID-19 brought a great share of operations to a halt and sent whole countries into lockdown, we realized how ill-prepared businesses were for the new challenges, which spanned remote work, cybercrime, and IT talent shortages. A recent survey from Accenture unearthed some curious facts about the pandemic's lasting impact on business continuity.

Even though most companies successfully migrated their apps and data to the cloud and implemented IT tools supporting efficient communication outside the office walls, one-third of executives are planning to accelerate their digital transformation efforts, redesign processes, change the way they interact with customers, and, what's more important, upend their disaster recovery plans. Neal Weinberg of Network World explores what it takes to build a truly resilient business in times of uncertainty.

One-Third of Companies Now Use Artificial Intelligence. Which Technologies are Trending Up and Why?

Even though AI adoption across enterprises somewhat flattened last year, the latest report from IBM indicates that 74% of companies are currently using or exploring artificial intelligence, with one-third of the respondents relying on AI to mitigate the pandemic's aftermath.

Among the technologies that sparked the most interest within the enterprise segment are natural language processing, which is used by a whopping 42% of organizations to improve customer service and reduce operating costs, and AI-powered automation software, which has already been deployed by 61% of the participants. Not surprisingly, companies are investing in intelligent automation tools to make their employees more productive during the pandemic.

Check out the full article by VentureBeat to investigate the state of enterprise AI, compare the IBM survey findings to those of IDC and O'Reilly, and discover the key barriers companies face when implementing artificial intelligence.

What Can Crossword-Solving AI Teach Us About Language Usage?

Although AI's winning the American Crossword Puzzle Tournament wasn't the first time algorithms beat humans by leaps and bounds (remember how AlphaGo defeated Lee Segol, one of the world's best Go players?), the unlikely victory could throw some light on how our brain functions when we play with words and how scientists could boost AI's linguistic skills.

A joint effort between the Berkeley Natural Language Processing Group and Matt Ginsberg, a renowned computer scientist, Dr. Fill is a state-of-the-art AI program trained on thousands of crossword puzzles published in various media outlets. The application was deliberately designed as a closed system, meaning it can't just Google answers, which makes it similar to the imperfect human brain. When approaching a new task, Dr. Fill draws on its impressive database of eight million clues and answers to come up with a list of words that could match the query. In the next step, the AI program analyzes the "environment" (i.e., other answers and letters that have already been guessed correctly) to establish semantic connections and select the best possible candidates.

While the intelligent algorithm finished the competition at least two minutes earlier than human cruciverbalists, the researchers have a long way to go before they fully understand how the machine arrives at its conclusions. Their ultimate goal is to eventually teach Dr. Fill to connect bits of knowledge into a reasoning chain (aka "multi-hop inference"). Currently, few machine learning models can combine more than two facts meaningfully before drifting away from the topic — and that's something we need to tackle if we ever want to create machines with human-like reasoning.

Product Design Gets an AI Overhaul. What's the Deal?

In a world where it takes six to nine months to design a new product of moderate complexity (there's no speaking about a fancy car or novel CT scanner here!), industrial engineers are turning to technology to speed up the process and meet the ever-growing consumer demand. One of these game-changing innovations could be artificial intelligence.

With the help of AI-powered simulation software, major brands like Amazon and Renault significantly improve the user experience for design engineers while collecting all sorts of data that helps avoid costly mistakes later in the development process. In particular, Renault is leveraging artificial intelligence to create automated manual transmission (AMT) systems. AMTs feature electro-mechanical actuators, smart sensors for monitoring a vehicle's performance, and embedded software that controls the engine. Because of this complexity, it used to take engineers up to twelve months to fine-tune the system's performance. But once the French car manufacturer equipped engineers with Siemens' Simcenter Amesim software, they could simply drag and drop UI components on a visual board, analyze the relationship between different AMT components, and predict the behavior of the entire system without building a single physical prototype!

How to: Approach Your First AI Project and Keep AI Development Costs Down

In the How to section, the ITRex team provides practical, easily applicable tips to accelerate your company's digital transformation. 

Back in January, we published an article about AI implementation challenges, citing technology roadblocks, a lack of explainability, and resistance to change as the major factors hindering company-wide artificial intelligence deployments. At that time, we were toying with the idea of writing a separate piece dedicated solely to financial matters, so we deliberately excluded unclear ROI perspectives and the presumably high cost of AI development from that list.

One of the reasons why 80% of enterprise AI projects never deliver on their promise or fail outright lies in the C-Suite's misunderstanding of artificial intelligence capabilities — and the efforts and resources needed to transform a clunky prototype into a truly intelligent system providing a 360-degree view into a company's workflows.

Here at ITRex, we devised a custom approach that helps our clients build and deploy AI solutions faster and see an immediate return on their investments:

• Identify a business problem you can solve with artificial intelligence. At this stage, it is crucial to reach out to internal and external stakeholders and pinpoint processes and decision flows that can be enhanced or fully automated.
• Prioritize use cases. With the help of a product prioritization framework like Kano or MoSCoW, it would be a no-brainer for you to select business cases that promise the most value during the interim period and serve as a basis for further AI implementations.
• Select the optimum technology stack for your project. Combining off-the-shelf, open-source, and custom-made components like API-driven voice assistants or plug-and-play facial recognition engines is a surefire way to create a vendor-agnostic solution at a reasonable price.
• Focus on user experience. Your future AI system should incorporate a user-friendly interface that allows stakeholders (regardless of their technical background!) to ask artificial intelligence questions, get instant insights, or automate tasks in the most natural way.
• Prepare data for algorithmic analysis. To help AI interpret your business data, it is essential to gather information, assess its quantity and quality, and bring it into a unified format. For this, several data collection, preparation, and normalization techniques can be applied.
• Build an MVP version of your AI system. Starting with a minimum viable product that supports the essential use cases is one of AI development best practices, as it allows you to test the feasibility of your concept, pinpoint areas for improvement, and start scaling the system across different use cases and departments.
• Treat AI implementation as a work in progress. Once you put artificial intelligence to work, you may not get perfect results right from the onset. To deliver more accurate predictions and become fully autonomous, your AI system will need to consume new information under the supervision of human specialists. Therefore, it is important to continue gathering feedback from your company's stakeholders, making the necessary changes to the system, and repeating the steps enumerated above when introducing new features and use cases.

That's all for today. Make sure to follow ITRex Group on LinkedIn to keep up with the IT industry trends and get practical tips from AI, QA, cloud computing, and intelligent automation experts.

See you next month!

Cheers,

Vitali Likhadzed

ITRex Group CEO and Co-Founder