WTF is Artificial Intelligence Anyways?
Quick note before you dive in:
This post (largely based on recent McKinsey Global Institute and Executive Office of the President reports) includes some doom and gloom about the negative impacts artificial intelligence may have on your life and career.
The last thing I want is for you to leave feeling freaked out, so try reading to the end, where I discuss the potential AI has to change your work for the better. If you can’t make it that far, save it for later or share it with a friend so you remember to read it the next time you hear about AI or automation in the news.
Either way, you’re taking a step towards preparing yourself and a friend to be more successful in the future. You rock!
WTF Is Artificial Intelligence Anyway?
The better you understand what artificial intelligence is, the better you’ll understand whether or not your job is in danger of being automated. Unfortunately, the confusion (and intentional misuse) around the term has made life hard for everyone. Ben Thompson recently explained what makes artificial intelligence so difficult to define.
First, there are two types of artificial intelligence: Artificial General Intelligence, that is, a computer capable of doing anything a human can. That is in contrast to Artificial Narrow Intelligence, in which a computer does what a human can do, but only within narrow bounds. — Ben Thompson, Stratechery
Treasury Secretary Steven Mnuchin was the laughing stock of many Twitter feeds recently when his lack of understanding of AI was put on display. He said he was not at all concerned about artificial intelligence automating jobs, but that self-driving cars will be shuttling us from coast to coast in the near future.
“In terms of artificial intelligence taking over American jobs, I think we’re so far away from that that it’s not even on my radar screen. I think it’s 50 or 100 more years. That to me isn’t artificial intelligence, that’s computers and using real technology we have today. But those types of things are very real. That’s very different from artificial [intelligence], you know, R2-D2 taking over your job.” — Steven Mnuchin, Treasury Secretary
The statement is scream-worthy because millions of truck and taxi driving jobs will be eliminated by the same self driving technology (enabled by narrow artificial intelligence) that he thinks isn’t too far away. They will be among the 38% of US jobs reported to be at high risk of being automated within the next 15 years.
It will not take a robot that is better than humans at all intellectual tasks (Artificial General Intelligence) to eliminate these jobs. It will only take a form of existing AI-enabled software, that is better than humans at one or two tasks, like driving from place A to place B, to be introduced into the market.
So, while 15 years may seem like a lot of time, the truth is many work activities that you or someone you love spends most of their time doing, will be automated within 1 to 2 years. There are less protections and regulations in place to keep you from feeling the impacts of AI-driven automation than there are to protect entire industries across the US Economy.
How AI Will Automate Certain Jobs Out of Existence
“By learning from a greater volume of information than we [humans] can process in our own lifetimes, AI software gives us the ability to reach new heights when solving complex problems. AI shows us that today’s state-of-the-art solution is no longer a global maximum, but in fact only a local maximum.” — Nathan Benaich, AI Investor
The AI-related stories landing in your Facebook and Twitter feeds are mostly about doomsday Artificial General Intelligence scenarios. While interesting, they distract from the more pressing issue at hand.
AI-driven automation will create new jobs and help people to be more productive, but the painful truth is that AI-driven automation of both specific work activities and entire jobs will be incredibly disruptive to hundreds of millions of people around the globe. The McKinsey Global institute and The Obama Whitehouse, respectively, believe that “60% of all occupations have at least 30% of activities that are technically automate-able,” and “47% of U.S. jobs are at risk of being replaced by AI technologies” over the next 10–20 years.
When a company stands to gain enormous financial benefits from automating a certain work task, and said task is automate-able with current artificial intelligence techniques, you can expect for it to be automated quickly. If a job consists mostly doing that single task, you can expect people to be replaced, rather than complemented, by AI software. While automation is incredibly painful for us who lose jobs, it can simultaneously improve the quality of life, and in some cases save, the lives of hundreds of millions of people.
Here are the activities, jobs, and industries that experts predict to experience the impacts of AI-driven automation in the next 5 years…
Recognizing known patterns
If you, or someone you love, has ever suffered from a medical issue that went undetected by doctors, you know the importance of good injury and disease detection. Diseases are a class of “patterns” that an AI algorithm could help humans recognize.
In the US alone, there are ~38,000 radiologists that make an average of $490,000 annually. According to recent FDA statistics, these radiologists look at 39,275,011 mammograms annually to detect breast tissue abnormalities that require further examination. The UK’s National Health Service recently showed that standard breast cancer checks aren’t sensitive enough to detect ~17% of cases. This is why people were so excited by Google’s recent announcement that it had developed an algorithm to “flag [potential breast cancer] that a human will miss.”
“The algorithm helps you localize and find these tumors. And the doctor is really good at saying, ‘This is not cancer.’ The technology will be especially useful in parts of the world where there’s a shortage of physicians. For patients who don’t have access to a pathologist, an algorithm — even if imperfect — would be a meaningful improvement.” — Matt McFarland, Washington Post
If the technology makes it out of the lab and into hospitals, many thousands of lives and millions of dollars could be saved by early diagnoses and treatment for patients — when procedures are most effective. While the opportunity to automate this skill is massive, it’s more likely that an AI will complement a radiologist than automate her out of a job.
Unlike other jobs that are largely composed of a single task, radiologists have many responsibilities. In addition to recognizing patterns in medical images, they are tasked with consulting physicians to direct patients’ care and working with doctors from different fields to decide on additional treatments to be considered. Neither of these are activities likely to be automated soon, meaning that a machine will augment radiologists long before it puts them out of work.
Driving Cars and Delivering Goods
Morgan Stanley predicts that self-driving truck technology could save the freight transportation industry $168 billion annually and Boston Consulting Group predicts that self driving cars could create a $42 billion market by 2025.
Demonstration of a self-driving Tesla, which relies partially on computer vision to detect objects on the road. Source
Freight companies will save $70 billion annually by putting many of the 1.6+ million people who drive heavy trucks in the US out of jobs. And, they’re estimated to save $36 billion from reduced accidents: 3,852 people died in large truck crashes in 2015 alone.
The Obama White House projected that the jobs of ~1.4 million (taxi, bus, self-employed, etc.) drivers are threatened by autonomous vehicle technology. Researchers also estimate that self-driving cars will reduce traffic fatalities by ~90%. By the 2015 numbers, that’s nearly 1,125,000 lives saved worldwide in a year, 11,250,000 saved over the course of a decade, and 56.3 million fatalities prevented in a half-century. To put that in perspective, in a single year, autonomous vehicles would save the combined populations of Fiji and The Bahamas, the total population of Belgium in 10 years, and the combined populations of South Africa and Botswana in 50 years.
“If it were true that the algorithms are demonstrably, measurably, statistically better than a human driver, then we should not let human drivers on the roads. If you wanted to drive, go to Lego Land….driving can be a fun recreational activity, we just don’t need you on the road.” — Frank Chen, Andreessen Horowitz
The risk of AI automating driving and delivery jobs is so high because of the clear cost (and life) savings opportunities for companies that find a way to remove humans from the picture. The major challenge for society will be training the drivers for new work. So far, much talk has been about skills being automated in lower wage professions, but skills will also be automated in high paying professions (like the radiologist example).
Much has been written about Amazon using robots, that rely on narrow AI, to dominate its competition and cut warehouse operating expenses by 20% (that’s many billions of $). The little robots speed across fulfillment center floors, lifting heavy goods and bringing them to Amazon’s human workers so they don’t have to waste time walking in search of a product. It’s not hard to imagine similar kinds of robots finding their way into other industries.
Major US companies in the waste industry, for example will be incentivized to replace the (as of May 2015) 48,620 waste collectors who earn $34,610 annually with AI-enabled robots that reduce costs to make their biggest revenue center even more profitable (how fast the industry seizes the opportunity is another question). Replacing humans in this role could benefit the environment through more efficient disposal of waste and pick up routes. It would also limit the amount of backbreaking, thankless work done by garbagemen and women — but many people will lose their jobs in the process.
Zume Pizza has even built robots with narrow AI that make pizza.
Searching for and gathering information
AI-driven automation won’t only be felt in faraway truck yards, but also in the office building you’re likely sitting in right now. Every year, knowledge workers spend at least 600 million human hours searching for and gathering information, an activity that a machine can do better (faster and more efficiently).
We now have a methodology to automate people in these [white collar] roles. What this means is that if I have a company, I may not fire people — companies tend to try to minimize firings. But I may dramatically slow the rate at which I hire new people and instead invest in automation. Ultimately, this leads to fewer job opportunities in the long term in these areas. — Jack Clark, OpenAI
You can imagine the excitement of a manager who is told that narrow AI makes it possible for software to show her salespeople the information they need, improving their productivity by allowing them to instead spend those 2 hours they lose every day searching, to an activity that is both far more valuable for the business and for narrow AI software to automate — helping a new customer to have a better experience with the product they’re being convinced to pay for.
Like the radiologist whose job has layers beyond spotting abnormalities in a medical image, the salesperson (and many other types of knowledge workers) is more than her ability to track down information. Therefore, she’s more likely to be complemented by AI than automated — more on this in the next section. Demand for analysts for example, who are tasked primarily with conducting research, however will likely shrink as a “blend of machine learning algorithms and distributed labor” can replace much of their work. (“Machine Intelligence Will Let Us All Work Like CEOs”)
Most knowledge workers spend less than half of their time doing things they’re really good at (i.e., what they’ve been hired to do). The rest is spent doing research, arranging meetings, coordinating with other people, and performing other minutia of office life. These tasks could be done just as well by a machine intelligence service. — Shivon Zilis, Bloomberg Beta
The truth is, impacts of AI-driven automation will be both positive and negative. While some jobs are particularly susceptible to being fully automated and not having any need for a human in the loop, it’s more likely that specific activities will be automated, without taking the rest of the job with it. Only time will tell if our society will find a way to employ the people who suffer most from automation.
It’s not worth devoting time and energy to mastering the above activities. AI-software will be better than most humans it in the very near future. A better use of your time would be to work on the skills below, ones that will be highly complementary to the advancements made possible by AI.
And what are those skills, you ask?
How to Thrive in An Automated Workplace
One of the major promises of AI is freeing people from mindless tasks, so they can do more meaningful work.
The introduction of AI enabled technologies into our work will change all of our jobs and industries, but tomorrow’s most effective individuals won’t be very different than today’s.
She will bring new ideas to the table, communicate effectively throughout and organization, use logical reasoning to be more convincing, and navigate the social and emotional waters of the company. While AI will be able to recognize known patterns, navigate autonomously, gather information and move objects, it won’t be able to think creatively, empathize, or logically reason and solve problems: a successful employee will.
Candidly, guidance in this section will be most helpful for knowledge workers who are fortunate enough to find themselves in the positions that already pay a premium for thinking. No one can predict all of the ways that narrow AI will change knowledge work. To position oneself to be augmented, rather than replaced by AI, one should embrace the benefits of AI enabled technology and invest in the “soft” skills that will empower her to stand out as an adaptable, personable and multi-faceted employee.
For many, these skills will be difficult to develop. However, they are worth building now, as they are unlikely to reach median level human performance before 2040 (“Harnessing Automation for a Future that Works”). Those who do develop the skills will be complementary to the new AI-powered technologies that are guaranteed to arrive in our workplace in the next few weeks, months and years.
Generating creative new ideas
As AI software automates certain white collar activities, successful employees will separate themselves from others by investing their time in the most impactful activities.
In an earlier section, we alluded to Jack Clark’s point that while certain white-collar jobs might not disappear due to automation, the demand for them may decrease as companies invest in automation rather than hiring. For example, the analyst we mentioned earlier, tasked primarily with conducting research, may find her work less valued if she simply follows instructions sent down to her by a manager.
The industrial revolution freed humanity from much repetitive physical drudgery; I now want AI to free humanity from repetitive mental drudgery, such as driving in traffic. — Andrew Ng, Founder of Google Brain
However, thanks to AI software that could gather information and give her 20% of her time back, the analyst can use her surplus time more effectively. Equipped with an holistic understanding of her business (that AI wouldn’t possess), the successful analyst would dedicate her time to identifying problem and opportunity areas for the business that hadn’t been previously considered. As is true today, her insights will be most valuable if she can effectively communicate them to other people throughout the organization.
Actionable 1st step: Build idea generation muscle
I promise this is worth the 25 seconds.
Despite recent advances, we humans are still much better than machines at understanding language and all the nuances of human interaction. If you’ve ever used Amazon’s Alexa, a Google Home, or Siri, you’re aware of these limitations.
An employee who packages her insights in way that others can understand and act on will always be a valued asset to her team. As narrow AI automates more rote tasks, the effective employee will dedicate more of her time to using her social and emotional capabilities to develop others on her team.
Actionable 2nd step: Improve your emotional intelligence and ability to communicate
Logical reasoning & problem solving
Until AI systems are able to explain in detail how they arrived at certain predictions, effective employees can separate themselves from the pack by logically explaining their recommendations.
There is much work being done on developing technology that will help people to interpret the steps a narrow AI program took to arrive somewhere. But today, technologists must make a trade off between the accuracy one would get by solving a problem with a deep learning technique versus the the transparency that could be realized by solving a problem with a regression analysis.
AI systems aren’t able to solve problems in organized ways with contextual information nearly as well as humans can, and humans also tend to over-respect reason. An individual who uses logic to solve problems and explain her solutions will be more likely to gain trust and respect than those who are not.
Actionable 3rd step: Learn about and apply mental models as needed
An employee who finds a way to master her core job function, and masters the soft skills, she’ll become a strong candidate for augmentation, not replacement, by narrow AI software.