It’s 2016.

Experts say it would be at least ten more years until AI could beat a world-class human player at Go. This ancient game has 10¹⁷⁰ possible combinations, more than the number of atoms in the known universe and is a googol — the digit 1 followed by a hundred zeros — times more complex than chess.

On March 9 of that year, history was made when AlphaGo, a computer program from the company DeepMind, beat the 18-time Go World Champion Lee Sedol in the first game of a five-game series. This incredibly advanced program used search trees and deep neural networks both to select the next move and predict the winner of the game based on the current state of the board. A few days later, AlphaGo had won four matches to one.

Around the world, 200 million people watched in surprise as the computer program went against conventional wisdom to play the now-renowned move 37, one of many incredibly creative winning moves AlphaGo played during the series, leaving even Lee Sedol amazed:

“I thought AlphaGo was based on probability calculation and that it was merely a machine. But when I saw this move, I changed my mind. Surely, AlphaGo is creative.”

Yet, just over a year later, DeepMind released AlphaZero, “a single system that taught itself from scratch how to master the games of chess, shogi, and Go, beating a world-champion program in each case.”

We’re at an inflection point in the progress of AI, where it shifts from being able to accomplish a narrow range of tasks — say, playing either chess or Go — to an increasingly broad skillset that will truly change the world. What AI’s been able to do up until now is only the beginning.

The “Landscape of Human Competence” by Hans Moravec. Creative tasks and original work are highest.

If we place most human tasks on terrain, elevated to represent the difficulty for computers, the capabilities of AI are a rising sea level as pictured in the image above. Already submerged are several low-lying plateaus while the water laps at the base of the foothills of driving, investment, and translation.

Even so, before we can really start talking about AI, we first have to define intelligence. In his book Life 3.0, physicist and MIT professor Max Tegmark defines intelligence as not uniquely human and the “ability to accomplish complex goals”.

Just as it would be futile to try and draw a line to classify intelligence as an all-or-nothing trait, we must approach intelligence as a spectrum that is determined by the degree of ability in accomplishing different goals — a system that can translate from one language to another in all its intricacies and subtleties is more intelligent that one that can play chess.

Read on for three major industries that will be completely transformed by AI in the coming decades and how exactly we could get there.

Photo by Javier Allegue Barros on Unsplash

Healthcare

Nine out of the ten leading causes of death in the United States occur from preventable causes such as heart disease, diabetes, or cancer. Globally, 43 million people are affected by medical errors and there is a pressing need for more advanced healthcare services in low-income countries, especially in the context of a global pandemic.

With AI, we’ll be able gain unique and so-far-unseen insights into many different diseases and conditions with big data, diagnose more accurately with machine learning, and use personalized medicine to tailor treatment to the individual person along with a faster and more fruitful drug development process. What’s more, integrating with smart wearable devices will allow for real-time monitoring of health, alerting a diabetic if their blood-sugar levels surpass a certain threshold for example.

Already, lots of progress has been made. At Harvard University’s teaching hospital, doctors are using AI-enhanced microscopes to scan for harmful bacterias like E. coli in blood samples quicker than manual scanning with an astounding 95% accuracy.

Lung cancer results in over 1.7 million deaths per year, making it the deadliest of all cancers worldwide and the sixth most common cause of death globally. Unfortunately, it has the worst survival rate among all cancers as it is usually caught far too late when treatment is less successful. In 2019, Google Health published research that showed that by using an artificial intelligence model they could detect 5% more cancer cases while reducing false-positives by more than 11% compared to radiologists.

In applying AI to breast cancer detection, pathologists were able to halve the average time they needed to spend to find small metastases in the lymph nodes. Moreover, the program was able to correctly distinguish a slide with metastatic cancer from a slide without cancer 99% of the time while being able to “accurately pinpoint the location of both cancers and other suspicious regions within each slide, some of which were too small to be consistently detected by pathologists” (source).

Just recently, in November 2020, DeepMind revealed that their AlphaFold 2 AI system has effectively solved the protein folding problem — a challenge in biology that goes back 50 years. Any one protein is estimated to have 10³⁰⁰ possible conformations, meaning that it would take millions of years to model all the possibilities. Using deep learning and almost 200,000 proteins whose structure is already known, AlphaFold 2 could then compare its predictions with researchers by modelling proteins that scientists are still working to determine their structure.

On a scale of 1–100, the AI achieved a remarkable median score of 92.4, considered on par with a team of human researchers, although significantly faster — days versus years. Ultimately this will lead to faster drug discovery and even specialized protein design. DeepMind’s chief executive, Demis Hassabis remarks on AlphaFold 2’s success:

“I do think it’s the most significant thing we’ve done, in terms of real-world impact.”

Clearly, there’s a lot of incredible development happening at the intersection of AI and healthcare which offers plenty of hope for a healthier future and millions of lives saved every year.

However, there are some challenges and questions that need to be addressed before AI can reach its full potential. For one, since machine learning models require large amounts of patient data to effectively train on, hospitals and government regulators alike need to tackle the issue of data privacy and ownership.

As a patient, if you receive care, are you able opt-out of having your medical record and health information uploaded to a database that can be accessed by AI models? What about if your genetic information puts you at an elevated risk for developing certain conditions — is your insurance company entitled to know about it? If so, is it ethical for them to raise your premiums?

Another sobering challenge is the issue of responsibility; if a hospital uses an AI program to help diagnose you or an algorithm to determine your treatment, perhaps the first time that it goes wrong is at that moment — and with malpractice, where does the responsibility lie?

These questions need to be answered for AI to enter the mainstream healthcare systems of the world and while we won’t discuss potential solutions here (I’ll be writing about some of those in another article), it’s important that we understand both the promise and the risk that AI holds.

Photo by Sharon McCutcheon on Unsplash

Finance

The stock market aside, there are many opportunities for AI to change the way we interact with and manage money.

Whether to promote good credit or more accurately assess under-served minority communities, tools to manage risk and detect fraud for larger companies, or to offer personalized banking to an increasingly online world, there will be a shift in the coming years toward an interconnected and accessible financial system brought about in part by AI.

Using machine learning to evaluate borrowers with little to no credit information or history, Zest AI reduces loses while more accurately predicting risk. By adopting their product, banks were able to give out more loans while reducing the default rate by more than 30% — benefiting both parties and the economy as a whole.

Kavout, a startup founded in 2015, uses AI to identify real-time patterns in financial markets and condenses massive amounts of unstructured data into a numerical rank for stocks. Its top-ranking stocks have outperformed the S&P 500 by almost double over the last five years.

The traditional banking experience is very impersonal and new tools such as financial advice chatbots are using AI to create a better and more personalized customer experience. Money-saving assistant Trim cancels money-wasting subscriptions, finds more cost-effective services such as insurance or cell phone plans, and even negotiates bills — saving the average person almost $1,500 a year.

Of course, with so much of our day-to-day financial activity happening online, fraud detection is often a time-consuming process for financial institutions. Shape Security protects more accounts from fraud than all other financial security firms in the world combined. The software used by most of the largest banks in the US was trained on billions of interactions, allowing it to distinguish between real customers and bots using machine learning.

While the work that these companies — among others — have done to date is extraordinary considering the strict regulation in this space, they are just part of the transition to what the financial system will look like in the future. Centralized banks and financial institutions will gradually become less prominent as loans and even credit will become partially crowd-sourced or on the blockchain (more about this soon). Kiva is a great example, connecting over 3.5 million borrowers in 77 different countries with 1.9 million lenders, resulting in over $1.5 billion in zero-interest loans to date.

Photo by Radek Kilijanek on Unsplash

Transportation

In the US alone, more than 38,800 people die every year in car accidents.

Another 4.4 million are injured seriously enough to require medical attention.

That’s $871 billion a year, according to the NHTSA.

Solving the issue of self-driving cars is a moral imperative. In creating a fully-fledged autonomous driving system that will reduce accidents on the road, AI has a crucial part to play — being able to take in data from the car’s surroundings and analyze it in real-time to determine if a situation requires a certain response or to keep cruising along. What’s more, the average person spends 54 hours wasted in traffic every year — what if that time could be used for something more productive?

To put things in perspective however, humans are really good at driving when it comes down to it. In 2018, there were about 1.22 deaths per 100 million miles driven, roughly three-quarters of the distance from the Earth to the Sun. For an AI-driven system to have a meaningful impact on the number of deaths caused per year by car accidents, it has to be even better. As self-driving car pioneer Sebastian Thrun put it:

“To [build a self-driving car] that manages 90% of the problems encountered in everyday driving can literally be done over a weekend, to do 99% percent might take a month, and then there’s 1% left […] it keeps going until you get to that 0.01% and then it’s hard.”

So what advancements there are have been in this AI-powered technology?

Heralded as the “future of driving”, Tesla’s Autopilot feature promises full self-driving capabilities in the near future thanks to a combination of cameras, ultrasonic sensors, and radar. In order to achieve this remarkable feat, Tesla uses a neural network that they’ve trained to emulate human driving at its best, using data from over three billion miles driven. Today, Autopilot enables the car to steer, accelerate, and brake automatically within its lane as well as be summoned to its driver within a parking lot. A new upgrade allows for lane changes and taking highway interchanges or exits based on the destination. Via over-the-air software updates, Tesla is able to incrementally improve its Autopilot feature over time.

Google’s self-driving project, now called Waymo, takes a different approach, first mapping out detailed maps with information such as road profiles, sidewalks, lane markers, traffic lights, and stop signs. Instead of integrating cameras and sensors into its vehicles, it instead adds them on to the cars it uses. Waymo’s software has driven billions of miles in simulations using AI-generated images and is currently operating two services in both testing and limited stages: Waymo One — an autonomous vehicle taxi service — and Waymo Via, a transporting service.

Other companies such as Daimler, Embark, and TuSimple have already put driverless-equipped trucks on the road for testing. Tesla plans to produce a fully electric autonomous truck called the Semi which provide enough fuel savings to pay back the entire cost of the truck in only two years.

The autonomous transportation industry experiences perhaps the most hype and it’s reflected in soaring stock prices and ballooning company valuations. However, the potential that exists in AI-fuelled self-driving vehicles and the impact they could have is — within all likelihood — going to be made reality in the coming decades.

Photo by Aaron Burden on Unsplash

What Does This Mean for Us Humans?

Many people fear that the rise in AI-empowered industries will lead to a swift decline in jobs for humans. Take for example that more than 3.5 million people work as truck drivers in the US alone — what will happen to their jobs once autonomous trucking reaches the mainstream? What about radiologists whose tumour-identifying role will be taken over by superior machine learning systems?

If you foresee a future where human job prospects are grim and ever-dwindling then you’re only looking at one side of the coin. Sure, there will be plenty of jobs that will pretty much disappear, but it won’t happen overnight. On the flip side, the amount of jobs that will be created could bring about even more progress and advancements in technology — just imagine what a million more computer scientists or engineers could do for the world.

The changes that will be brought about by AI in the coming years and decades only stress the need for current and future generations to be multi-skilled, having the invaluable ability to learn and then re-learn in a continuous cycle that will last a lifetime. If you’re interested in this art of reinvention, I encourage you to read this article by historian and writer Yuval Noah Harari.

Photo by Benjamin Davies on Unsplash

Summary

In short, AI will revolutionize the world as a whole in the coming 20 years, changing not only the systems we interact with, but even our everyday lives. We had a look at three key industries to give us a glimpse of the future:

• Healthcare: getting insights into different diseases with big data, diagnosing more accurately with machine learning, using personalized medicine to tailor treatment, and accelerating drug discovery
• Finance: promoting good credit for individuals, managing risk and detecting fraud for larger companies, and offering personalized banking
• Transportation: reducing the thousands of fatal accidents on the road and saving people time

Whatever change does happen over the next two decades, you can be sure that it’s going to be an exciting and rapidly-changing time. Humanity is going to need a new generation of problem-solvers, creative thinkers, and doers to overcome the greatest challenges in the history of our species. The only thing that we know for certain in the future is that change itself is the only certainty.

I hope you’re ready.