Part of:

The Promises and Pitfalls of Machine Learning


While machine learning in healthcare and marketing has provided consumers with many benefits, the technology's use by police and other government agencies is still controversial.

More people would have died in the pandemic had it not been for Robovision’s international Imaging COVID-19 AI initiative. “Our algorithm helps radiologists accurately diagnose COVID-19,” Erik Ranschaert, co-founder of that machine learning (ML) program told me. "Physicians all over the world will use our deep learning platform to settle questions like: What’s the probability that this is COVID, should this patient be rushed to the ICU, and how deep is this lung tissue affected?”

Repeated in the retail, educational and hospitality industries, among others, it was our ML-trained robot counterparts that eased our lives, healed us, educated us, fed us, shopped for us and so forth. Machine learning certainly has its advantages. On the flip side, ML has its disadvantages, too.

What is Machine Learning?

Data scientists have this ambitious project of training machines to think like us. That includes the capabilities of detecting, predicting and recommending or prescribing solutions. Right now, more human-like abilities (like emotion or creativity) are beyond them, but in the past 40 years, or even in the last five years, machines have come an incredibly long way. (Read also: What is the Impact of AI on Art?)

Machine learning is done by scientists feeding hundreds of thousands of images of an item into the machine algorithms so that the algorithm recognizes the differences between that item and different items. Algorithms are tested and retrained until they pass their tests. These algorithms are then put to work in real life to achieve their purposes.

Machine learning is famously used in innovations like voice recognition systems, digital customer assistants (e.g., Alexa and chatbots), self-driving cars, credit card fraud detection systems and virus and spam detection software.

In addition, ML is proving itself to be an event-driven defence against cybercrime. ML first gains an understanding of what is normal (measured against a baseline) then actively detects and prevents attacks perpetrated by cybercriminals. It does this by actively blocking traffic from suspicious IP addresses, or preventing malicious activity against network files and maintaining confidentiality, integrity and availability of data.


Machine Learning Benefits

1. Machine Prediction

Machines are better and faster at analyzing huge data sets with complex variables than any human. In fact, researchers say predictive modeling techniques can detect illnesses better than doctors and determine the best treatment with 40 percent better patient outcomes.

Just from recording our voices or gait, machines can tell whether humans are more likely to die from cancer in the next few years or get neurological diseases like Alzheimer's, ALS and Parkinson's. Their predictions help doctors prevent illness and death.

When it comes to COVID-19, studies show ML algorithms like those deployed by Robovision predicted onset Covid with 95% accuracy.

2. Machine Detection

Machines outperform humans by detecting objects five percent more accurately than us – an ability that helps them help us across industries. Hailey Peng, former Marketing Manager at the technology company, DJI, told me how ML-trained drones choked forest fires in the German town of Hechingen, in August 2018, and saved 765 firefighters. Accurately analyzing the situation, the drones helped Hechingen’s fire chief dispatch crews faster to the scene with precise manpower, units and supplies.

“The biggest advantage came to light during the search for hotspots and extinguishing them,” Hechingen’s Fire Chief Commander Bulach later reports DJI: “The simultaneous deployment of the XT and X4S provided me with exact information about where to delete the hotspots and how long until we reached a safe state.”

3. Machine Prescription

Modern machines can analyze data and come up with solutions 100 times faster than any human. Using this ability, models help farmers and agronomists identify, prevent and treat problems such as soil erosion and irrigation issues, unhealthy plants, and livestock disease. It also helps them apply the precise amount of chemicals to fungible crops. (Read also: Adventures in AgroTech: 7 Can't-Miss Developments.)

In 2010, Vijay Bhaskar Reddy of India created a mobile device to help farmers monitor their pumps and save their fields from flooding. “Travelling 14 miles multiple times a day to just water the fields takes a lot of time and a lot of petrol.”

A farmer from Telangana’s Karimnagar district told me, “Since I started using this KisanRaja app about five years back, I no longer have to guess when to switch on the pumps and go back and forth to my home and wake up late at night to go in the fields among the snakes and switch on the pumps. I can turn them on from wherever I am, even driving my daughter to school.”

ML is used across industries in countless ways. These include:

  • Real-Time Recommendations – Business websites create individualized recommendations for customers based on their browsing activity, among other factors.

  • Virus and Spam Protection – Virus and spam detection software uses AI to detect new types of virus and spam, without relying on virus signature detection.

  • Automated Stock Trading – Robo-advisers crunch millions of data points and recommend which stock to invest in.

  • Household Robots – These are robots that help you with all aspects of your house (and business), from making beds to cooking and cleaning.

  • Autopilot Technology – Advanced AI equips self-driving cars with cameras, radar, and LiDAR sensors. Also, drones to monitor oil pipelines, sensors to monitor smart factories, smart parking where drivers use apps to find available parking lots, and sensor-equipped waste and recycling stations that communicate real-time to streamline waste management operations.

Machine Learning Disadvantages

For all its benefits, machine learning has its shortfalls.

1. Data Privacy

Around 30 states allow law enforcement to run database searches of driver’s license and ID photos. The Los Angeles Police Force drive “smart cars” that stop passers-by at random, while San Diego allows law enforcement officers from nearly 25 agencies to stop people on the street and use their mobile phones to photograph them. Altogether, at least five major police departments – including agencies in Chicago, Dallas, and Los Angeles – use surveillance cameras to track pedestrians as they stroll the streets.

Companies like AnyVision sell cameras for reasons that include school surveillance, while facial recognition software, Churchix, helps clergy monitor how often members attend the services.

“What’s so scary,” Jameson Spivack, a Policy Associate at the Center on Privacy & Technology, Georgetown Law, told me, “is that, for the most part, there are no regulations for how law enforcement and other parts of the government, (or companies) use face recognition technology. And the public has no idea what this entails. This is a powerful, error-prone, biased surveillance technology operating completely outside the public view.”

…for the most part, there are no regulations for how law enforcement and other parts of the government, (or companies) use face recognition technology. And the public has no idea what this entails.

2. Data Security

Companies like Google and Facebook mostly keep their operations secret – but our data is accessible to their use. (Read also: Encrypted Messenger Apps: Are Any Actually Safe?)

Back in 2019, Facebook was repeatedly pulled to the curb for failing to keep its user privacy promises. In April, half a billion user records were found exposed on Amazon Cloud. Around that time, journalists reported that Facebook had harvested users' Android data for five years, since 2015. Four months later, investigators found Facebook paid hundreds of contractors to transcribe users’ audio, where contractors frequently listened to users’ voice messages. This opens a whole other discussion, bringing into question such things as Data Consent, Use Purpose and concerns around compliance with the GDPR as well as other Privacy Acts. (Read also: Data Breach Notification.)

3. Other Ethical Issues

Data control is only one of the various ML issues. Other ethical concerns include:

  • Bias in decision-making Computer models cannot think objectively because they're programmed by subjective humans. Results lead to bias in areas such as criminal justice, healthcare, financial services such as approval for Mortgage & Loan applications and hiring.

  • Reasoning constraints – Certain human cognitive processes are beyond the machine's capacity. These include imagination, the ability to ask questions and contextualize. So even if you have robot therapists and robot priests, don't expect them to empathize with you just yet or to be able to make distinctions between one human and another.

  • Machine morality – We’d like to expect robots not to harm. Where do we draw the line as things previously thought to be the realm of science fiction become closer to possible? For example, what will happen if individuals digitally download their physical minds into a computer, hoping to live forever through a computer simulation – or where other people morph robot parts into their physical bodies to exceed their potential? The possibilities of AI can be scary.

Bottom Line

Machine learning has its undisputed benefits.

“We are confident,” Robovision’s founder and director Jonathan Berte told me, “that our AI tool will also be useful after coronavirus when we will have to face mutations of the virus … We will always need AI assistance to detect these mutations and act faster, so we’ll never have a COVID again.”

On the other hand, ML can be abused too with ethical violations like multiple examples of data invasion as detailed by Business Insider.

For good or for bad, ML technology patrols our daily lives – and it doesn't look like this is stopping anytime soon.


Related Reading

Related Terms

Kaushik Pal
Technology writer

Kaushik is a technical architect and software consultant with over 23 years of experience in software analysis, development, architecture, design, testing and training. He has an interest in new technologies and areas of innovation. He focuses on web architecture, web technologies, Java/J2EE, open source software, WebRTC, big data and semantic technologies. He has demonstrated expertise in requirements analysis, architectural design and implementation, technical use cases and software development. His experience has covered various industries such as insurance, banking, airlines, shipping, document management and product development, etc. He has worked on a wide range of technologies ranging from large scale (IBM…